Ramped SSTC - Power Supply Question + General Improvements

[ZakW]

Hello Everyone,

Recently I have started working on my ramped SSTC again and have encountered a similar issue as before. See thread https://highvoltageforum.net/index.php?topic=1878.msg14063#msg14063

I would like to scope the bridge during operation, so I am attempting to power the bridge from a transformer instead of mains or my variac. [Variacs ARE NOT isolated, learned that from before ]

I do not currently have a transformer with sufficiently high output voltage to power the bridge (plus I had issues before), so I attempted to create an isolated mains supply using two UPS transformers.

Two identical (tested) UPS power supply transformers, see attached picture for schematic.

Normal operation:
Primary: ~8.5v AC
Secondary:120v AC

Variac or mains -> T1 120v AC winding -> 8.5v AC output -> T2 8.5v AC -> ~120v AC output

I know there will be some losses, and I will not get a perfect 120v output but that is fine.


The problem:

When powering the bridge from the isolation transformer setup the coil output is very weak. The arcs are thin and light blue, only about an inch long. When connected directly to mains or the variac they are about 12 inches.

Previously, Dave mentioned issues related to transformers not working too well with halfwave rectification, which is what I am currently using. I tried connecting the output to a full bridge rectifier instead but achieved the same results… I can attach scope shots later if that will help. Connecting the isolation transformer setup to mains gives me 120vAC through the full bridge rectifier. I don’t think low voltage is an issue.



Additional notes:

The transformers DO NOT hum or get hot during operation.
In an attempt to simulate a load on the transformers I wired in a 60w bulb. It did not make a difference.

My ideas so far:

I have only noticed this issue with this particular mains-synced type of coil (this is currently my 3rd iteration) so it is not limited to just this version. Being that it is synced to the mains cycle, I suspect that is part of the issue, but I do not know enough about transformers to understand why my variac works without issues but a step down or isolation transformer will not work.

Could it be that the type of rectification is not the issue but rather some sort of phase lag between the transformer secondary voltage and current? I am using an optocoupler to sense the zero crossing of the AC voltage. From what I understand in normal conditions voltage and current are synced unless it is a reactive/inductive load. If that is case, why wouldn’t my variac be having the same issue?

Thanks in advance!
-Zak

[davekni]

Could it be that the type of rectification is not the issue but rather some sort of phase lag between the transformer secondary voltage and current? I am using an optocoupler to sense the zero crossing of the AC voltage. From what I understand in normal conditions voltage and current are synced unless it is a reactive/inductive load. If that is case, why wouldn’t my variac be having the same issue?

If you are running a low duty cycle, say one half-cycle every 10 cycles, issue with half-wave rectified transformer output will be reduced.  Transformer magnetization can recover during the intervening cycles.

Might be something about phasing of your line-synced interrupter and and loaded transformer output phase.  If so, must be a subtle effect, perhaps delaying startup.  Or something changing round capacitance affecting antenna feedback.

Idea for scoping:  Wrap one turn tightly around middle of primary winding with twisted-pair leads (or coax) out to scope probe.  That will allow approximate scoping of primary voltage independent of half-bridge grounding.  (Will include some component of secondary current too.  Still good for comparison purposes.)   Compare that single-turn voltage between functioning and problematic runs.  If you have enough scope channels, simultaneously scope interrupter output and line voltage for reference.
Other scoping option is to add a current transformer (with burden resistor on CT output) to primary lead (half-bridge output).

BTW, after looking at the driver circuit from your other thread again, realized that the spikes you showed there after oscillation stopped may be from the upper left 555 timer used to inject startup pulses into antenna input.

[ZakW]

Hello again Dave, thanks for taking a look!

If you are running a low duty cycle, say one half-cycle every 10 cycles, issue with half-wave rectified transformer output will be reduced.  Transformer magnetization can recover during the intervening cycles.

I normally run the bps on the lower end (~1-1.5/sec) during operation.

Might be something about phasing of your line-synced interrupter and and loaded transformer output phase.  If so, must be a subtle effect, perhaps delaying startup.  Or something changing round capacitance affecting antenna feedback.

If it makes a difference I am using a CT for this build and not an antenna.

Idea for scoping:  Wrap one turn tightly around middle of primary winding with twisted-pair leads (or coax) out to scope probe.  That will allow approximate scoping of primary voltage independent of half-bridge grounding.  (Will include some component of secondary current too.  Still good for comparison purposes.)   Compare that single-turn voltage between functioning and problematic runs.  If you have enough scope channels, simultaneously scope interrupter output and line voltage for reference.
Other scoping option is to add a current transformer (with burden resistor on CT output) to primary lead (half-bridge output).

I will give this a shot today after work. I have 3 channels I can test with. For the 'line voltage reference' can I use a small step down transformer since I do not own a differential probe? I have a 15v AC transformer that I use for comparing the optocoupler output against.

What I plan on testing:
Test 1 - powering with isolation transformer: Scoping a single turn on the middle of the primary, the interrupter output, and AC input (isolation transformer).
Test 2 - powering with variac: Scoping a single turn on the middle of the primary, the interrupter output, and AC reference voltage (from small AC transformer)...

BTW, after looking at the driver circuit from your other thread again, realized that the spikes you showed there after oscillation stopped may be from the upper left 555 timer used to inject startup pulses into antenna input.

You could be right, if I recall I think it might have been related to an incorrect resistor value on the 74hc14 to kick start the oscillation. It was too low of a frequency and was overpowering the CT feedback from the secondary. I have since wound an 80T CT using an N87 type core and removed the resistor between pins 1 & 2. I will have to double check to see if those spikes are still there.

Another mystery I solved were the weird bridge scope measurements I was getting https://highvoltageforum.net/index.php?action=dlattach;topic=1878.0;attach=13192;image. Turns out I must have damaged/killed my 100x HV probe from probing the bridge when it was connected to my variac. Testing the probe later I found it was not working properly.

I will report back later with my results.

[davekni]

If it makes a difference I am using a CT for this build and not an antenna.

I have since wound an 80T CT using an N87 type core and removed the resistor between pins 1 & 2. I will have to double check to see if those spikes are still there.

Do you mind sketching and posting your input circuit (coupling cap, protection diodes w/type, etc., up to HC14 input)?  I personally prefer current feedback as you now have.  There are one or two problematic schematics for such that are sometimes copied, resulting in issues.  If you still have a capacitor between CT and protection diodes, it should be fine.

What I plan on testing:
Test 1 - powering with isolation transformer: Scoping a single turn on the middle of the primary, the interrupter output, and AC input (isolation transformer).
Test 2 - powering with variac: Scoping a single turn on the middle of the primary, the interrupter output, and AC reference voltage (from small AC transformer)...

Sounds great.  The transformer powering optical coupler is lightly loaded, so would be fine for scoping instead of yet another transformer.

Do you have a resistor across H-bridge output or across cap from H-bridge to primary?  If not, that could be causing startup unreliability.  If H-bridge output happens to already be in the state that the HC14 also happens to be in when enable starts, there is no initial current to trigger oscillation.

[ZakW]

Do you mind sketching and posting your input circuit (coupling cap, protection diodes w/type, etc., up to HC14 input)?  I personally prefer current feedback as you now have.  There are one or two problematic schematics for such that are sometimes copied, resulting in issues.  If you still have a capacitor between CT and protection diodes, it should be fine.

I am using ceramic caps. 

Test 1 - powering with isolation transformer: Scoping a single turn on the middle of the primary, the interrupter output, and AC input (isolation transformer).
Y = AC Input Probe is connected after single diode for halfwave rectification
B = Interrupter output


Y = AC Input (before rectification)
B = Interrupter output


Y = AC Input (before rectification)
B = Interrupter output


Acquire mode set to 'Average'. Seemed to clean up the signal a bit to make it easier to see. Probably not very accurate.


B = Opto output. You can see that it is NOT crossing zero at the correct time.


Purple = UCC output. You can see that the coil is trying to start well into the positive cycle. I have accounted for this delay in


The above opto scope shots are from this design.


Here is the opto output after following the design below. My original idea was to use this design since the 560nf cap + 100 resistor was a lot more efficient and did not require a large resistor to dissipate a lot of energy. However, I did not realize this would cause a phase delay(shift?) in the AC signal. Messing around with a simulation in Falstad I was able to come up with a design to compensate for the phase shift which seemed to work well but was sensitive to the AC input voltage. Cutting out if the voltage fell too low.



Test 1 Notes:

- I made one turn around the primary and twisted the wire together before connected it to my probe. Not sure if there was such little current flowing but after setting the scope Volt/Div to 500mV it was still a very small signal. It looked just like what I see when I hang a probe near the scope during operation.
- While being powered from the isolation transformer setup the output of the coil was extremely weak. The arcs were white/light blue almost not visible and the sound was very snappy, loud, and sharp.
- Seems like the voltage on the transformer drops to almost zero when the coil turns on...
- Adjusting the BPS did not seem to make a difference
- Adjusting the interrupter duty cycle made the arc pop a bit bigger every once in awhile but it was not consistent.

If it helps I can take a video of the output during operation. My main goal is to test the bridge under while it is close to normal operating voltage. I doubt there is something wrong with any of the AC transformers I am using but instead a fundamental issue with how I have the coil setup to operate. Having to sense the zero crossing with an optocoupler etc.
-------------------------------------------------

Test 2 - powering with variac: Scoping a single turn on the middle of the primary, the interrupter output, and AC reference voltage (from small AC transformer)

Since the 'single turn on the middle of the primary' signal was so weak I removed it and did not test under normal operation. I can try again when I have more time after work (not enough hours in the day).

Do you have a resistor across H-bridge output or across cap from H-bridge to primary?  If not, that could be causing startup unreliability.  If H-bridge output happens to already be in the state that the HC14 also happens to be in when enable starts, there is no initial current to trigger oscillation.

No resistor. It is configured just like the schematic shows. J4-Output is going to my primary of course.

Here are a couple pictures of the driver board and coil for reference. I wound a new small GDT on a new core and I have not trimmed the wires short since I was just testing it. Please cut me some slack 





Next steps:

- as I have time I want to post in more detail my ZCD circuit + simulation data as well as the interrupter circuit incase they are to blame for the strange behavior. Also, my gate drive voltage & wave form are not ideal but that is another issue for another time. 

Falstad simulation link https://tinyurl.com/2l525bnj



For now this is all I have. Enjoy this video of a full power test run

Connected to my variac, tuned up the ZCD. 15in arcs from my 1.6in coil.

[davekni]

Wow, lots of information!  First, thank you for the CT input schematic.  Looks fine.  R1 is larger than in other designs I've seen.  May make primary voltage lead secondary current some, depending on CT inductance.  Some lead can be helpful to compensate for delays through circuitry.  SSTCs don't typically add intentional phase lead.  Since your coil is working well on variac output, probably no reason to change this.
As with most CT and antenna input circuits, the initial state of HC14 is indeterminate.  That combined with an indeterminate state of initial bridge output can cause startup issues.  At interrupter lead edge, bridge may be driven to same state it is already in, preventing any initial half-cycle to trigger oscillation.  That's the use of a resistor across bridge output.  Causes initial state to be centered, so there will be some initial output no matter which direction bridge is driven initially.

- Seems like the voltage on the transformer drops to almost zero when the coil turns on...

Yes, I think that is accurate.  Transformer impedance is much too high for the load your coil presents.  You will need a isolation transformer capable of your coil power draw to be useful.

I doubt there is something wrong with any of the AC transformers I am using but instead a fundamental issue with how I have the coil setup to operate.

Your scope traces show the opposite.  AC transformers are too small (too low power) to be useful here.

Since the 'single turn on the middle of the primary' signal was so weak I removed it and did not test under normal operation. I can try again when I have more time after work (not enough hours in the day).

I suspect the main reason signal is weak is because power to coil is weak.  Will be valuable to see that signal under normal operation.
However, there is still something odd about the single-turn scope captures you do have.  Looks like a DC shift when coil is operating (during enable pulse).  Single turn of wire should not be able to sustain such a voltage.  Perhaps the actual waveform has short higher-voltage spikes that are over-ranging scope input.  Might be worth increasing sample rate (zooming in horizontally) to the beginning of enable pulse to see if such pulses exist.  A picture of your single-turn and connected scope probe would be useful too.

Here are a couple pictures of the driver board and coil for reference. I wound a new small GDT on a new core and I have not trimmed the wires short since I was just testing it. Please cut me some slack 

Did you intend to say "new small CT"?  Lead length is not an issue for CT secondary.  GDT appears to be soldered to ECB with short leads.  Or is that GDT not connected now?

Concerning ZCD, I'd suggest using something more like your original circuit.  Use a transformer output to reduce power dissipation, such as 12VAC.  Reduce resistor to 330 ohms or however low your opto LED input can handle (max LED current rating).  Then add just a small capacitor and series resistor across the main 330 ohm resistor to adjust zero-crossing earlier if needed.

Connected to my variac, tuned up the ZCD. 15in arcs from my 1.6in coil.

Nice performance!

[ZakW]

Wow, lots of information!  First, thank you for the CT input schematic.  Looks fine.  R1 is larger than in other designs I've seen.  May make primary voltage lead secondary current some, depending on CT inductance.  Some lead can be helpful to compensate for delays through circuitry.  SSTCs don't typically add intentional phase lead.  Since your coil is working well on variac output, probably no reason to change this.
As with most CT and antenna input circuits, the initial state of HC14 is indeterminate.  That combined with an indeterminate state of initial bridge output can cause startup issues.  At interrupter lead edge, bridge may be driven to same state it is already in, preventing any initial half-cycle to trigger oscillation.  That's the use of a resistor across bridge output.  Causes initial state to be centered, so there will be some initial output no matter which direction bridge is driven initially.

You're welcome, I'm happy to provide all the info I can.

Regarding R1 I was following your advice from this thread https://highvoltageforum.net/index.php?topic=834.20. I assumed it was advice on best practice for the input stage of the HC14. If not I can try adjusting R1.

What value resistor could I use on the bridge output? It seems like a good idea to include in my design.

Yes, I think that is accurate.  Transformer impedance is much too high for the load your coil presents. You will need a isolation transformer capable of your coil power draw to be useful.

Your scope traces show the opposite.  AC transformers are too small (too low power) to be useful here.

I am still new to understanding impedance so hopefully I have this right - You're saying that the impedance of the transformer output is too high and it cannot supply the required current to the coil during those instances when it is on? I don't know if impedance correlates to power VA rating of a transformer but these UPS transformers are pretty decent. Googling the UPS model "APC 550" I found 'It yields an average output power capacity of 550VA at 330W and even delivers battery-powered AC from its outlets.' 550VA @ 120v = ~4.58Amps. So while the transformers are rated for such power they are unable to supply the current required during the short on times of the coil?

I suspect the main reason signal is weak is because power to coil is weak.  Will be valuable to see that signal under normal operation.
However, there is still something odd about the single-turn scope captures you do have.  Looks like a DC shift when coil is operating (during enable pulse).  Single turn of wire should not be able to sustain such a voltage.  Perhaps the actual waveform has short higher-voltage spikes that are over-ranging scope input.  Might be worth increasing sample rate (zooming in horizontally) to the beginning of enable pulse to see if such pulses exist.  A picture of your single-turn and connected scope probe would be useful too.

Sorry for the confusion. None of the above pictures contain a picture of the single turn around the primary. I think you are referencing the optocoupler output. Here is a picture of the single turn setup. If this is correct I will grab a few pictures of the waveforms.

Did you intend to say "new small CT"?  Lead length is not an issue for CT secondary.  GDT appears to be soldered to ECB with short leads.  Or is that GDT not connected now?

Concerning ZCD, I'd suggest using something more like your original circuit.  Use a transformer output to reduce power dissipation, such as 12VAC.  Reduce resistor to 330 ohms or however low your opto LED input can handle (max LED current rating).  Then add just a small capacitor and series resistor across the main 330 ohm resistor to adjust zero-crossing earlier if needed.

I meant to say GDT. You have given me a lot of advice on them in the past and from what I can tell you are an advocate for abolishing all stray inductance with short, clean connections. I was testing out a new GDT core https://www.mouser.com/ProductDetail/871-B64290L0697X087 and left the connections to the ECB about 3/4" long instead of cutting it flush to the ECB. On my next board I will have it mounted a lot closer.

What are your thoughts on the GDT core and size? It is N87 material which from what I can find online seems like a good choice. I am also using a similar N87 core for my CT.

Concerning the original ZCD design using a transformer - my goal for this build is to make it as compact as possible. Currently my new ECB design using mostly SMD components is 2.8x2.8in (71mmx71mm). Shrinking down the DC blocking caps and the GDT saved a lot of space.

I have a couple ideas for providing the ZCD but the easiest so far is the optocoupler.

Nice performance!

Thank you! I need to rewind a new secondary. The resonant frequency ended up being a bit too low with this one. First time using epoxy to insulate the coil and it has worked great! this 44awg wire usually pops if there is a flashover. I added 4 coats and I have had very few issues.

Working on uploading pictures below

In this shot I was able to capture a time when the ZCD was not 'tuned' via the 200k POT (RV3). The coil was just snapping with a very weak output then there was one louder pop where it looks like it ran normally for a single arc.

Yellow = HC14 pin4 output || Blue = Opto output || Purple = 12vAC transformer for reference



Normal operation





Now here is a shot focusing on the problem. When the output is very weak. The HC14 is only on for a very short time at the beginning and end of the AC cycle. I do not know what is causing this... It appears to only be on for 85uS?

 || Purple = 12vAC transformer for reference.





Here is a shot of a normal pulse/arc HC14 output (pin 4) and input (pin 1).

Yellow = HC14 pin4 output || Blue = HC14 pin1 input (form CT)



Zoomed




Here is what is looks like when it is weak





Notes:


I see two things happening here:
1. If the output of the ZCD is not tuned correctly the interrupter will not output a signal. If the tuning is close it will output a signal but it will be weak and snappy.
2. The CT feedback on those weaker pulses seems to be dying out causing the coil to shut off.
3. I noticed in some pictures the HC14 output is either pulled high at 5v when it turns on or it is pulled low before turning on. Is that supposed to happen?

Update:

Here is a short video I took explaining the issue. Ideally the ZCD would work regardless of the input voltage (with operating range of the coil at least). As this is my 3rd revision I am constantly testing components and not being able to power the coil from an isolated lower voltage transformer is a big drawback. For the final version I do not plan on adjusting the input at all and just having it connected directly to mains, so tuning wont be an issue at that point. However it is during testing that the ZCD (as it is now) introduces a lot of inconsistency.

I suppose for testing I could have a breakout on the board for two AC inputs. One for the ZCD connected to mains (tuned for that voltage), and the other from a step down transformer so I can at least scope the bridge. The issue remains though that I cannot get the coil to work on any step down transformer I have used so far. My DC supply works but it is limited to 30v. Since using my variac and tuning the ZCD causes a similar issue to occur when using a step down transformer I figured the issues were related but I do not know the cause.

[davekni]

Regarding R1 I was following your advice from this thread https://highvoltageforum.net/index.php?topic=834.20. I assumed it was advice on best practice for the input stage of the HC14. If not I can try adjusting R1.

Oh, I see.  I was editing a secondary CT feedback design that has diodes in the wrong place.  Didn't think much about resistor value.  May be fine, or might be better with a lower value.

What value resistor could I use on the bridge output? It seems like a good idea to include in my design.

High enough to keep power dissipation reasonable.  Low enough to mostly discharge blocking capacitor(s) in series with primary winding between enable pulses.  Given your infrequent enables, a wide range of values should work.

I am still new to understanding impedance so hopefully I have this right - You're saying that the impedance of the transformer output is too high and it cannot supply the required current to the coil during those instances when it is on? I don't know if impedance correlates to power VA rating of a transformer but these UPS transformers are pretty decent. Googling the UPS model "APC 550" I found 'It yields an average output power capacity of 550VA at 330W and even delivers battery-powered AC from its outlets.' 550VA @ 120v = ~4.58Amps. So while the transformers are rated for such power they are unable to supply the current required during the short on times of the coil?

Do you know how much current your coil draws?  Likely well above the ~3A capability of transformers.
One new possibility comes to mind:  Perhaps coil is somehow misbehaving, such as oscillating at resonant frequency of primary and blocking caps, causing excess line current draw.  Perhaps it does this briefly at the start of bursts, and locks to desired secondary frequency if enough current is available to get it past some initial mis-locking (ie. when powered by variac).

Here is a picture of the single turn setup.

Oops, looks like I didn't explain that correctly.  The single turn should go all the way around the primary coil form, not around the wire circumference.  For your two-turn primary, a wire laid into the groove between the two turns (on top of the two turns) would be ideal.  Goal is to make a transformer, 2:1 in this case, from primary coil to added isolated turn.  Should have almost 50% of primary voltage.  Bit less due to coupling factor below 1 and some inductance in primary lead wires.

I meant to say GDT. You have given me a lot of advice on them in the past and from what I can tell you are an advocate for abolishing all stray inductance with short, clean connections. I was testing out a new GDT core https://www.mouser.com/ProductDetail/871-B64290L0697X087 and left the connections to the ECB about 3/4" long instead of cutting it flush to the ECB. On my next board I will have it mounted a lot closer.

Yes, I am a fan of minimal parasitic inductance.  3/4" isn't that critical, but might as well eliminate that eventually.

What are your thoughts on the GDT core and size? It is N87 material which from what I can find online seems like a good choice. I am also using a similar N87 core for my CT.

Yes, that is a common material that will work well.

Concerning the original ZCD design using a transformer - my goal for this build is to make it as compact as possible. Currently my new ECB design using mostly SMD components is 2.8x2.8in (71mmx71mm).

Do you have any remaining 60Hz transformers within your coil build?  If so, output would likely work fine.  That way you don't need the opto.  Just a comparitor and clamp diodes to keep comparitor input within limits.

Shrinking down the DC blocking caps and the GDT saved a lot of space.

Is capacitance of blocking caps lower now?  That would raise primary resonant frequency, making it more likely that oscillation might inadvertently lock to that.

2. The CT feedback on those weaker pulses seems to be dying out causing the coil to shut off.

Perhaps bridge bus voltage is dropping too low.  I notice that frequency is dropping towards the end.  Not sure if that is a cause or symptom, nor why frequency is changing.

3. I noticed in some pictures the HC14 output is either pulled high at 5v when it turns on or it is pulled low before turning on. Is that supposed to happen?

That is what I meant by saying initial HC14 state is indeterminate.  That is why a resistor across bridge can help.  Makes some initial half-cycle current signal no matter which initial state HC14 may be in.

However it is during testing that the ZCD (as it is now) introduces a lot of inconsistency.

Yes, your ZCD circuit looks voltage-sensitive.  That's why I'd recommend a comparitor on transformer output.  Or, there may be suitable commercial ZCD chips.  I know there are ZCD optical-coupled TRIAC gate driver chips.  We used those at work a couple decades ago.

Perhaps the updated single-turn primary scoping turn will allow more scope testing with variac.

Given your half-wave rectification, scoping when variac-powered shouldn't be hard.  Make sure variac is wired correctly (output referenced to neutral) and that your diode is connected to line hot.  Then bridge Vbus- will be at neutral potential independent of variac setting.  Neutral should be close to ground potential.  Of course, measure to verify.  Then connect scope ground clip through a capacitor (0.1 to 1uF) to Vbus- (neutral).  This will ground for high-frequencies without tying neutral directly to safely ground.

[ZakW]

Oh, I see.  I was editing a secondary CT feedback design that has diodes in the wrong place.  Didn't think much about resistor value.  May be fine, or might be better with a lower value.

Looking at Steve's Mini design schematic https://www.stevehv.4hv.org/SSTC5/miniSSTCfnlsch.JPG he does not have any resistors on the input. If I were to test different values with my setup what is their purpose? Is it just protecting the input from too much current? You mentioned phase lead as well.

High enough to keep power dissipation reasonable.  Low enough to mostly discharge blocking capacitor(s) in series with primary winding between enable pulses.  Given your infrequent enables, a wide range of values should work.

Is this correct placement for the resistor (R3)? If so, it would be dissipating mains voltage constantly?

Do you know how much current your coil draws?  Likely well above the ~3A capability of transformers.
One new possibility comes to mind:  Perhaps coil is somehow misbehaving, such as oscillating at resonant frequency of primary and blocking caps, causing excess line current draw.  Perhaps it does this briefly at the start of bursts, and locks to desired secondary frequency if enough current is available to get it past some initial mis-locking (ie. when powered by variac).

I do not know but I would like to figure out the current draw. Will Mad's section on calculating primary current work for a ramped design? If so I can take some measurements of my primary coil and get an approximation. https://kaizerpowerelectronics.dk/tesla-coils/sstc-design-guide/


I also have a miscellaneous CT I salvaged, here are some pictures. I tried searching the markings on it but I was not able to find any data on it. Are unknown CT's worth using or should I invest in a CT clamp probe for my scope?

Oops, looks like I didn't explain that correctly.  The single turn should go all the way around the primary coil form, not around the wire circumference.  For your two-turn primary, a wire laid into the groove between the two turns (on top of the two turns) would be ideal.  Goal is to make a transformer, 2:1 in this case, from primary coil to added isolated turn.  Should have almost 50% of primary voltage.  Bit less due to coupling factor below 1 and some inductance in primary lead wires.

  that makes way more sense! I will try that.

Do you have any remaining 60Hz transformers within your coil build?  If so, output would likely work fine.  That way you don't need the opto.  Just a comparitor and clamp diodes to keep comparitor input within limits.

I do but I want to keep the form factor as small as possible. I understand that the opto/ZCD might be the cause of my issues now and if that is the case I can live with it. I just wanted to make sure I was not doing something wrong with my step down transformers. It sounds like the current draw is too much for them to handle.

I also bought these mini 1:1 transformers to experiment with https://www.electroschematics.com/voltage-sensor/. I have yet to try and feed the output signal into the original staccato interrupter design since mine omits the input diode and transistor. https://www.loneoceans.com/labs/sstc3/schema_sstc3_staccato.jpg. There did not seem to be any noticeable phase delay when scoping the output of the ZMPT101B and comparing it to the output of that small AC xfrm. As long as the load resistor is selected correctly I might be able to use these to sample the zero crossing when testing at different voltages as well as mains.

Is capacitance of blocking caps lower now?  That would raise primary resonant frequency, making it more likely that oscillation might inadvertently lock to that.

They are, I went from 1uf caps to .68uf. It is what I had on hand. Calculating the reactance at 400kHz puts them around .59 ohms. While the reactance is low I did not consider unwanted resonance. I can order higher value caps if there is an issue.

Perhaps bridge bus voltage is dropping too low.  I notice that frequency is dropping towards the end.  Not sure if that is a cause or symptom, nor why frequency is changing.

A total guess but could it be due to arc loading? I wanted to wind a smaller coil to increase the resonant frequency so I could use a small top load to help stabilize that arc a bit more. Right now I only have a thin metal disc and the arcs are ~10 times the length of the secondary.

Yes, your ZCD circuit looks voltage-sensitive.  That's why I'd recommend a comparitor on transformer output.  Or, there may be suitable commercial ZCD chips.  I know there are ZCD optical-coupled TRIAC gate driver chips.  We used those at work a couple decades ago.

I looked into to some ZCD chips like the BM1Z series https://www.mouser.com/ProductDetail/ROHM-Semiconductor/BM1Z001FJ-E2?qs=BJlw7L4Cy78Ptf%252B6AVl2XA%3D%3D. They look like they would be a great solution but they are not galvanically isolated and that is a concern.

Are you aware of any phase lead/lag circuits that I could implement to adjust for the latency associated with the opto design?

Perhaps the updated single-turn primary scoping turn will allow more scope testing with variac.
Given your half-wave rectification, scoping when variac-powered shouldn't be hard.  Make sure variac is wired correctly (output referenced to neutral) and that your diode is connected to line hot.  Then bridge Vbus- will be at neutral potential independent of variac setting.  Neutral should be close to ground potential.  Of course, measure to verify.  Then connect scope ground clip through a capacitor (0.1 to 1uF) to Vbus- (neutral).  This will ground for high-frequencies without tying neutral directly to safely ground.

I will give this a try. I popped my IGBT's again last night. I am running a bit low on IGBT's and fuses. I don't want to get ahead of myself before I figure out my current issues. My next steps will be to ensure I am driving and protecting my IGBT's adequately and maybe some assistance on choosing the correct parts.

Thanks for all of your advice so far! This has been a long project and I can't wait to share everything once it is all complete.

[ZakW]

Here are the scope measurements of the gate drive and single turn around the primary.

Yellow trace if connected to Q2 Gate. Connected scope ground to neutral (pin 3 of Q2) via .1uf cap.
Purple trace is the single turn around the primary.










Notes:

- Why does the V_GE voltage rise and then fall again? It seems to be getting dangerously high for the IGBT's. I am using FGA60N65SMD at the moment. https://www.mouser.com/ProductDetail/onsemi-Fairchild/FGA60N65SMD?qs=5kMjoDYFkgiyVEfTbs%2FZRw%3D%3D. I was using FGA3060ADF (https://www.mouser.com/ProductDetail/512-FGA3060ADF) before but moving to 2 turns on the primary was popping them every time.
- I swapped out my 12v regulator for a 15v when I started using IGBT's.
- I am using a 7.1ohm resistor on each gate in addition to an 1N5819 diode. Should I consider using back-to-back zenner diodes to clamp the voltage?
- When moving to IGBT's I also removed the freewheeling diodes since I read the IGBT's have proper built in diodes. Is that a good idea?
- The last image of the Gate waveform looks a lot like the miller effect from Richie's site https://www.richieburnett.co.uk/temp/gdt/gdt2.html

[davekni]

Looking at Steve's Mini design schematic https://www.stevehv.4hv.org/SSTC5/miniSSTCfnlsch.JPG he does not have any resistors on the input.

That circuit is antenna feedback.  Antenna is high impedance, so a resistor isn't required.  The 1N60 diodes he uses are germanium, so have low voltage drop.  If using silicon signal diodes (such as 1N914 or 1N4148), a resistor between diodes and HC14 input should be added to minimize HC14 input current.

If I were to test different values with my setup what is their purpose?

Resistor from diodes to HC14 input is to limit HC14 current (avoid latch-up).  The other input resistor isn't necessary.  CT outputs current.  Diodes need to be capable of handling that current.  With no input resistor, voltage switches as current switches.  Delay through rest of circuit (HC14, driver chip, bridge) make output voltage lag current a bit.  That isn't critical for SSTCs.  Adding the input resistor makes the CT output voltage higher.  CT inductance and that resistance form a high-pass filter, adding phase lead.

Is this correct placement for the resistor (R3)? If so, it would be dissipating mains voltage constantly?

Oops, I forgot you have a half-bridge.  Half-bridge requires two resistors, one across C19 and one across C20.

Will Mad's section on calculating primary current work for a ramped design? If so I can take some measurements of my primary coil and get an approximation. https://kaizerpowerelectronics.dk/tesla-coils/sstc-design-guide/

That appears to be for a full-bridge, so +-320V across primary coil.  Adjust to the peak voltage for your coil.  Is your input 120VAC?  In that case peak is 170V, so +-85V into primary coil.  Also, that impedance formula is for sine waves.  With square wave input, inductor current is a triangle wave, with peak current PI/2 times higher.  Voltage drops through IGBTs etc. will make actual peak current a bit lower than that theoretical value.  Yes, I'd recommend doing this calculation.

I also have a miscellaneous CT I salvaged, here are some pictures. I tried searching the markings on it but I was not able to find any data on it. Are unknown CT's worth using or should I invest in a CT clamp probe for my scope?

I've collected some similar transformers from scrapped power supplies at work.  They are used for current sense of switching FETs/IGBTs on the line-side, PFC or main converters.  Usually in series with source or drain lead.  Output runs through diode.  That way when current drops to zero, voltage reverses (secondary diode reverses), generating a voltage spike as needed to reset magnetic flux.  Usually 100 turns on core and single turn wire passing through center.
At your 400kHz these should be good current transformers.
Another possible use is for ZCD.  Wind a few turns through core and connect to scope.  Then run AC line (or xformer output for testing) through resistor and then through built-in 100T coil.  Scope output vs line voltage.  If current is high enough (resistor value low enough), you will see voltage spikes at zero-crossings on the few turns you wound.  For a real circuit, adjust turn count to properly drive a BJT Vbe or FET Vgs.  (Core saturation is what converts input sine wave current into voltage spikes.)  I've used this for ZCD before.

I do but I want to keep the form factor as small as possible. I understand that the opto/ZCD might be the cause of my issues now and if that is the case I can live with it.

Are you aware of any phase lead/lag circuits that I could implement to adjust for the latency associated with the opto design?

You got me thinking about ZCD circuits.  Here's on option I designed/simulated this evening.  Generates zero-crossing current pulses to opto LED.  Circuitry on opto output needs to extend that trigger pulse to quarter-cycle or however long you want.  It is an option instead of the above use of tiny CT backwards.  Uses less current than CT likely requires, only 1mA RMS for what I simulated.  Simulated at 120Vrms (170Vpeak) and at 35Vrms (50Vpeak).  LED current drops, but should still be sufficient for opto.  Very little timing change.
This circuit detects line voltage rising edges, not both edges.
BTW, no idea why I'm suddenly getting attachments moved to the end.  Seems to be happening to other people too.

 [ You are not allowed to view attachments ]

They are, I went from 1uf caps to .68uf. It is what I had on hand. Calculating the reactance at 400kHz puts them around .59 ohms. While the reactance is low I did not consider unwanted resonance. I can order higher value caps if there is an issue.

Once you calculate primary inductance, also calculate primary resonant frequency.  See how close that is to 400kHz.

Here are the scope measurements of the gate drive and single turn around the primary.

Much more like expected.

- Why does the VGE voltage rise and then fall again? It seems to be getting dangerously high for the IGBT's. I am using FGA60N65SMD at the moment.

Two answers.  First, Mads points out that IGBTs usually don't die until ~80Vge.  People here often aim for +-24V to get more current capability, with spikes hitting 30V and sometimes higher.  For SSTC I wouldn't intentionally drive above ~+-15Vge, but no damage for doing so.
Second, that rise and fall is probably an artifact of scoping a neutral-referenced circuit with a ground-referenced scope.  Try scoping neutral.  It probably has that same curve.  Probably the neutral wiring (in your house) voltage drop due to coil current.

- I swapped out my 12v regulator for a 15v when I started using IGBT's.
- I am using a 7.1ohm resistor on each gate in addition to an 1N5819 diode. Should I consider using back-to-back zenner diodes to clamp the voltage?

Your Vge waveform looks great.  No reason to change drive.  Back-to-back zeners are a good idea for any unexpected anomalous situations.  (Bidirectional TVS diodes are the same thing, designed specifically to handle such transients.)

- When moving to IGBT's I also removed the freewheeling diodes since I read the IGBT's have proper built in diodes. Is that a good idea?

Yes, IGBT diodes are designed for freewheeling.  No reason to have external diodes too.  (Unless internal diodes are very slow, and external ones have enough lower voltage drop to take current instead of internal ones.  Unlikely.)

- The last image of the Gate waveform looks a lot like the miller effect from Richie's site https://www.richieburnett.co.uk/temp/gdt/gdt2.html

Yes, exactly as expected.  Notice that the falling-edge flat spot is missing.  That is because diodes bypasses 7.1 ohm gate resistor.  Speeds up turn-off exactly as intended.

Great work with nice detailed measurements!


Edit:  Here's a bit fancier version of line zero-cross-detect.  Adds a second transistor to make LED current turn-on latching.  Tweaked other values too (bit higher line current, 1.6mA).  Now LED current pulse is constant and delay changes only 46us from 35Vrms to 120Vrms input.  Initial version is probably just fine for your use.  Adding this one just to satisfy my perfectionist tendencies:)

[ZakW]

That circuit is antenna feedback.  Antenna is high impedance, so a resistor isn't required.  The 1N60 diodes he uses are germanium, so have low voltage drop.  If using silicon signal diodes (such as 1N914 or 1N4148), a resistor between diodes and HC14 input should be added to minimize HC14 input current.

Got it, that makes sense. I forgot the Mini was antenna based.

Resistor from diodes to HC14 input is to limit HC14 current (avoid latch-up).  The other input resistor isn't necessary.  CT outputs current.  Diodes need to be capable of handling that current.  With no input resistor, voltage switches as current switches.  Delay through rest of circuit (HC14, driver chip, bridge) make output voltage lag current a bit.  That isn't critical for SSTCs.  Adding the input resistor makes the CT output voltage higher.  CT inductance and that resistance form a high-pass filter, adding phase lead.

As a quick test I bridged R1 (1K) and did not notice any change in performance. I might just keep it as is but make a note of it incase it cases issues.

Oops, I forgot you have a half-bridge.  Half-bridge requires two resistors, one across C19 and one across C20.

That is what I was thinking as well. I wasn't sure how a single resistor was going to work. I am not really sure how to calculate the wattage or resistance. At mains (120vAC) Each cap is charged to +- 85v, right? Am I trying to discharge each cap after every pulse or only once the coil is power off completely?

That appears to be for a full-bridge, so +-320V across primary coil.  Adjust to the peak voltage for your coil.  Is your input 120VAC?  In that case peak is 170V, so +-85V into primary coil.  Also, that impedance formula is for sine waves.  With square wave input, inductor current is a triangle wave, with peak current PI/2 times higher.  Voltage drops through IGBTs etc. will make actual peak current a bit lower than that theoretical value.  Yes, I'd recommend doing this calculation.

Thanks for clarifying, I will give it a shot and see what I come back with.

BTW, no idea why I'm suddenly getting attachments moved to the end.  Seems to be happening to other people too.

I have always had this issue. The only workaround I have found so far is to upload the photos as attachments and then use the URL of the photo from the bottom and add it to the thread where I want using the "Insert Image" button. [img] URL HERE [//img]

Once you calculate primary inductance, also calculate primary resonant frequency.  See how close that is to 400kHz.

will do, I will report back.

Two answers.  First, Mads points out that IGBTs usually don't die until ~80Vge.  People here often aim for +-24V to get more current capability, with spikes hitting 30V and sometimes higher.  For SSTC I wouldn't intentionally drive above ~+-15Vge, but no damage for doing so.
Second, that rise and fall is probably an artifact of scoping a neutral-referenced circuit with a ground-referenced scope.  Try scoping neutral.  It probably has that same curve.  Probably the neutral wiring (in your house) voltage drop due to coil current.

Very interesting. I have always used MOSFETs so when I saw it was so high I was worried. I did not realize people drive them so hard. IGBTs are awesome! So if I under stand correctly, the Vmax is increasing but overall Vpp is likely staying the same so it is still within limits of the IGBT?

Your Vge waveform looks great.  No reason to change drive.  Back-to-back zeners are a good idea for any unexpected anomalous situations.  (Bidirectional TVS diodes are the same thing, designed specifically to handle such transients.)

Awesome! Glad to hear that it is acceptable. I picked up some bidirectional TVS diodes as well. I will have to take a look and see what they are rated for. I will probably include them just in case.

Yes, IGBT diodes are designed for freewheeling.  No reason to have external diodes too.  (Unless internal diodes are very slow, and external ones have enough lower voltage drop to take current instead of internal ones.  Unlikely.)

Yes, exactly as expected.  Notice that the falling-edge flat spot is missing.  That is because diodes bypasses 7.1 ohm gate resistor.  Speeds up turn-off exactly as intended.

I do see that. It is really cool to see that in action.

Great work with nice detailed measurements!

Thank you!! These last 6 months or so I have been learning more and more. It is a lot of work but I appreciate you taking the time to review everything and provide feedback.

You got me thinking about ZCD circuits.  Here's on option I designed/simulated this evening.  Generates zero-crossing current pulses to opto LED.  Circuitry on opto output needs to extend that trigger pulse to quarter-cycle or however long you want.  It is an option instead of the above use of tiny CT backwards.  Uses less current than CT likely requires, only 1mA RMS for what I simulated.  Simulated at 120Vrms (170Vpeak) and at 35Vrms (50Vpeak).  LED current drops, but should still be sufficient for opto.  Very little timing change.
This circuit detects line voltage rising edges, not both edges.

Edit:  Here's a bit fancier version of line zero-cross-detect.  Adds a second transistor to make LED current turn-on latching.  Tweaked other values too (bit higher line current, 1.6mA).  Now LED current pulse is constant and delay changes only 46us from 35Vrms to 120Vrms input.  Initial version is probably just fine for your use.  Adding this one just to satisfy my perfectionist tendencies:)

Thanks! I will simulate the schematic you included as well and give it a shot. I appreciate you taking the time to draw something up.

D5 looks to be a zenner diode. What value would it be?

-Zak


Update

Here are my calculations. I was not sure about the capacitor value. Since Mad's is based off a full bridge with two 0.68uf caps in series with the primary. I am also using two 0.68uf caps but do I also add them like Mad's did?

The primary res frequency is a little under half of the secondary which I suspect is too close and is resulting in issues.

Current draw is high but not likely 60A. But that probably explains why my FGA3060ADF IGBT's started dying at 2 primary turns. They are only rated for 60A continuous. Would a current probe tell me more accurately what the current draw is during switching? I would like to invest in a differential probe as well as a current probe.

[davekni]

That is what I was thinking as well. I wasn't sure how a single resistor was going to work. I am not really sure how to calculate the wattage or resistance. At mains (120vAC) Each cap is charged to +- 85v, right? Am I trying to discharge each cap after every pulse or only once the coil is power off completely?

Goal for half-bridge is to equalize the two cap voltages.  Between enable cycles, caps charge to +-85V.  Goal is to make it close to +-85V, not close to +0-170V or +170-0V.  Each resistor will have ~85V across it most of the time, so power needs to be calculated that way.

Very interesting. I have always used MOSFETs so when I saw it was so high I was worried. I did not realize people drive them so hard. IGBTs are awesome! So if I under stand correctly, the Vmax is increasing but overall Vpp is likely staying the same so it is still within limits of the IGBT?

I think FETs are also fairly robust for the low total use times of hobby projects, though not likely to 80V.  And SiC FETs are quite sensitive to Vgs.
No, Vmax is not likely increasing.  It is an artifact of the scoping technique I recommended.  Neutral voltage isn't quite identical to ground.  If you want to scope more precisely, use another probe to scope neutral, then use scope math to subtract neutral voltage from gate voltage.  (Or, use the same probe and save neutral to reference.  That presumes conditions repeat accurately.  Or purchase a differential probe.)

D5 looks to be a zenner diode. What value would it be?

D5 is a 9.1V zener.  I just picked a random 9.1V part from LTSpice's built-in library.

I was not sure about the capacitor value. Since Mad's is based off a full bridge with two 0.68uf caps in series with the primary. I am also using two 0.68uf caps but do I also add them like Mad's did?

Simulation is probably best.  My first guess would be to take one 0.68uF capacitor in parallel with the other two 0.68uF caps in series, so 1.02uF.  (Presuming you have a 0.68uF capacitor across Vbus.)

The primary res frequency is a little under half of the secondary which I suspect is too close and is resulting in issues.

May not be, depends on whether any startup cases manage to lock to this lower frequency.  I haven't seen any of your posted scope captures at that low a frequency.  The other issue is that the capacitor impedance subtracts some from primary coil inductance at secondary resonant frequency (400kHz).  Simulation is the best way to see this.  It does further increase primary current beyond the PI/2 factor of next paragraph.

Current draw is high but not likely 60A. But that probably explains why my FGA3060ADF IGBT's started dying at 2 primary turns. They are only rated for 60A continuous.

60A is peak IGBT current, not line current.  IGBT current draws power from line for a quarter-cycle, then feeds power back to line for a quarter-cycle.  Line current depends on arc load power and power of loss through the circuitry and coil windings.
As I'd mentioned, that formula is for sine waves.  With square wave drive, peak current is PI/2 times higher.  Here's a crude LTSpice simulation showing such:

Would a current probe tell me more accurately what the current draw is during switching? I would like to invest in a differential probe as well as a current probe.

I'd prioritize the differential probe.  Current transformers as you retrieved from power supplies or home-built ones will suffice fine for most current measurement needs.

In-line pictures worked this time   No idea why.

[ZakW]

Hello,

Here is an update on my progress so far. New PCBs should be here Tuesday!

This is the 5th iteration of my ramped SSTC I have made. Making general improvements and tweaks for better performance and stability. In the end I plan on sharing my final schematic which is not far off from the original by LoneOceans, however, I have a lot of notes on the overall design that I need to finalize. Explaining why I choose that specific component/value, what its function is, and some alternatives. Mainly all the things I have learned through out this process, starting back in December of 2021. Over a year already!

Here is the main PCB, coming in at 71mm (slightly smaller than a Post-it Note). That is a 50% reduction in area compared to v4. This is also my first time using SMD components so that will be a learn process as well. A big goal of this project was to make the smallest coil with the best output.








I am going to be testing two new mini coils. Both wound with 44awg wire. Using 2in and 2.5in PCV for the coil form.

Instead of coating the coils with epoxy like usual I decided to pour in the epoxy. Giving me about ~2.5mm of epoxy plus the thickness of the outer form. It should be slightly higher coupling than what I was able to achieve before without flash over so hopefully there will be some output improvement.

I 3D printed an end cap with two sections to hold the secondary coil in the center of the outer coil.












My winding setup. Only snapped one coil this time while winding. The magnifier helps a lot, it was much more difficult before.






Dave,

Thank you again for the schematic and advice. I was too invested in my optocoupler design to make changes at this point. I am happy to report though that by changing the AC input capacitor/resistor for the optocoupler from 560nf to ~280nf allowed me to tune the ZCD way more accurately. Lowering the cap value significantly increase the impedance (from around ~4.7k down to ~9.5k). Maybe too much current was flowing through the LED causing the signal to be less adjustable? Either way I am able to adjust it over a wide range of input voltages while maintaining proper zero crossing timing.

Goal for half-bridge is to equalize the two cap voltages.  Between enable cycles, caps charge to +-85V.  Goal is to make it close to +-85V, not close to +0-170V or +170-0V.  Each resistor will have ~85V across it most of the time, so power needs to be calculated that way.

I did not get around to adding the resistors across each cap on this design. I DO plan on adding them at some point.

May not be, depends on whether any startup cases manage to lock to this lower frequency.  I haven't seen any of your posted scope captures at that low a frequency.  The other issue is that the capacitor impedance subtracts some from primary coil inductance at secondary resonant frequency (400kHz).  Simulation is the best way to see this.  It does further increase primary current beyond the PI/2 factor of next paragraph.

To be safe I increased the DC blocking caps back to 1uf instead of .68uf.

60A is peak IGBT current, not line current.  IGBT current draws power from line for a quarter-cycle, then feeds power back to line for a quarter-cycle.  Line current depends on arc load power and power of loss through the circuitry and coil windings. As I'd mentioned, that formula is for sine waves.  With square wave drive, peak current is PI/2 times higher.  Here's a crude LTSpice simulation showing such:

You mention the formula was for sine waves, can you point me to the formula for square waves? I don't understand what I should be doing differently. PI/2= ~1.57. Am I simply multiplying the sine wave derived current value by 1.57?

In-line pictures worked this time   No idea why.

I also did not have any issues with pictures this time thankfully.

[davekni]

Instead of coating the coils with epoxy like usual I decided to pour in the epoxy. Giving me about ~2.5mm of epoxy plus the thickness of the outer form. It should be slightly higher coupling than what I was able to achieve before without flash over so hopefully there will be some output improvement.

Does make for a very nice looking secondary, and will very likely reduce or eliminate flash-overs.  Hopefully 2.5mm isn't thick enough to introduce too much added capacitance and dielectric loss.  My low-frequency (100kHz) QCW coil experiment has ~25mm epoxy internally, with an internal primary.  Epoxy added as much capacitance as top-load, and lowered secondary Q way down.  I don't recall exactly how much.  My DRSSTC has ~6mm epoxy without issue, but is also 1.1m high and only 80kHz.

You mention the formula was for sine waves, can you point me to the formula for square waves? I don't understand what I should be doing differently. PI/2= ~1.57. Am I simply multiplying the sine wave derived current value by 1.57?

Yes, peak current for a square wave voltage feeding a perfect inductor is a factor of 1.57 times higher than for a sine wave with the same peak voltage.
Current through an inductor is the time integral of voltage / inductance.  For either input voltage waveform, current is zero at the center of peak voltage and maximum at voltage zero-crossing.  So peak current is calculated by integrating over 1/4 cycle, from center of peak voltage to zero-crossing.  If cycle period is called "T", integrate over a time period of 0.25T.  For a square wave of peak voltage V, integral is just 0.25T * V / L.  For a sine wave of peak voltage V, integral is 0.25T * 2/PI * V / L.

Looks like a great project!

[ZakW]

Does make for a very nice looking secondary, and will very likely reduce or eliminate flash-overs.  Hopefully 2.5mm isn't thick enough to introduce too much added capacitance and dielectric loss.  My low-frequency (100kHz) QCW coil experiment has ~25mm epoxy internally, with an internal primary.  Epoxy added as much capacitance as top-load, and lowered secondary Q way down.  I don't recall exactly how much.  My DRSSTC has ~6mm epoxy without issue, but is also 1.1m high and only 80kHz.

It seems to work great! I did notice a little heating on the clear coil. Can you explain a bit more on what causes the heating. Before when I was using a roll of plastic for the primary insulation/form and I did not notice much heating.

Yes, peak current for a square wave voltage feeding a perfect inductor is a factor of 1.57 times higher than for a sine wave with the same peak voltage.
Current through an inductor is the time integral of voltage / inductance.  For either input voltage waveform, current is zero at the center of peak voltage and maximum at voltage zero-crossing.  So peak current is calculated by integrating over 1/4 cycle, from center of peak voltage to zero-crossing.  If cycle period is called "T", integrate over a time period of 0.25T.  For a square wave of peak voltage V, integral is just 0.25T * V / L.  For a sine wave of peak voltage V, integral is 0.25T * 2/PI * V / L.

Based on my previous post. I landed somewhere at 60amps. Would I simply multiply 60A by 1.57 = 94A? If that gets me in the range of the actual current I will need to redo my calculations since my primary dimensions have changed slightly.

Update:

I am assembling my new boards now but I was testing my new secondary coils yesterday and I killed two pairs of IGBTs. I am trying to get the most out of the driver so I am only using 2 turns on the primary so I am pushing it pretty hard (it has worked in the past with similar sized secondaries). Having said that I am wondering if the IGBTs are failing from over voltage spikes due to hard switching. I am using FGA60N65SMD 600v rated IGBTs. I installed some lower power IGBTs (FGA3060ADF) that I have used before and added a turn for a total of three turns. Everything was working fine until I added a larger top-load, that's when they failed again. By the way, all of the failures are silent.

I am considering adding some snubber capacitors directly across the C-E pins but I am not sure what value to use. I know Madd's has a calculator but I wanted to double check if there was another way calculate the require value.

Finally, I checked Goa's Ramped SSTC schematic https://www.loneoceans.com/labs/sstc3/schema_sstc3v1complete.jpg and noticed he is using a large bus capacitor (350v 680uf). Wouldn't that smooth the half wave input too much defeating the purpose of the AC ramp? I added a 1uf bus cap to help reduce voltage spikes and I observed that around 2uf I would start getting a loud snapping sound in the arc and the length would decrease. I am not sure how he was able to use such a large bus capacitor for a ramped build. Any ideas?

[davekni]

Can you explain a bit more on what causes the heating. Before when I was using a roll of plastic for the primary insulation/form and I did not notice much heating.

Heating is from dielectric loss.  It is due to AC electric field across epoxy.  No issue with magnetic field.  Urethane epoxy, at least what I used, has a higher dielectric loss than any other plastics I've used.  If I'd known that ahead of time, wouldn't have designed my low-frequency QCW experiment using so much epoxy.  Polypropylene, polyethylene, and polytetrafluoroethylene (Teflon) have very low dielectric loss.  PVC is a bit higher, but much less than epoxy.

Based on my previous post. I landed somewhere at 60amps. Would I simply multiply 60A by 1.57 = 94A? If that gets me in the range of the actual current I will need to redo my calculations since my primary dimensions have changed slightly.

Yes.  Though this formula is for perfect square waves into perfect inductors.  Actual peak current will differ some.
Looking back again at Mad's example, it is for a full-bridge.  For a half-bridge, peak current is half.  So perhaps you need to multiply by 0.5 * 1.57.
I've lost track of all your details.  I think this is 170V input to a half bridge?

Having said that I am wondering if the IGBTs are failing from over voltage spikes due to hard switching.

That board layout you posted looks fairly good for low parasitic inductance.  Seems somewhat unlikely that spikes would extend from 170V to 650V.  Unless C18, C19, and C20 have particularly high parasitic inductance.  BTW, if turning the ECB again, you could extend the Vbus (top node of C18) and connection between two IGBTs into plane sections to fill in empty space there.  Would decrease parasitic inductance slightly more.  Even if not exceeding 650V, spikes increase switching energy of IGBTs, which may be the real cause of failure.

I am considering adding some snubber capacitors directly across the C-E pins but I am not sure what value to use.

Adding CE capacitors (calculating value) is tricky.  May require increasing dead-time to accommodate.  SSTCs typically are hard switching for turn-off transitions only.  CE capacitance slows voltage rise, so reduces switching energy.  IGBT tail current is dissipated across a lower voltage drop, so dissipates less energy.  Ideally approaches ZVS conditions.  Voltage rise is slower, so opposite IGBT needs to turn on later, after voltage rise is complete.

If coil primary-to-secondary coupling factor is high and arc loss reasonably low, it is possible to get SSTCs into ZCS conditions.  That is a problem for added CE capacitance.  If current is low (closer to zero) at IGBT turn-off, voltage takes an even longer time to rise.  Voltage may not finish rising before it starts falling again (current polarity switches) or not finish before opposite IGBT turns on.  Turning-on across a charged capacitor dissipates that capacitor's energy in the IGBT that is turning on.

Finally, I checked Goa's Ramped SSTC schematic https://www.loneoceans.com/labs/sstc3/schema_sstc3v1complete.jpg and noticed he is using a large bus capacitor (350v 680uf).

That looks like a normal SSTC design, not a ramped one.

[ZakW]

Heating is from dielectric loss.  It is due to AC electric field across epoxy.  No issue with magnetic field.  Urethane epoxy, at least what I used, has a higher dielectric loss than any other plastics I've used.  If I'd known that ahead of time, wouldn't have designed my low-frequency QCW experiment using so much epoxy.  Polypropylene, polyethylene, and polytetrafluoroethylene (Teflon) have very low dielectric loss.  PVC is a bit higher, but much less than epoxy.

That is good to know. I was not aware it would heat up like that. Thankfully it doesn't seem to be too much for this small coil. Since I have never used epoxy before I just bought some cheap stuff off of Amazon to test out. I will have to look to see what kind it is.

Were you limited to short runs do to the epoxy heating too much?

Yes.  Though this formula is for perfect square waves into perfect inductors.  Actual peak current will differ some.
Looking back again at Mad's example, it is for a full-bridge.  For a half-bridge, peak current is half.  So perhaps you need to multiply by 0.5 * 1.57.
I've lost track of all your details.  I think this is 170V input to a half bridge?

I am using 120vAC so ~170. Since it is a half bridge, the voltage would be +- 85v. Do I still need to half the final value then? Seems like that is being addressed by only using 85v as my primary voltage.

Here is the picture of my maths from earlier in the thread.

That board layout you posted looks fairly good for low parasitic inductance.  Seems somewhat unlikely that spikes would extend from 170V to 650V.  Unless C18, C19, and C20 have particularly high parasitic inductance.  BTW, if turning the ECB again, you could extend the Vbus (top node of C18) and connection between two IGBTs into plane sections to fill in empty space there.  Would decrease parasitic inductance slightly more.

Thank you! I watched a lot of videos on best practices and tried to keep the layout and routing clean. Regarding extending the Vbus and connections between the two IGBTs I am using copper pours for those connections. I had them on outline mode in the original picture (see updated pic below). Red = top layer Blue = bottom neutral connections. Are you saying I could have increased the connections more on the top layer?

Even if not exceeding 650V, spikes increase switching energy of IGBTs, which may be the real cause of failure.

Can you explain a bit more about "switching energy"? Would that be the total current the IGBT is switching? My small heatsink hardly gets warm after running the coil for a few 30 sec runs. I plan on installing a small fan in the final design.

I have referenced Richie's site a lot https://www.richieburnett.co.uk/sstate3.html.

Regarding primary turns, he says:

"The magnetizing current does not perform any transfer of real power, and merely represents current sloshing back and forth between the primary winding and the drive circuit."

and

"If Lp is too small, the magnetizing current becomes unacceptably high. Although this does not contribute to the supply current of the driver, it does contribute to heating in the switches. It also increases the current ripple seen by the supply reservoir capacitors."

I have a feeling 2 primary turns might be too low and is overloading the driver. I get great results though!  :'(  I'd like to test that with my newer version to see if condensing everything and changing to SMD parts makes a difference. If not I will have to increase it a bit (if that is the issue).

Adding CE capacitors (calculating value) is tricky.  May require increasing dead-time to accommodate.  SSTCs typically are hard switching for turn-off transitions only.  CE capacitance slows voltage rise, so reduces switching energy.  IGBT tail current is dissipated across a lower voltage drop, so dissipates less energy.  Ideally approaches ZVS conditions.  Voltage rise is slower, so opposite IGBT needs to turn on later, after voltage rise is complete. If coil primary-to-secondary coupling factor is high and arc loss reasonably low, it is possible to get SSTCs into ZCS conditions.  That is a problem for added CE capacitance.  If current is low (closer to zero) at IGBT turn-off, voltage takes an even longer time to rise.  Voltage may not finish rising before it starts falling again (current polarity switches) or not finish before opposite IGBT turns on.  Turning-on across a charged capacitor dissipates that capacitor's energy in the IGBT that is turning on.

Like you pointed out I probably have low enough stray inductance that adding snubbers might cause more harm than good. I will test the new boards and see.

That looks like a normal SSTC design, not a ramped one.

I took a closer look and that is indeed from his Ramped SSTC page, but looking at the final build pictures before adding the driver to the enclosure I can see he did not install the bus cap. I was confused because no where in his schematic did he have a way to disconnect the bus cap so it would not filter the input.

[davekni]

I am using 120vAC so ~170. Since it is a half bridge, the voltage would be +- 85v. Do I still need to half the final value then? Seems like that is being addressed by only using 85v as my primary voltage.

Yes, you have already taken care of the half-bridge by using 85V.  So just multiply 60.3A by 1.57 to get 94.7A.  Simulation including DC blocking caps would be more accurate.

Are you saying I could have increased the connections more on the top layer?

I'm not quite clear.  Is the semi-transparent red showing showing what could be added, or what is already there?  Semi-transparent red area on Vbus+ (top of C18 and C19) is definitely helpful.  Output to primary coil is less important.  That current is already sine-wave due to primary inductance.  What does matter more is the path from high-side emitter to low-side collector.  This part of the one primary output node is important.  Current rapidly swaps from one IGBT to the other through the portion of this trace connecting the two IGBTs.  Some of that trace could be widened, the portion across C20 and C19.

Can you explain a bit more about "switching energy"? Would that be the total current the IGBT is switching? My small heatsink hardly gets warm after running the coil for a few 30 sec runs. I plan on installing a small fan in the final design.

So average power dissipation is clearly low.  Peak power at the end of a ramp will be higher.  IGBT junction temperature may momentarily get too high.  Or, I may be wrong about what is causing occasional IGBT failures.
Switching energy isn't a simple topic, depending on many factors.  For basic understanding, I suggest analog simulation (LTSpice or other).  I've had poor success with IGBT simulation models, but FETs will suffice for simulation.  Added gate series resistance can slow switching to make FETs behave a bit more like IGBTs.  Add parasitic inductance in series with Vbus supply and/or between high and low side IGBTs (FETs) and compare power.  (In LTSpice, hold Alt key down while clicking on a device in order to plot device power dissipation.)

I have a feeling 2 primary turns might be too low and is overloading the driver. I get great results though!  :'(  I'd like to test that with my newer version to see if condensing everything and changing to SMD parts makes a difference. If not I will have to increase it a bit (if that is the issue).

"Overloading" just means that primary current is too high for your IGBTs.  "Too high" may be driven by switching losses as much or more than conduction losses.  New design with lower parasitic inductance will help.

Of course, IGBT failure may have a different cause.  Scoping will show if voltage spikes are the issue (above 650V).  It is also possible that there is some erroneous circuit behavior at high load, perhaps due to secondary arc or due to switching spikes coupling into low-voltage circuitry.  (Not enough to fry low-voltage circuitry, just enough to cause faulty operation.)  Or perhaps some occasional condition causes coil to drop to 182kHz primary resonance and current thus increase way more than normal.  Everything in this paragraph is brainstorming ideas, not anything likely.

Just noticed what may be a critical layout issue.  Looks like high-side GDT emitter connection trace comes from low-side collector.  Switching spikes across power trace from high emitter to low collector will be imposed across high-side Vge.  That may be either causing erroneous operation or directly frying high-side IGBT from excess Vge.  When IGBTs fry, which pins are shorted or open?  Fix is to make the two high-side GDT traces parallel (like differential pairs).  This is much more important than keeping the high-side gate trace wide as you have it now (trace from D6 to GDT).  If hand-patching, cut trace from GDT pin 4 and wire GDT to D8 pin 1 or high-side emitter directly, routed along GDT to D6 trace.

Good luck!

[ZakW]

Yes, you have already taken care of the half-bridge by using 85V.  So just multiply 60.3A by 1.57 to get 94.7A.  Simulation including DC blocking caps would be more accurate.

Great! I am glad that is a rough estimate.

I'm not quite clear.  Is the semi-transparent red showing showing what could be added, or what is already there?  Semi-transparent red area on Vbus+ (top of C18 and C19) is definitely helpful.  Output to primary coil is less important.  That current is already sine-wave due to primary inductance. 

Yes, in KiCad the semi-transparent red is a (top copper fill/pour).

What does matter more is the path from high-side emitter to low-side collector.  This part of the one primary output node is important.  Current rapidly swaps from one IGBT to the other through the portion of this trace connecting the two IGBTs.  Some of that trace could be widened, the portion across C20 and C19.

I see what you mean. I will adjust the routing.

Of course, IGBT failure may have a different cause.  Scoping will show if voltage spikes are the issue (above 650V).  It is also possible that there is some erroneous circuit behavior at high load, perhaps due to secondary arc or due to switching spikes coupling into low-voltage circuitry.  (Not enough to fry low-voltage circuitry, just enough to cause faulty operation.)  Or perhaps some occasional condition causes coil to drop to 182kHz primary resonance and current thus increase way more than normal.  Everything in this paragraph is brainstorming ideas, not anything likely.

I appreciate the ideas. I only have standard probes at the moment. Could I connect my scope probe ground to a 1uf cap to mains neutral again like before?

I don't think the coil is locking on the lower 182kHz fres.

I fried two more pairs of IGBT's. I am out of switches until I order more. sigh... I think 90% of the cases where they have died is during primary-secondary flash over. Tonight when I was adjust the primary I had it a bit too high and it arced to the PCV form and pop, blew the fuse. Another time was when I added a larger top load, normally I have to lower the primary a bit to avoid flashover. When I powered up the coil it did not even arc it just blew the fuse.

Just noticed what may be a critical layout issue.  Looks like high-side GDT emitter connection trace comes from low-side collector.  Switching spikes across power trace from high emitter to low collector will be imposed across high-side Vge.  That may be either causing erroneous operation or directly frying high-side IGBT from excess Vge.  When IGBTs fry, which pins are shorted or open?  Fix is to make the two high-side GDT traces parallel (like differential pairs).  This is much more important than keeping the high-side gate trace wide as you have it now (trace from D6 to GDT).  If hand-patching, cut trace from GDT pin 4 and wire GDT to D8 pin 1 or high-side emitter directly, routed along GDT to D6 trace.

YIKES! thanks for spotting that. That's what I get for following the ratsnest and not checking with my eyes! Looks like I will be ordering more boards in the future. In the meantime I will cut it and solder a wire in like you suggest.  I also forgot the component silkscreen check box when I exported the files 

When IGBTs fry, which pins are shorted or open?

Out of all the dead IGBTs I kept, I would say it is an 80/20 split.

Most of them are shorted C-E. For a few all 3 pins were shorted. Obviously, flashover is bad but I have had it happen several times in the past without killings IGBTs. Besides not having it happen is there anything I can do to mitigate the effects (if that is the issue)?

So far the FGA60N65SMD seemed to have worked the best. Initially, I was using the FGA3060ADF and even tried the middle ground with the FGA40N65SMD. All died.

I don't know if I need to order parts that have higher voltage or current ratings or both at this point. I need to order something so I can test the board but they will likely die during a full power run. Do you have any recommendations for fast IGBTs? Another consideration is the UCC27425 driver I am using. To date the FGA60N65 had the largest typical total gate capacitance of around 200nC. The UCC chip never got warm.

In this thread https://highvoltageforum.net/index.php?topic=1634.0 you said:

If you stay with IGBTs, modulation by phase or dead-time will lead to some hard switching.  Frequency is too high for IGBT hard switching, so you will need to change to FETs.  (IGBT modulation options are pulse-skipping or separate modulated bus supply to the half-bridge.)

It looks like there coil was running about 220kHz to 320kHz. Mine is ~400kHz. Since there is a bit of delay due to signal processing and CT feedback, does that mean I might be hard switching as well?

Per Loneoceans page, he used a Half bridge of FGH40N60SMD TO-247 IGBTs at 450-420kHz Secondary Frequency (with and without loading). I would like to think my build is comparable and should work similar frequencies.

[davekni]

I only have standard probes at the moment. Could I connect my scope probe ground to a 1uf cap to mains neutral again like before?

Yes, presuming your scope probe can handle 650V spikes if there are any.  I doubt spikes will be that high.  Would be nice to rule that out.

I don't think the coil is locking on the lower 182kHz fres.

I presume it is reliably starting at secondary frequency.  Was just wondering (still unlikely) if under some conditions such as a long arc at the end of a ramp the frequency might switch.

I think 90% of the cases where they have died is during primary-secondary flash over.

Obviously, flashover is bad but I have had it happen several times in the past without killings IGBTs.

That is the sort of event that might cause frequency to change to 182kHz or cause other circuit disruption.  Sounds like you need a grounded guard rail around primary.  Common for DRSSTCs, but your SSTC appears to need that.

Most of them are shorted C-E. For a few all 3 pins were shorted.

Gate oxide breakdown often leads to GE short without CE short.  Looks like layout issue hasn't fried gate oxide.  TVS diode probably prevents that.  Fixing the layout is still important, reducing switching energy.

It looks like there coil was running about 220kHz to 320kHz. Mine is ~400kHz. Since there is a bit of delay due to signal processing and CT feedback, does that mean I might be hard switching as well?

Yes, though presumably turn-off only.  Hard switching just refers to switching at or close to full current.  Turn-on hard switching is usually even more problematic than turn-off hard switching.  At ~94A, even turn-off switching may be causing some of your IGBT failures.  The layout fix will reduce switching energy at least some, so should help or even fix IGBT failure issues.

Per Loneoceans page, he used a Half bridge of FGH40N60SMD TO-247 IGBTs at 450-420kHz Secondary Frequency (with and without loading). I would like to think my build is comparable and should work similar frequencies.

Per Loneoceans page, he is using a 6uH primary, so significantly lower current than yours.  He calculated 20.8A, but has a couple mistakes in that calculation.  Will be higher, but not close to 94A.

I don't know if I need to order parts that have higher voltage or current ratings or both at this point. I need to order something so I can test the board but they will likely die during a full power run. Do you have any recommendations for fast IGBTs? Another consideration is the UCC27425 driver I am using. To date the FGA60N65 had the largest typical total gate capacitance of around 200nC. The UCC chip never got warm.

The layout fix and primary guard ring may be enough to keep IGBTs working, so FGA60N65 may be fine.  You could consider a bit higher gate drive voltage.  For high peak current (especially for DRSSTC), many designs run Vge at +-18V to +-24V.  For your SSTC, wouldn't go too high, but +-18V might get you slightly more current capability.  There are other driver chips capable of 20V and higher, such as IX4340, UCC27624, and IVCR2401.
For other IGBTs, my DRSSTC uses STGW60H65DRF.  I'm building a 450kHz QCW now using FGH75T65SHDTLN4.  They happen to be available cheap because that exact sub-variant is discontinued.  It is a 4-pin device (kelvin emitter connection), great for high frequency gate drive, but requires layout to match.  There is an almost identical 3-pin version FGH75T65SHD, and another 4-pin non-obsolete version FGH75T65SHDTL4.  These are all lower gate charge than you have now.

[ZakW]

Yes, presuming your scope probe can handle 650V spikes if there are any.  I doubt spikes will be that high.  Would be nice to rule that out.

I will give it a shot. Is it okay to use a 10x probe? That is what I mean by standard probe.

That is the sort of event that might cause frequency to change to 182kHz or cause other circuit disruption.  Sounds like you need a grounded guard rail around primary.  Common for DRSSTCs, but your SSTC appears to need that.

A ground ring will be interesting, I will see if I can fit one. Do I need to worry about high current strikes to ground, especially over such short distances?

Edit: to be more specific, the primary/secondary flash over is happening from the primary on the outer coil form/epoxy to the secondary coil. Looks like thin purple arcs of corona on the PCV. I am not getting arcs downward to the primary coil if that make sense.

Yes, though presumably turn-off only.  Hard switching just refers to switching at or close to full current.  Turn-on hard switching is usually even more problematic than turn-off hard switching.  At ~94A, even turn-off switching may be causing some of your IGBT failures.  The layout fix will reduce switching energy at least some, so should help or even fix IGBT failure issues.

I think this looks a lot better. What do you think?

The layout fix and primary guard ring may be enough to keep IGBTs working, so FGA60N65 may be fine.

The 60N65SMD variant is not in stock at Mouser at the moment. I am going to try the FGH75T65SHD-F155 in the meantime. It is a little more robust.

You could consider a bit higher gate drive voltage.  For high peak current (especially for DRSSTC), many designs run Vge at +-18V to +-24V.  For your SSTC, wouldn't go too high, but +-18V might get you slightly more current capability.  There are other driver chips capable of 20V and higher, such as IX4340, UCC27624, and IVCR2401.

Thanks, I will see about picking up some 18v regulators and a few UCC27624DR chips to have on hand. For now I will try to stick with 15v.

If I decide to drive the gate at 18v, should I get TVS diodes around 18 or 20v? I am using 15v diodes (P6SMB15CA-Q), which have a breakdown Voltage of 14.3 V and a clamping Voltage of 21.2 V.

______________________

Lastly, I have a question regarding the bypass caps and their value. Awhile back I was learning about the response curve (Cimpedance vs Freq.) and since SMD parts + proper layout has such low inductance you can use larger values, instead of the universal 0.1uf. For this build I used 1uf ceramic caps for everything on the board. I also included a 10uf ceramic + 1uf on the UCC chip. Off hand, was 1uf too high? I forgot to scope the power input on each chip before I killed the last pair of IGBTs I will have to look next time.

I made sure to keep the caps as close to the power/ground pins as possible. Hopefully noise was not a contributing factor to the failures as well.

[davekni]

I will give it a shot. Is it okay to use a 10x probe? That is what I mean by standard probe.

I just checked two of my cheap 10x probes.  Both are rated for 600V.  Doubt spikes will be that high.  Also doubt that the probe will be damaged even if brief spikes go above 600V.

Edit: to be more specific, the primary/secondary flash over is happening from the primary on the outer coil form/epoxy to the secondary coil. Looks like thin purple arcs of corona on the PCV. I am not getting arcs downward to the primary coil if that make sense.

Oh, I'd misunderstood.  Makes me wonder if that corona is a symptom rather than cause of an issue.  TC secondaries do have higher frequency resonant modes.  The next one above normal is where center of secondary is maximum voltage and top and bottom are both low voltage.  Wild guess might be 2 or 3MHz for your coil.  Oscillation at such high frequency might fry IGBTs and lead to corona where secondary voltage is high.

Just to double-check my understanding, primary/secondary flash over is occurring only occasionally and associated with IGBT failure.  Correct?  If there is corona along secondary adjacent primary normally, then secondary volts/turn may be too high there.  Normal advice in this case is to reduce coupling, but that tends to reduce performance.  If a 3-turn primary with windings spread out could get close to the same inductance as two close-spaced turns, that would reduce volts/turn and increase coupling.  Or, a 2-turn primary with slightly larger diameter and added space between the two turns.

A ground ring will be interesting, I will see if I can fit one. Do I need to worry about high current strikes to ground, especially over such short distances?

Yes, for an SSTC, ground strikes (and strikes from top to primary) may both be problematic.  A low-resistance arc would short secondary, eliminating secondary resonant frequency.  Secondary current would then match primary, a bit above 182kHz due to shorted secondary effect on primary inductance.  Current could then ramp up indefinitely until interrupter period ends.  This is a theoretical concern.  I have no personal experience with such an SSTC issue.  I do know that ground strikes on my DRSSTC cause primary current to ramp up until either OCD or interrupter ends the pulse.  (I was going to edit my last post with this concern after thinking about it during my sunset walk this evening.)

I think this looks a lot better. What do you think?

Yes, looks great!  Since I'm a perfectionist, one tiny tweak you could make:  Route GDT pin 5 to D9 and low-side emitter on red layer instead of on ground (neutral) plane.  That keeps gate return path completely away from power path.  (Of course, IGBT emitter lead is still common to power and gate paths.  That is the reason some IGBTs come in 4-lead packages, to avoid that common path.)

I am going to try the FGH75T65SHD-F155 in the meantime. It is a little more robust.

I suspect you will be happy with that part.  Significantly lower gate charge, slightly faster switching, and a bit higher current capability too.  Given the lower gate charge (lower capacitance), resistors in series with gates may need to increase in value a little to maintain dead time.  Although, perhaps not.  FGH75T65SHD-F155 is faster, with less difference between turn-off and turn-on times, so shouldn't require quite as much dead time.

If I decide to drive the gate at 18v, should I get TVS diodes around 18 or 20v? I am using 15v diodes (P6SMB15CA-Q), which have a breakdown Voltage of 14.3 V and a clamping Voltage of 21.2 V.

For commercial designs, it is necessary to handle worst-case part variations.  For home projects, I measure the breakdown voltage any time I purchase TVS diodes.  Possible that your 15V parts are high enough (at least 18V at a few mA), but probably necessary to change to 18V or 20V as you mentioned.

Lastly, I have a question regarding the bypass caps and their value. Awhile back I was learning about the response curve (Cimpedance vs Freq.) and since SMD parts + proper layout has such low inductance you can use larger values, instead of the universal 0.1uf. For this build I used 1uf ceramic caps for everything on the board. I also included a 10uf ceramic + 1uf on the UCC chip. Off hand, was 1uf too high? I forgot to scope the power input on each chip before I killed the last pair of IGBTs I will have to look next time.

For SMD ceramic caps, doubt you will have any issue with too-high capacitance.  However, do pay attention to dielectric type.  Z5U or Y5V caps have such steep temperature and voltage sensitivities that I avoid them completely, both at home and work.  Use X7R or at least X5R caps.  Even those can loose a lot of capacitance due to DC bias.  Some parts drop 90% at rated DC voltage.  Using two different values by critical chips is a great feature.  Even X7R capacitors can be a bit piezoelectric, having high impedance at their mechanical resonant frequency.  Two different parts will usually avoid resonances at the same frequency.

I made sure to keep the caps as close to the power/ground pins as possible. Hopefully noise was not a contributing factor to the failures as well.

Great!  Good design practice.

[ZakW]

I just checked two of my cheap 10x probes.  Both are rated for 600V.  Doubt spikes will be that high.  Also doubt that the probe will be damaged even if brief spikes go above 600V.

Excellent! Thanks for confirming that for me.

Oh, I'd misunderstood.  Makes me wonder if that corona is a symptom rather than cause of an issue.  TC secondaries do have higher frequency resonant modes.  The next one above normal is where center of secondary is maximum voltage and top and bottom are both low voltage.  Wild guess might be 2 or 3MHz for your coil.  Oscillation at such high frequency might fry IGBTs and lead to corona where secondary voltage is high.

I am not very familiar with the resonant modes or maybe I don't understand enough. Is that similar to "poles" that I hear around QCW coils?

Just to double-check my understanding, primary/secondary flash over is occurring only occasionally and associated with IGBT failure.  Correct?  If there is corona along secondary adjacent primary normally, then secondary volts/turn may be too high there.  Normal advice in this case is to reduce coupling, but that tends to reduce performance.  If a 3-turn primary with windings spread out could get close to the same inductance as two close-spaced turns, that would reduce volts/turn and increase coupling.  Or, a 2-turn primary with slightly larger diameter and added space between the two turns.

That is correct. I get the coupling as close as I can before it arcs, then I back it off a bit. Once I established the spacing I slide the primary up/down to find the best performance. I have noticed with the really small coils I am winding and the thick 12awg silicone wire I am using, adjusting the height by a small amount can easily cause it to go too high. That results in arcing on the secondary. Normally with my previous build that did not result in a failure unless I have really over done it. Except with this new board (which has those routing issues) the IGBTs have died during the flash over.

I am using 44awg wire in order to shrink the secondary way down so the primary is probably seeing a lot of the windings. I have noticed too that I get the flash over if the top load is adjusted higher or swapped for a bigger size. I normally have to move the primary lower for a larger top load.

Yes, looks great!  Since I'm a perfectionist, one tiny tweak you could make:  Route GDT pin 5 to D9 and low-side emitter on red layer instead of on ground (neutral) plane.  That keeps gate return path completely away from power path.  (Of course, IGBT emitter lead is still common to power and gate paths.  That is the reason some IGBTs come in 4-lead packages, to avoid that common path.)

Great, that is an easy fix, I will update that trace. Looking at it now it makes so much sense. I took a break halfway through the design and should have paid closer attention when I started working on it again. That is good to know about the 4-pin package.

I suspect you will be happy with that part.  Significantly lower gate charge, slightly faster switching, and a bit higher current capability too.  Given the lower gate charge (lower capacitance), resistors in series with gates may need to increase in value a little to maintain dead time.  Although, perhaps not.  FGH75T65SHD-F155 is faster, with less difference between turn-off and turn-on times, so shouldn't require quite as much dead time.

I will keep that in mind. Thanks for taking a look and comparing it with the 60n65.

Regarding the gate resistors, I used 1/4w SMD resistors. Didn't really get a chance to test them out but I have not noticed any issues with 1/4watt through hole resistors either. Should I get higher wattage resistors? Loneoceans used 2W which seems overkill.

For commercial designs, it is necessary to handle worst-case part variations.  For home projects, I measure the breakdown voltage any time I purchase TVS diodes.  Possible that your 15V parts are high enough (at least 18V at a few mA), but probably necessary to change to 18V or 20V as you mentioned.

Just to be clear, are you saying I should use 18v or 20v TVS diodes when powering my gate drive IC at 15v?

For SMD ceramic caps, doubt you will have any issue with too-high capacitance.  However, do pay attention to dielectric type.  Z5U or Y5V caps have such steep temperature and voltage sensitivities that I avoid them completely, both at home and work.  Use X7R or at least X5R caps.  Even those can loose a lot of capacitance due to DC bias.  Some parts drop 90% at rated DC voltage.  Using two different values by critical chips is a great feature.  Even X7R capacitors can be a bit piezoelectric, having high impedance at their mechanical resonant frequency.  Two different parts will usually avoid resonances at the same frequency.

I have heard you mention that before about the Z5U and Y5V caps, I will be sure to stay away.


As always, thank you for sharing your knowledge and providing insight to all of my questions.

[davekni]

I am not very familiar with the resonant modes or maybe I don't understand enough. Is that similar to "poles" that I hear around QCW coils?

Similar only in that there are multiple frequencies.  Think of a musical instrument such as trumpet.  Multiple notes with one fingering depending on how many nodes and antinodes are along the tube length.

Just to be clear, are you saying I should use 18v or 20v TVS diodes when powering my gate drive IC at 15v?

No, 15V is fine for existing design.  The UCC driver chips with BJT output stages drive a bit less than full supply voltage, so your Vge is slightly under +-15V.  And typical 15V rated TVS diodes don't conduct much until 16 or 17V.

Regarding the gate resistors, I used 1/4w SMD resistors. Didn't really get a chance to test them out but I have not noticed any issues with 1/4watt through hole resistors either. Should I get higher wattage resistors? Loneoceans used 2W which seems overkill.

One quick estimate:  Measure +15V current, presuming most is going to UCC driver chip.  Power into driver chip is roughly split between driver chip and gate resistors.  A bit more than half in driver chip and a bit less than half in resistors.  Average power is probably fine.  During one ramp may be a bit high.  Thin-film resistors are better than thick-film for pulse power handling.  Thin film 1206 resistors are usually rated 0.5W.  On the other hand, if resistance isn't drifting (usually up) with use, then you should be fine with standard thick-film 1/4W 1206 resistors.

As always, thank you for sharing your knowledge and providing insight to all of my questions.

You are welcome.  I enjoy seeing your interest in detail, and I learn from your experiments too.

[ZakW]

The next one above normal is where center of secondary is maximum voltage and top and bottom are both low voltage.

Do you think moving the primary up the small secondary could cause this condition? Oscillating at those frequencies does sound destructive.

I found an older video of a flashover. Note that the coil was running for about 30 seconds prior to the flash over. The bright yellow flash is from the insulation burning. Normally, it is just diffuse purple arcing/corona (as you can see in the screenshot).

/>

One quick estimate:  Measure +15V current, presuming most is going to UCC driver chip.

I assume I would do that with my scope. Unfortunately, my scope/probe knowledge is lacking. How would I go about taking that measurement?

Thin-film resistors are better than thick-film for pulse power handling.  Thin film 1206 resistors are usually rated 0.5W.  On the other hand, if resistance isn't drifting (usually up) with use, then you should be fine with standard thick-film 1/4W 1206 resistors.

Glad to hear that they should be okay. I could always double the value and solder two stacked for higher peak currents right?

I googled thick vs thin resistors since I was not familiar with the differences and I found this site that mentions pulse power: https://passive-components.eu/resistor-types-and-construction/ They say "Pulse handling for thin film is better for longer term pulses, but thick film resistors are better for short pulses of 0.0001 second and shorter.".

10 kHz = 0.0001 s(p) | 400 kHz = 0.0000025 s(p)

I am sure this is splitting hairs but I am curious - since I am running around ~400kHz wouldn't the thick film be better for such a frequency?

 

[davekni]

Do you think moving the primary up the small secondary could cause this condition?

Yes.  Capacitance from middle of secondary to primary will lower the frequency and raise the Q of that mode, so make it more likely.

I found an older video of a flashover. Note that the coil was running for about 30 seconds prior to the flash over. The bright yellow flash is from the insulation burning. Normally, it is just diffuse purple arcing/corona (as you can see in the screenshot).

Interesting video!  I see several bursts with no hint of corona around the primary, and several with obvious bright corona.  That would suggest some abrupt change in operation (such as switch in secondary mode).  If corona were a result of normal operation, would expect to see some hints of it normally.  Mode change is likely occurring at the end of a burst when arc load is maximum.  That lowers Q of normal mode, perhaps below that of second mode in your case.  If your scope has a sufficiently long record length or can be triggered by the end of the burst, you could look for high frequency operation.  (Would be even better if you scope has a pulse-width trigger function.  Then you could trigger on only such anomalous events.)

I assume I would do that with my scope. Unfortunately, my scope/probe knowledge is lacking. How would I go about taking that measurement?

No, current probes tend to be expensive.  I was thinking of a meter in series with 15V supply input (or higher voltage input to 15V regulator).  Measure quiescent current and operating current at some duty cycle.  Difference will be mostly driver chip current.

Glad to hear that they should be okay. I could always double the value and solder two stacked for higher peak currents right?

Yes, I stack resistors sometimes for peak power capability in home projects.

I am sure this is splitting hairs but I am curious - since I am running around ~400kHz wouldn't the thick film be better for such a frequency?

Haven't seen that 100us distinction before.  If valid, probably varies with resistance.  High-value resistors may have different pulse handling capability than do low-value resistors.
Also points out another subtly:  Gate resistors experience pulses of power at 400kHz (~300ns wide) enabled by longer 1/4-line-cycle pulses (~4ms wide).  There are two pulse durations of interest.  I'd guess the longer pulse envelope is more relevant here, but don't know.  If a resistor includes a pulse rating curve (energy or power as a function of pulse width), then you could check both pulse widths to see which is closer to spec limit.  Relatively few resistors are actually spec'ed and rated for pulse duty.

[ZakW]

Interesting video!  I see several bursts with no hint of corona around the primary, and several with obvious bright corona.  That would suggest some abrupt change in operation (such as switch in secondary mode).  If corona were a result of normal operation, would expect to see some hints of it normally.  Mode change is likely occurring at the end of a burst when arc load is maximum.  That lowers Q of normal mode, perhaps below that of second mode in your case.

I am glad it provided some additional context. This test was done with my first PCB (driver only), the H-Bridge was a copy of your low inductance design with the coper tape. Likely no issue there. If I recall I think I had issues with skipping pulses and generally unreliable operation. Likely due to interference, poor layout, etc. Probably more of a driver/feedback issue compared to what I am dealing with now.

Like I mentioned before, with this latest design as soon as I would power up the coil (full power @ 120v) the flash over would occur and the IGBTs would die. The previous PCB (not the one in the video...too many versions  ) seemed to be more resilient to these start up flash overs and would not instantly die.

I need to be more careful about full power tests after adjusting the primary height.

If your scope has a sufficiently long record length or can be triggered by the end of the burst, you could look for high frequency operation.  (Would be even better if you scope has a pulse-width trigger function.  Then you could trigger on only such anomalous events.)

I have seen the video capture function before. I will play around with it once I get more IGBTs.

No, current probes tend to be expensive.  I was thinking of a meter in series with 15V supply input (or higher voltage input to 15V regulator).  Measure quiescent current and operating current at some duty cycle.  Difference will be mostly driver chip current.

That I can do, I will add that to my list to check.

Haven't seen that 100us distinction before.  If valid, probably varies with resistance.  High-value resistors may have different pulse handling capability than do low-value resistors.
Also points out another subtly:  Gate resistors experience pulses of power at 400kHz (~300ns wide) enabled by longer 1/4-line-cycle pulses (~4ms wide).  There are two pulse durations of interest.  I'd guess the longer pulse envelope is more relevant here, but don't know.  If a resistor includes a pulse rating curve (energy or power as a function of pulse width), then you could check both pulse widths to see which is closer to spec limit.  Relatively few resistors are actually spec'ed and rated for pulse duty.

Thanks for elaborating. I was just curious.

______________

I was looking at getting this differential probe on Amazon https://www.amazon.com/gp/product/B0795XQR68/ref=ox_sc_act_title_1?smid=A1LBN2V0Z4JGUA&psc=1

Seems to have good reviews. It doesn't look anything special but might be a decent budget probe. Will this work for this sort of application?

[davekni]

Like I mentioned before, with this latest design as soon as I would power up the coil (full power @ 120v) the flash over would occur and the IGBTs would die. The previous PCB (not the one in the video...too many versions  ) seemed to be more resilient to these start up flash overs and would not instantly die.
I need to be more careful about full power tests after adjusting the primary height.

Yes, testing first at lower voltage is recommended.  Subtle changes may make second mode more or less likely to oscillate.  Your initial driver layout may have been slower due to parasitic inductances etc., so not as capable of oscillating at the higher second-mode frequency.

I was looking at getting this differential probe on Amazon https://www.amazon.com/gp/product/B0795XQR68/ref=ox_sc_act_title_1?smid=A1LBN2V0Z4JGUA&psc=1

Seems to have good reviews. It doesn't look anything special but might be a decent budget probe. Will this work for this sort of application?

Yes.  It is roughly a copy of Tektronix P5205.  I have one of each, DP10013 purchased new and an old used P5205 from EBay.  Both work fine.  Only caution is that the DP10013 has no specification for how voltage capability drops at high frequency.  They claim it is good to full 1300V up to 100MHz.  Mads burned one out somewhere within that range, at somewhat high voltage and somewhat high frequency.  He's posted about that to the forum.  I don't recall details.  The P5205 starts dropping max voltage at about 300kHz.  Good to 800V at 500kHz or so per its spec graph.  Don't know if DP10013 is worse or not.  Cheap probes usually derate faster.  Your 170V or even 340V is probably fine at 500kHz.

[ZakW]

Yes.  It is roughly a copy of Tektronix P5205.  I have one of each, DP10013 purchased new and an old used P5205 from EBay.  Both work fine.  Only caution is that the DP10013 has no specification for how voltage capability drops at high frequency.  They claim it is good to full 1300V up to 100MHz.  Mads burned one out somewhere within that range, at somewhat high voltage and somewhat high frequency.  He's posted about that to the forum.  I don't recall details.  The P5205 starts dropping max voltage at about 300kHz.  Good to 800V at 500kHz or so per its spec graph.  Don't know if DP10013 is worse or not.  Cheap probes usually derate faster.  Your 170V or even 340V is probably fine at 500kHz.

Thanks for the info, I am excited to test it out. I have to order new boards so it is going be a week or so. I will be back with more updates!

[ZakW]

I received my new boards. Got one all put together and everything was working great! The redesigned bridge portion was a significant improvement. I also played around with my differential probe which will be a great tool for testing and troubleshooting.
 
Bad news: Pushed the driver too hard messing around with it the other night. Had the primary-secondary flash over occur and killed the IGBTs. I replaced them the next morning only to find the the coil would not run. I was getting a weird snapping/popping sound but no output.
 
I replaced a lot of components that I thought might be causing the issue to no avail. Here is what I cannot figure out, everything was working correctly except the signal from the CT was only a super short pulse. The snapping was the coil turning on for a short time. Nothing I replaced would fix what I assume is a feedback issue. Specifically, I was scoping the output of the CT and it was just a very short pulse. The coil was trying to turn on for the duration of the interrupter signal, but it was not oscillating after ~90us and the coil would shut off. Every now and then I would get it to fire and the feedback from the CT at that time looked normal.
 
I connected my signal generator to a single loop of wire through the CT and ran it at 400kHz (3v sine wave). Driving it like that I could get the coil to run but to ideal since it was at a fixed frequency. The interrupter was working as it should, the ZCD (another issue) seemed to be working, the 74hc14 was receiving the signal and outputting a perfect looking 400kHz square wave, the UCC output looked perfect as well. I tested all this several times before and after installing new IC's, diodes, caps, etc...  but nothing fixed it.
 
I have had this issue before, but it has come and gone with different builds without a clear solution (maybe an issue with the ZCD I am using since it spans many builds). This may be a symptom or directly related but the adjustment pot I have for the ZCD optocoupler was extremely sensitive. Before the IGBT failure I could freely adjust the zcd timing without really impacting the coil. It would get a little snappy if the timing was too off etc, no big issue. However, after I replaced the IGBTs the only way I could get any output at low and higher voltages was to get the signal just perfect. Even then the coil was stuttering and skipping pulses.
 
Yellow = CT feedback. You can see I am zoomed in one short pulse. For some reason the CT or the 74hc14 just stops producing feedback?
 [ Invalid Attachment ]
 
Having said all of that, I have to rebuild the driver on a spare board and can no longer test the one that was having issues. In the process of replacing the 556 timer I ended up peeling up a couple micro traces so it is not salvageable. I am concerned that I will have this same issue again if I rebuild it. If it works after being rebuilt, I would really like to know what causes it to get in that mode because it is so frustrating! Does anyone have any ideas or other tests I can try if it happens again?
 
At this point this is probably a dumb question but what is the mechanism that kickstarts the coils oscillation? Is it the interrupter then sends a short pulse that kickstarts the coil which in turn resonates and is picked up by the CT feedback?

[davekni]

Yellow = CT feedback. You can see I am zoomed in one short pulse. For some reason the CT or the 74hc14 just stops producing feedback?

Does anyone have any ideas or other tests I can try if it happens again?

Scope capture will be more helpful if other traces are identified too.  I'd rather not spend time speculating.

At this point this is probably a dumb question but what is the mechanism that kickstarts the coils oscillation? Is it the interrupter then sends a short pulse that kickstarts the coil which in turn resonates and is picked up by the CT feedback?

Here's a post where I describe startup.  I've written others too, but this is the one I found in a search:
https://highvoltageforum.net/index.php?topic=840.msg5667#msg5667

[ZakW]

After a bit of a break I managed to get my board working for the most part. I am feeding an Arduino a low voltage AC signal that it tracks to detect the zero crossing. From there I can adjust the delay to fine tune the output. However, I am running into a similar issue that I was dealing with when using the optocoupler. The coil does not seem to have an output at the actual start of the zero crossing. Instead the soonest I can get the coil to have an output is if I adjust the interrupter to about ~1ms into the positive AC half cycle. This was not as big of an issue in my previous versions but has become more and more apparent.

Here are some scope shots to illustrate what is happening.

Yellow trace = low voltage AC reference
Purple trace = CT output
Blue = Low side Gate/Source

Here is the interrupter tuned to the zero crossing.



Here the interrupter output is delayed ~1ms into the half cycle.



I feel that if the coil is turning on 1ms into the half cycle that it is having a negative impact on the output since the arc is not growing from the beginning of the ramp. In the zoomed in version of the zero crossing scope capture, it appears that the CT feedback just sort of dies out. I did try adding 5.6Kohm resistors across the bridge capacitors. It did not seem to make a difference.

Any thoughts or ideas would be appreciated.

[davekni]

I feel that if the coil is turning on 1ms into the half cycle that it is having a negative impact on the output since the arc is not growing from the beginning of the ramp.

Coils need some voltage for an arc to start.  1ms may be too high for your coil.

In the zoomed in version of the zero crossing scope capture, it appears that the CT feedback just sort of dies out. I did try adding 5.6Kohm resistors across the bridge capacitors. It did not seem to make a difference.

5.6kohm should be fine for balancing half-bridge voltage.  I suggest leaving them in just so bridge balance doesn't pop up as an issue in the future.

The coil does not seem to have an output at the actual start of the zero crossing.

Coil cannot start at 0V unless you use a self-oscillating driver (PLL or the self-oscillating mod's I've suggested several places on the forum).  However, it may be able to start at sufficiently low voltage without driver change.  You may need a stiffer load across the entire half-bridge supply.  The reason you see an initial brief burst at 0V is because remaining charge on half-bridge supply caps (1uF if I recall correctly) is sufficient to power coil very briefly.  When line-crossing is set to start at say 0.5ms, that brief higher power burst may pull voltage down to 0 causing oscillation to drop out.  Try adding a power resistor (perhaps a 4W or 7W incandescent bulb) across half-bridge supply.  That will help discharge 1uF caps between half-cycles.  Perhaps then coil will start before 1ms.

Thank you for the clearly defined scope traces.

[ZakW]

Coils need some voltage for an arc to start.  1ms may be too high for your coil.

By too high, do you mean 1ms might be to long of a delay? Like the arc should happen sooner into the cycle? I did pick up a differential probe so I can check out the bridge during normal operation.

5.6kohm should be fine for balancing half-bridge voltage.  I suggest leaving them in just so bridge balance doesn't pop up as an issue in the future.

I will leave them on. The pictures in my previous post were before I added them. I wonder if that initial spike is gone now. I can check later this week.

Coil cannot start at 0V unless you use a self-oscillating driver (PLL or the self-oscillating mod's I've suggested several places on the forum).

The mod you are referring to is the resistor across pin 1&2 (Something that is in the ballpark of the res frequency) of the HC14, right?

You may need a stiffer load across the entire half-bridge supply.

What do you mean by stiffer load? I only have two turns of thick wire for the primary.

The reason you see an initial brief burst at 0V is because remaining charge on half-bridge supply caps (1uF if I recall correctly) is sufficient to power coil very briefly.  When line-crossing is set to start at say 0.5ms, that brief higher power burst may pull voltage down to 0 causing oscillation to drop out.

I could test/observe this with my differential probe, right? If I connect it across the primary and adjust closer to zero crossing. If what you speculate is true I should see a corresponding voltage spike and then nothing?


I understand that the voltage has to increase enough before there will be an output but I am still hung up on why the coil does not output anything if the delay is adjusted closer to or at the zero crossing? It is because the interrupter is only making the driver output for a split second and if there is insufficient voltage on the bridge the coil doesn't produce an arc so there is no feedback via the CT and the loop fails to start?

If the driver is only producing that single pulse and not a continuous one,  I could add a resistor across to the HC14 to make it self oscillate which might make it switch on sooner into the half cycle? I am just guessing at this point.

Thank you for the clearly defined scope traces.

Happy to provide them, I appreciate the help!

[davekni]

By too high, do you mean 1ms might be to long of a delay? Like the arc should happen sooner into the cycle? I did pick up a differential probe so I can check out the bridge during normal operation.

Yes.  I'm just agreeing with you that voltage at 1ms may be more than necessary for arc to start, so may not be optimal for long arc performance.

The mod you are referring to is the resistor across pin 1&2 (Something that is in the ballpark of the res frequency) of the HC14, right?

Yes, though slightly more change needed for current feedback.  Referring to your schematic in reply #4 of this thread:  Resistor needs to go from HC14 pin 2 back to right side of C3 (left side of R1) rather than directly to pin 1.  C3 will need to be a much smaller value to achieve frequency with reasonable resistor value.  And will likely need an additional resistor across CT secondary (burden resistor for CT, perhaps 1kohm) because the small value of C3 will prevent R1 from being the CT burden resistor.  Several people have used this self-oscillation patch with antenna feedback.  Others have used it with UD2.7 or similar current feedback (where there is already a CT burden resistor and feedback is from primary TC coil current).  Don't know of anyone trying it yet with 74HC14 and current feedback from TC secondary.  Should work, but might take a bit of value experimenting.

What do you mean by stiffer load? I only have two turns of thick wire for the primary.

By stiffer I mean lower value resistance than the two 5.6kohm resistors in series.  However, perhaps 11.2k is sufficient if there is a long enough time between line half-cycles being applied to bridge.  Your scope test will show that.

I could test/observe this with my differential probe, right? If I connect it across the primary and adjust closer to zero crossing. If what you speculate is true I should see a corresponding voltage spike and then nothing?

Yes, though you can already see that with your previous scoping of CT output and low side Vgs (or GDT input).  Just repeat that with 5.6kohm resistors in place.  And/or include your suggested differential probe measurement too.

I understand that the voltage has to increase enough before there will be an output but I am still hung up on why the coil does not output anything if the delay is adjusted closer to or at the zero crossing?

Per your previous CT output trace, the coil is operating very briefly, likely creating a very tiny faint arc that isn't even noticed, or perhaps not quite enough voltage to actually arc.  That brief operation is powered by energy stored in 1uF bridge caps.  With 5.6kohm resistors in place, that brief operation may go away.  If so, no need for additional load resistor across bridge supply.

If the driver is only producing that single pulse and not a continuous one,  I could add a resistor across to the HC14 to make it self oscillate which might make it switch on sooner into the half cycle? I am just guessing at this point.

Self-oscillation at a close-to-correct frequency will almost certainly fix the startup issue.  However, just a bit complex as discussed above.

[ZakW]

Yes.  I'm just agreeing with you that voltage at 1ms may be more than necessary for arc to start, so may not be optimal for long arc performance.

I don't recall this being as much of an issue with previous designs but at the same time nothing is jumping out at me in regard to any big changes that would result in such a difference in operation. I have always had to adjust the timing to tune it but each design became more and more sensitive to this issue. I assume it is an error in my schematic or a mismatched component value that is causing this.

Yes, though slightly more change needed for current feedback.  Referring to your schematic in reply #4 of this thread:  Resistor needs to go from HC14 pin 2 back to right side of C3 (left side of R1) rather than directly to pin 1.  C3 will need to be a much smaller value to achieve frequency with reasonable resistor value.  And will likely need an additional resistor across CT secondary (burden resistor for CT, perhaps 1kohm) because the small value of C3 will prevent R1 from being the CT burden resistor.  Several people have used this self-oscillation patch with antenna feedback.  Others have used it with UD2.7 or similar current feedback (where there is already a CT burden resistor and feedback is from primary TC coil current).  Don't know of anyone trying it yet with 74HC14 and current feedback from TC secondary.  Should work, but might take a bit of value experimenting.

Thanks for the explanation. If adjusting the input of the Arduino does not correct the issue I will mess around with some values to see if I can get the self oscillation to work. Based on what you said is the additional resistor value the primary factor in determining the frequency?

By stiffer I mean lower value resistance than the two 5.6kohm resistors in series.  However, perhaps 11.2k is sufficient if there is a long enough time between line half-cycles being applied to bridge.  Your scope test will show that.

the 5.6kohm resistors are 5w resistors. Besides that, I only have 1/4w resistors. I could try to put two in parallel to increase wattage while aiming to reduce the resistance. I will determine if this is necessary after I scope the bridge output to confirm if the spikes are still there.

Per your previous CT output trace, the coil is operating very briefly, likely creating a very tiny faint arc that isn't even noticed, or perhaps not quite enough voltage to actually arc.  That brief operation is powered by energy stored in 1uF bridge caps.  With 5.6kohm resistors in place, that brief operation may go away.  If so, no need for additional load resistor across bridge supply.

This is exactly what is happening. There is a very thin white snapping spark that occurs if the timing is not adjusted correctly. For reference it can be heard in the video from my previous post (https://youtu.be/pWb63VKzJO8)

I suspect the issue a combination of two problems:
#1 this type/design is sensitive to timing the on pulse given that it is powered by the half cycle ramp (duh, thinking out loud) but the driver is not continuously generating an output.
#2 my interrupter is not using a buffer for the AC input/ZCD. I am currently feeding the AC signal directly into the analog input of the Arduino and might get better results by implementing the ZCD section from the original schematic https://www.loneoceans.com/labs/sstc3/schema_sstc3_staccato.jpg and feeding the output of that into the Arduino. Previous designs used an optocoupler with the output being directly connected to the 555 trigger pin via a small cap that had similar results.

[davekni]

Thanks for the explanation. If adjusting the input of the Arduino does not correct the issue I will mess around with some values to see if I can get the self oscillation to work. Based on what you said is the additional resistor value the primary factor in determining the frequency?

For antenna feedback, yes, resistor value is the primary factor, presuming antenna capacitance is constant.  For current feedback, C3 capacitance and resistor value determine frequency, presuming a burden resistor is added across CT and its value is low compared to added self-oscillation resistor.  In either case, 74HC14 hysteresis voltage also matters, so if 74HC74 part is replaced frequency may change some.
Another thought that I was going to edit into my last reply:  It may be sufficient to just add a high value resistor (1 to 10meg) from 74HC14 pin 2 to C3/R1 (or even to HC14 pin 1) without changing any other values.  That should self-oscillate at a low frequency, much too low to drive coil, but sufficient to keep 74HC14 pin 1 biased close to threshold so sensitive to smaller CT feedback signal amplitude.  (Or two resistors to form a voltage divider between ground and P5V to bias 74HC14 pin 1 around 2 to 2.5V.)
Yet another option is to add a resistor from interrupter input to C3/R1 connection as in Steve Ward's DRSSTC-0.5 schematic.  Steve used 10k for that resistor.  Provides one initial edge on interrupter enable to help start oscillation.

I could try to put two in parallel to increase wattage while aiming to reduce the resistance.

If scoping shows lower resistance is needed, leave the two 5.6k resistors in place, then add new 5.6k resistor(s) across full supply voltage.  More effective than pairs of resistors in series.

#2 my interrupter is not using a buffer for the AC input/ZCD. I am currently feeding the AC signal directly into the analog input of the Arduino and might get better results by implementing the ZCD section from the original schematic https://www.loneoceans.com/labs/sstc3/schema_sstc3_staccato.jpg and feeding the output of that into the Arduino.

Are you seeing jitter in timing?  Probably no need to change ZCD unless existing circuit is generating unstable timing.  (I don't personally use Arduinos, so don't know exact capability of an "analog input".)

[ZakW]

For antenna feedback, yes, resistor value is the primary factor, presuming antenna capacitance is constant.  For current feedback, C3 capacitance and resistor value determine frequency, presuming a burden resistor is added across CT and its value is low compared to added self-oscillation resistor.  In either case, 74HC14 hysteresis voltage also matters, so if 74HC74 part is replaced frequency may change some.
Another thought that I was going to edit into my last reply:  It may be sufficient to just add a high value resistor (1 to 10meg) from 74HC14 pin 2 to C3/R1 (or even to HC14 pin 1) without changing any other values.  That should self-oscillate at a low frequency, much too low to drive coil, but sufficient to keep 74HC14 pin 1 biased close to threshold so sensitive to smaller CT feedback signal amplitude.  (Or two resistors to form a voltage divider between ground and P5V to bias 74HC14 pin 1 around 2 to 2.5V.)
Yet another option is to add a resistor from interrupter input to C3/R1 connection as in Steve Ward's DRSSTC-0.5 schematic.  Steve used 10k for that resistor.  Provides one initial edge on interrupter enable to help start oscillation.

Thanks for the suggestions. I will try and see what works. It shouldn't be too hard to add a resistor and cap here and there.

If scoping shows lower resistance is needed, leave the two 5.6k resistors in place, then add new 5.6k resistor(s) across full supply voltage.  More effective than pairs of resistors in series.

From Page 1, my first post.


I have a 5,6kohm resistor across C19 & 20. Are you saying to add an additional resistor across C18 instead of decreasing the resistance across C19&20?

Are you seeing jitter in timing?  Probably no need to change ZCD unless existing circuit is generating unstable timing.  (I don't personally use Arduinos, so don't know exact capability of an "analog input".)

I am seeing the output of the Arduino move relative to the AC input, almost a scrolling effect depending on what signal I am triggering the scope on. That is why I might try connecting the single 2n3906 ZCD output to the input of the Arduino. I thought it might be having a hard time sampling or tracking the AC voltage. I have virtually zero experience when it comes to Arduino so I am not aware of any pitfalls or limitations of implementing this kind of thing.

[davekni]

Are you saying to add an additional resistor across C18 instead of decreasing the resistance across C19&20?

Yes.  Given fixed value of 5.6k, will reduce voltage faster that way, across C18.

I am seeing the output of the Arduino move relative to the AC input, almost a scrolling effect depending on what signal I am triggering the scope on. That is why I might try connecting the single 2n3906 ZCD output to the input of the Arduino. I thought it might be having a hard time sampling or tracking the AC voltage. I have virtually zero experience when it comes to Arduino so I am not aware of any pitfalls or limitations of implementing this kind of thing.

"Scrolling effect" sounds more like a problem with Arduino code than with input signal processing.  I'd guess the AC input pin is being sampled to infrequently.

[ZakW]

"Scrolling effect" sounds more like a problem with Arduino code than with input signal processing.  I'd guess the AC input pin is being sampled to infrequently.

Here is a video to show what is happening:

Yellow= Arduino Interrupter output
Blue= ZCD output (going into the input of the Arduino)
Purple= reference AC signal from small transformer

/>

I added a simple ZCD circuit (a single 2n3907) that now feeds a square wave into the input pin of the Arduino. That seems to have allowed me to get ever so slightly closer to zero crossing since the arc length has improved a bit. The remaining issue now is that scrolling/jitter (not sure what to call it). You can see from the video that it scrolls and then resets. Because the interrupter output has that slight degree of fluctuation or jitter, it's difficult to synchronize it accurately. I've been attempting to fine-tune the timing so that I hit that sweet spot where the coil won't output, will output , or will output but with shorter, more branched arcs.

As for the code, I used ChatGPT 4 to create it. I do not have any coding experience but I do have an oscilloscope, determination, and a dream. After a few failed attempts at creating a ZCD/synced output from scratch I used the code that LoneOceans created here https://www.loneoceans.com/labs/sstc4/StaccatoAttiny.ino. The piece I was missing is the fire() portion. I bashed that together with what I had created so far to match my inputs/outputs and I got this:

My code here: UNFINISHED This likely contain erroneous notes and text from using ChatGPT. The goal is to eventually review the code and clean everything up.

I can adjust the BPS and PW with two separate potentiometers. The phase delay pot was necessary to get the output to line up closer to the zero crossing. Without it the phase is off by a bit.

#define TRIGGER_PIN A0
#define OUTPUT_PIN 5
#define LED_PIN 3
#define POTENTIOMETER_PIN_BPS A1
#define POTENTIOMETER_PIN_PW A2
#define POTENTIOMETER_PIN_DELAY A3
#define POTENTIOMETER_SCALE 10

int bps = 0.5;
int pw = 4000;
int delayUs = 0;
int phaseDelayUs = 0;
bool triggerState = false;
bool triggerActive = false;

void fire();

void setup() {
  pinMode(OUTPUT_PIN, OUTPUT);
  pinMode(LED_PIN, OUTPUT);
  pinMode(TRIGGER_PIN, INPUT);
}

void loop() {
  int bpsReading = analogRead(POTENTIOMETER_PIN_BPS);
  int pwReading = analogRead(POTENTIOMETER_PIN_PW);
  int triggerReading = analogRead(TRIGGER_PIN);
  int delayReading = analogRead(POTENTIOMETER_PIN_DELAY);

  bps = map(bpsReading, 0, 800, 1, 40);
  pw = map(pwReading, 0, 1023, 1000, 9000);
  bool currentTriggerState = triggerReading > 512;

  if (!triggerState && currentTriggerState) {
    triggerActive = true;
  }

  triggerState = currentTriggerState;

  if (delayReading > 0) {
    delayUs = delayReading * POTENTIOMETER_SCALE;
    phaseDelayUs = delayUs;
  }

  if (triggerActive) {
    fire();
    triggerActive = false;
  }
}

void fire() {
  unsigned long periodMs = 1000 / bps;
  unsigned int onTimeUs = pw;
  unsigned long offTimeMs = periodMs - (onTimeUs / 1000);

  unsigned int delayCorrection = onTimeUs + phaseDelayUs;
  if (delayCorrection > periodMs * 1000) {
    delayCorrection = periodMs * 1000;
  }

  delayMicroseconds(delayCorrection);
  digitalWrite(OUTPUT_PIN, HIGH);
  digitalWrite(LED_PIN, HIGH);
  delayMicroseconds(onTimeUs);
  digitalWrite(OUTPUT_PIN, LOW);
  digitalWrite(LED_PIN, LOW);
  delay(offTimeMs);
}

The setup works for the most part besides the skipped pulses and jitter. Since I am using pots for the controls once I tune it where I want it it keeps that configuration even after a restart. I feel like I can get more performance out of the coil if I could reduce the variance in the output and get the coil to trigger closer to the zero crossing.

Yes.  Given fixed value of 5.6k, will reduce voltage faster that way, across C18.

This time around I focused more on the Interrupter but I did scope the output and saw that the spikes are still there. I will eventually test it with a resistor added to C18.

Final note, I have noticed the coil seems to have not killed itself when a flashover from the primary to secondary occurs. I have been a lot more careful with my adjustments but so far several strikes have happened when I have it adjusted too close. I wonder if the flashovers were interfering with my old opto style ZCD/interrupter which would subsequently cause a high frequency oscillation to occur that destroyed the bridge.

[davekni]

Here is a video to show what is happening:

Yellow= Arduino Interrupter output
Blue= ZCD output (going into the input of the Arduino)
Purple= reference AC signal from small transformer

Clearly an Arduino timing issue, probably low sample rate as I'd initially guessed.

As for the code, I used ChatGPT 4 to create it.

I'm surprised that any functioning code was generated by ChatGPT.  My thought is to either avoid Arduino or learn how to code it yourself.

[ZakW]

Clearly an Arduino timing issue, probably low sample rate as I'd initially guessed.

After messing with it a bit and reading up on Arduino jitter I think the issue is more nuanced and above my basic level of understanding. I did have some success in removing the delay potentiometer, as well as a multitude of other suggestions. I found when adjusting the delay that I could get the jitter to stop completely. After updating the code it appears to be working within an expectable range. So much so I can't tell when listening to the output.

I'm surprised that any functioning code was generated by ChatGPT.  My thought is to either avoid Arduino or learn how to code it yourself.

It did a good job. It was really quick to implement changes and when it made mistakes I could tell it "There is no output" and it would realize it made an error and update the code. I could then verify it was working via the oscilloscope and I was on my way.

I did a bunch of testing, here are my results:

Resistors across DC Bus caps (C19&20)

• without any resistors I saw a spike of ~56V @ around 100VAC input
• There was a louder snapping/popping with the arc

Yellow=Interrupter pulse
Blue=Primary Output

• With a 4.7kohm resistor (1/4w) across C19&20
• Quieter arc, no snap
• voltage spike reduced to ~22V

Resistor across AC input capacitor (C18)

• 4.7kohm 1/4w got way too hot. Paralleled two 1/4w 11kohm resistors instead
• voltage spike almost completely gone
• Need to replace with a higher wattage resistor at some point

Changes to Arduino Code and ZCD

• Added 100k bias resistor to the base of the ZCD transistor - no change
• I can control the phase by adjusting the pot (burden resistor I guess...?) I have on the output of the AC signal transformer
• Removed bits of code that might cause delays in the output. Implemented direct port manipulation for faster response from the Arduino
• Much less jitter in the output signal now. Not apparent via the output sound or skipped pulses - YAY!

Auto/Self Oscillation Mod

Yellow=Interrupter pulse
Blue=Gate/Source

• Added 1M pot from pin 2 -> C3/R1
• High resistance values over ~300kohm resulted in no oscillation
• measured the output and how close it was to ZC, tuned the pot value for the best performance = 18kohm
• soldered 18kohm in place

Gate/Source pic - you can see the small chirps the driver was outputting due to the 18k resistor



Here is the primary output during self oscillation - there is a brief turn on and then off off again before the coil outputs an arc.



Without Self Oscillation

• Without the 18k resistor the coil had virtually zero output. Every once and awhile it would have a small arc.
• Bringing my hand close to the GDT would cause the coil to run, I assume it was picking up on the 60Hz mains via my hand

Here are a couple pictures of an arc I captured when it was triggering intermittently

Gate drive input, starting and then stopping again


Drain/Source output



Arc Appearance Questions

Secondary Fres = ~470Khz, well within range of sword arcs

The output seems to be consistently forked, like a 'V' shape from the break out instead of a long straight arc. Adding a bigger top load causes more branching and lower Fres. Every 5-6 arcs I will get a nice straight one. Any advice on what might cause the splitting so consistently?

I feel like it has to do with when the coil is starting. If there is too much voltage present when it starts could the arc be splitting instead of staying as a single arc channel? That would coincide with my slightly variable timing on the ZCD/Arduino output.


Miscellaneous Observations

• Smaller coils with thinner wire (41awg is what mine is) are more prone to primary/secondary flash over. I assume this is due to high voltage per turn being induced into the secondary
• When adjusting the primary coil height (usually with a smaller coil) if it is too low the arc will make a screeching sound as the variac voltage approaches full mains AC. Raising the primary up a millimeter at a time and checking until the sound goes away results in the best output. Any further and a flashover will likely occur!

[ZakW]

More Testing!

I was able to get a lot more performance out of my coil by reducing the DC blocking capacitor value (C19&C20).

Here it is running with 0.47uf instead of 1uf. Hitting 18in at times with a 2in secondary.

/>


1uf was my base line for testing but I also tested 2uf and 1.68. All together I tested from 2uf, 1.68, 1, 0.47,0.2uf The lower the cap value the better the output. 0.2uf was causing the variac to thump/knock loudly (assuming that is a result from high peak current draw?). I was worried to push it to full mains voltage. Plus the coil output was getting very long and erratic for my small bench area and was starting to arc to everything.

I ended up going with some 0.47uf WIMA caps that I felt provided a decent improvement and a good balance between performance and stress on the inverter.

I was under the assumption per Mads write up https://kaizerpowerelectronics.dk/tesla-coils/sstc-design-guide/ that the lower the reactance the better? Higher DC blocking cap value = lower reactance...

Per Mads "Being 5 times lower than the resonant frequency, there is no risk at the primary LC circuit resulting in a DRSSTC condition which would destroy the MOSFETs." some quick calculations shows my primary Fres to be around 220kHz, with my secondary running around ~400kHz. Is this why I am seeing such a boost in the output?

Does higher impedance also play a part? I read that QCW coils tend to have higher impedance primaries.

[davekni]

Much less jitter in the output signal now. Not apparent via the output sound or skipped pulses - YAY!

Glad that bit is working well.

The lower the cap value the better the output. 0.2uf was causing the variac to thump/knock loudly (assuming that is a result from high peak current draw?). I was worried to push it to full mains voltage. Plus the coil output was getting very long and erratic for my small bench area and was starting to arc to everything.

My guess:  This might be an interaction of self-oscillating and primary resonance with 0.2uF.  At least one scope capture shows 238kHz self-oscillation.  Might be close enough to primary resonance to start oscillation at that frequency instead of intended secondary resonant frequency.
Unless you make the other changes I'd suggested (CT burden resistor, much smaller value for 0.1uF input capacitor), self-oscillation frequency is likely to be unstable.  I'd suggest either going back to 1meg just to bias 74HC14-1 to center-supply, or make other changes.

The output seems to be consistently forked, like a 'V' shape from the break out instead of a long straight arc. Adding a bigger top load causes more branching and lower Fres. Every 5-6 arcs I will get a nice straight one. Any advice on what might cause the splitting so consistently?

I feel like it has to do with when the coil is starting. If there is too much voltage present when it starts could the arc be splitting instead of staying as a single arc channel? That would coincide with my slightly variable timing on the ZCD/Arduino output.

Too much initial voltage, or perhaps initial start at wrong frequency due to self-oscillation too low, or perhaps not starting at all initially then starting after voltage is a bit too high.  Fixing self-oscillation to be stable would fix latter two of three guesses.  Or delay enable edge until voltage is a bit higher (not counting remaining voltage from previous half-cycle) so that feedback oscillation starts immediately and remains on.  (Perhaps counterintuitive, but a bit more delay in enable start might help.)

[ZakW]

My guess:  This might be an interaction of self-oscillating and primary resonance with 0.2uF.  At least one scope capture shows 238kHz self-oscillation.  Might be close enough to primary resonance to start oscillation at that frequency instead of intended secondary resonant frequency.

Sorry I was not more clear. The testing I did with different DC blocking capacitor values was independent of all the previous scope shots.

When I was testing with 0.2uf for each cap, the primary output voltage was significantly higher than when I used larger cap values. The output of the coil was very impressive but I wanted to dial it back before I broke something. I also wanted to conduct more testing to make sure I was staying within the parameters of the IGBTs. I can post some measurements of the primary output with the current 0.47uf caps as well as after I add some 1uf caps to bring the total to 1.47uf for comparison. Because I saw an increase in the output of the coil, presumably an increase in primary current I was not sure if I was approaching a potential primary resonance issue that might destroy the bridge.

Unless you make the other changes I'd suggested (CT burden resistor, much smaller value for 0.1uF input capacitor), self-oscillation frequency is likely to be unstable.  I'd suggest either going back to 1meg just to bias 74HC14-1 to center-supply, or make other changes.

I think the coil feedback is working with the addition of the 18k resistor. It seems very consistent and is not skipping pulses or snapping erratically. I am happy to explore your suggestions with the lower cap value and additional burden resistor to see if I can get any better results.

Too much initial voltage, or perhaps initial start at wrong frequency due to self-oscillation too low, or perhaps not starting at all initially then starting after voltage is a bit too high.  Fixing self-oscillation to be stable would fix latter two of three guesses.  Or delay enable edge until voltage is a bit higher (not counting remaining voltage from previous half-cycle) so that feedback oscillation starts immediately and remains on.  (Perhaps counterintuitive, but a bit more delay in enable start might help.)

Great suggestion - I did notice that while increasing the voltage via the variac I get much straighter arcs around ~90V AC. Once I hit full mains voltage (120V) they start to split in that V shape. I will mess around with the self-oscillation to see if that corrects the issue.

[ZakW]

Update: Hit 20inches!! That is a personal record  10x the secondary length

/>

Changes:

• Added 1k burden resistor across CT input
• Lowered C3 from 100nf to 1nf. That was a bit of trial and error while using the 1Meg pot.
• Lower the cap value the higher the oscillation frequency
• Measured the pot value right before oscillation cut out (at its highest frequency), 8.6k ohm. Installed an 8.2k ohm resistor from pin 2 to C3/R1

Those changes had a significant impact on the arc straightness. The burden resistor made the signal much neater and more square, while the new cap/resistor pair value oscillation frequency is now much higher than before. Somewhere around ~250kHz I think. There does not appear to be a gap at the start of the coil turning on now, likely allowing the arc to start forming earlier in the cycle.

• Added a slightly larger diameter metal disc, that also improved arc straightness
• Anything larger and the arcs start branching a lot. The load from the output is bringing the frequency down below 400kHz. I am going to wind a new secondary coil with a higher Fres so I can add a larger top load to help stabilize the output without lowering the frequency too much.
• Primary height now has an impact on arc appearance.
  
  • Too high = long but erratic and splitting
  • Too low = splitting and shorter
  • Middle = straightest and longest
  
  Here are some scope shots of the primary output with 0.2uf DC blocking caps installed.
  Blue = Primary output
  
  
  
  For comparison, here is the output with 0.47uf caps which is what is currently installed and working the best.
  
  
  
  
  
  
  Is the output waveform supposed to look like that?
  
  
  
  I decided to remove a couple components as well: All in the name of the longest arc
  
  
  • C18 (1uf) across the Mains input - decreased output slightly and caused a loud snapping sound in the arc. A bleeder resistor helped but also decreased output
  • C18 bleeder resistor - slightly decreased output

[davekni]

Lowered C3 from 100nf to 1nf. That was a bit of trial and error while using the 1Meg pot.

Looks like C3 in the range of 220pF to 470pF would be good, allowing self-oscillation to hit your operating frequency, around 450kHz if I recall correctly.

Is the output waveform supposed to look like that?

I'm not clear on what your input circuit looks like now and where the scope probe is connected.  Here's what I'm picturing, edited from your reply #4 image:



Is that accurate?  If not, what is the circuit?  Is scope probe on CT input (J1 pin 1)?

[ZakW]

I'm not clear on what your input circuit looks like now and where the scope probe is connected.  Here's what I'm picturing, edited from your reply #4 image:

Apologies, after so many updates it has gotten a bit confusing. You are spot on though. Here are some updated schematics with currently installed components. 

Is scope probe on CT input (J1 pin 1)?

The previous scope shots comparing the different cap values were taken with my differential probes connected to the bridge primary output as seen in the above bridge schematic.

For testing the self oscillation frequency I connected my differential probes to Q2 gate/source. There I was be able to see the self oscillation frequency vs when the coil was running.

Looks like C3 in the range of 220pF to 470pF would be good, allowing self-oscillation to hit your operating frequency, around 450kHz if I recall correctly.

I can try reducing it further. I did notice at some frequencies (typically much lower than what it is now) that the current draw from the driver increased to about 35mA while the bridge was off. Normally, it hovers around 13mA. I assumed this was due to the UCC tying to output constantly at the frequency of the self oscillation? It seems to have gone away at higher frequencies, figured I would mention that.

As I was writing this I grabbed a few more pictures. In them you can see the self oscillation running (~145kHz) until the coil feedback (~463kHZ) takes over. Note: coil secondary frequency is higher since this was not a full voltage test, arc loading etc...

Yellow= Interrupter output
Purple= 74HC14 Pin 4 output (self oscillation)
Blue= Q2 gate/source








So in my previous post (https://highvoltageforum.net/index.php?action=dlattach;topic=2338.0;attach=18447;image) the scope shot where I have the differential probe connected across the primary output of the bridge (0.47uf caps test), is the output waveform supposed to look like that? Sort of square-ish? The 0.2uf output looks more like a sinewave too. The output has improved drastically with the lower 0.47uf caps installed, I guess I wanted to make sure it is not killing my IGBTs to be running like that. The voltage spikes are well under 650v.


Next steps so far:

• Try installing a smaller cap to get the self oscillation frequency more in line with the operating frequency of the coil
• Swap out the Arduino Uno for a Nano and test the code to ensure it works the same
• Start redesigning the PCB to accommodate all the changes I have made so far

[davekni]

Apologies, after so many updates it has gotten a bit confusing. You are spot on though. Here are some updated schematics with currently installed components.

Thank you for the schematic update.  Very helpful.

The previous scope shots comparing the different cap values were taken with my differential probes connected to the bridge primary output as seen in the above bridge schematic.

Oh, that makes sense now.  With 0.2uF half-bridge caps, primary voltage swing is about 3x bus voltage, so you are getting close to primary resonance.  That explains why output increases.  Primary current is increasing too.  The step part of the waveform is half-bridge switching.  Remaining sine-wave portion is resonant voltage across 0.2uF caps.  Less when across 0.47uF caps as would be expected.

Might be ideal with a value between such as 0.33uF.  At 0.22uF, looks like half-bridge switching may be slightly past zero current.  At 0.47uF appears to be well before zero current.  Before is better than after, but just barely before is ideal.  SSTCs are not usually designed to hit ZCS conditions.  Tweaking that in would be an extra benefit compared to most SSTCs.  However, doing so requires looking carefully at the highest current/voltage portions of the half-cycle where it matters most, making sure switching doesn't get past zero-current point.

As I was writing this I grabbed a few more pictures. In them you can see the self oscillation running (~145kHz) until the coil feedback (~463kHZ) takes over. Note: coil secondary frequency is higher since this was not a full voltage test, arc loading etc...

Ideally self-oscillation frequency will be close to operating frequency, about 3x higher than now.  330pF would be roughly there.  However, better to go down to 220pF and increase R13 value to get back down to 450kHz.  Better to have some margin away from self-oscillation dropping out.  (Likely going to very high frequency rather than dropping out, too high to get through driver chip.  Could cause problems, so better to stay away from failure threshold.)  Or, if R10 is decreased, that should allow margin for lower R13.  Very-high-frequency mode is likely when the voltage divider ratio of R13 and R10 is high enough to trip 74HC14 threshold without waiting for C3 to charge.  Also, self-oscillation frequency will be more stable farther from that very-high-frequency failure threshold.

I can try reducing it further. I did notice at some frequencies (typically much lower than what it is now) that the current draw from the driver increased to about 35mA while the bridge was off.

Only guess I have is the above-mentioned very-high-frequency oscillation.  That would draw more current.  Was this with larger C3 and low R13?

Edit:  Just noticed this comment and resulting schematic change:

I decided to remove a couple components as well: All in the name of the longest arc
    C18 (1uf) across the Mains input - decreased output slightly and caused a loud snapping sound in the arc. A bleeder resistor helped but also decreased output
    C18 bleeder resistor - slightly decreased output

Likely reason removing C18 improved performance:  C18 "shorts" bridge power (what I call Vbus+) to neutral (Vbus-) for AC (450kHz) current.  That makes C19 and C20 effectively in parallel.  So removing C18 is similar to reducing value of C19 and C20.  Down side is that wiring inductance of bridge power and neutral now become a significant part of the circuit, especially if within an order of magnitude or so of primary inductance.  I'd consider putting C18 back into the circuit and lowering value of C19 and C20.  0.2uF may be fine for C19 and C20 with C18 back in place.
Yes, bleeder resistor is needed to discharge C18 (and C19/C20) between line half-cycles.  Once self-oscillation frequency is close to 450kHz, I'm guessing bleed resistor won't decrease performance as much.  If bleed is still an issue, a couple options are more effective than a plain resistor.  One is to add a large inductance such as primary of a small line-frequency transformer in series with bleed resistor.  That will keep current flowing longer as voltage approaches zero, so bleed more effectively without requiring as low a resistor value.  Another alternative is a constant-current source circuit instead of a resistor.

[ZakW]

Oh, that makes sense now.  With 0.2uF half-bridge caps, primary voltage swing is about 3x bus voltage, so you are getting close to primary resonance.  That explains why output increases.  Primary current is increasing too.  The step part of the waveform is half-bridge switching.  Remaining sine-wave portion is resonant voltage across 0.2uF caps.  Less when across 0.47uF caps as would be expected.

Might be ideal with a value between such as 0.33uF.  At 0.22uF, looks like half-bridge switching may be slightly past zero current.  At 0.47uF appears to be well before zero current.  Before is better than after, but just barely before is ideal.  SSTCs are not usually designed to hit ZCS conditions.  Tweaking that in would be an extra benefit compared to most SSTCs.  However, doing so requires looking carefully at the highest current/voltage portions of the half-cycle where it matters most, making sure switching doesn't get past zero-current point.

That makes a lot of sense now, thanks for breaking that down. I already have some 0.33uf caps in my shopping cart!

Is this what you mean by "slightly past zero current"

Ideally self-oscillation frequency will be close to operating frequency, about 3x higher than now.  330pF would be roughly there.  However, better to go down to 220pF and increase R13 value to get back down to 450kHz.  Better to have some margin away from self-oscillation dropping out.  (Likely going to very high frequency rather than dropping out, too high to get through driver chip.  Could cause problems, so better to stay away from failure threshold.)  Or, if R10 is decreased, that should allow margin for lower R13.  Very-high-frequency mode is likely when the voltage divider ratio of R13 and R10 is high enough to trip 74HC14 threshold without waiting for C3 to charge.  Also, self-oscillation frequency will be more stable farther from that very-high-frequency failure threshold.

I will lower C3 to 220pF and update R13 after I see how much margin I have via pot before the oscillation dies out. If it is not a lot I will try lowering R10.

Only guess I have is the above-mentioned very-high-frequency oscillation.  That would draw more current.  Was this with larger C3 and low R13?

That is correct, that happened when C3 was larger and R13 was lower. Doesn't seem to be an issue now, I was just curious. 

Likely reason removing C18 improved performance:  C18 "shorts" bridge power (what I call Vbus+) to neutral (Vbus-) for AC (450kHz) current.  That makes C19 and C20 effectively in parallel.  So removing C18 is similar to reducing value of C19 and C20.  Down side is that wiring inductance of bridge power and neutral now become a significant part of the circuit, especially if within an order of magnitude or so of primary inductance.  I'd consider putting C18 back into the circuit and lowering value of C19 and C20.  0.2uF may be fine for C19 and C20 with C18 back in place.
Yes, bleeder resistor is needed to discharge C18 (and C19/C20) between line half-cycles.  Once self-oscillation frequency is close to 450kHz, I'm guessing bleed resistor won't decrease performance.

Great! It is nice to understand more of C18s function and how it impacts C19 and C20. I will try reinstalling C18 and lowering C19/20 to 0.2uf again all with bleeder resistors.

[davekni]

Is this what you mean by "slightly past zero current"

Not quite.  Red circles are what I'm talking about, not green "zero".  You are scoping voltage across primary coil, which is the same as across either C19 or C20 during any given half-cycle.  Zero current through capacitor is when voltage slope is zero.  Voltage is roughly flat before switching as you have circled in red.  Hard to say exactly where within that relatively-flat voltage spot current is changing polarity.  In your similar plot with 0.47uF caps, voltage is still slewing rapidly at half-bridge switching points, so current is well before zero.

BTW, at 60Hz line, there is 8.33ms between positive half-cycles, perhaps 11ms from when C18 stops charging from previous positive half-cycle.  C18 + series of C19 and C20 is 1.1 to 1.2uF depending on C19 and C20 values.  Every RC time constant voltage drops by 1/e (1/2.781828).  Three time constants gets to roughly 5% of initial voltage, probably good enough.  11ms / 3 = 3.67ms time constant, or about 3k across 1.2uF.  About 2.5W for 3k.  If 2.5W is too much, that is where current source or series inductance would help, allowing lower power and better discharge.

[ZakW]

Is this accurate? Made using paint so it is not pretty. I wanted to make sure I understand the waveforms and where things are occurring.

In your similar plot with 0.47uF caps, voltage is still slewing rapidly at half-bridge switching points, so current is well before zero.

By slewing rapidly, do you mean the spikes after switching? If so, do larger spikes indicate switching well before zero current? Like the IGBTs are switching off while current is flowing hence the spikes?

BTW, at 60Hz line, there is 8.33ms between positive half-cycles, perhaps 11ms from when C18 stops charging from previous positive half-cycle.  C18 + series of C19 and C20 is 1.1 to 1.2uF depending on C19 and C20 values.  Every RC time constant voltage drops by 1/e (1/2.781828).  Three time constants gets to roughly 5% of initial voltage, probably good enough.  11ms / 3 = 3.67ms time constant, or about 3k across 1.2uF.

Thank you for this. I am a bit confused by what you are suggesting - are you saying to use a single 2.5W 3k resistor across C18 or 3k total across all caps (C18/19/20) or something else?

About 2.5W for 3k.  If 2.5W is too much, that is where current source or series inductance would help, allowing lower power and better discharge.

2.5W being too much as in 3k ohm is too low allowing too much current to flow, right?

Could you elaborate on 'current source and series inductance would help'?


Testing Yesterday

I tested a lot yesterday and somehow forgot to take pictures of the scope along the way.

Self oscillation changes

• changed C3 to 220pf - oscillating now at 450kHz
• Made R10 11k ohms to achieve 450kHz

Bridge changes - 0.2uf

• swapped C19/20 back to 0.2uf
• added C18 back (1uf)
• added 4.2k bleeders for C18/C19/20

Arcs grew very branched as I got close to 120V. I always notice really nice straight arcs around 100V. Pass that and things get branched. Lowering the primary as low as it could go on the secondary seemed to help with arc straightness. Still not as uniform as 0.47uf setup.

Bridge changes - 0.3uf

I figured 0.2uf was too little capacitance so I soldered a third 0.1uf cap to C19/C20 for a total of ~0.3uf

Messing with the primary a bit, I could not seem to get the arcs to be consistently straight either. The primary waveform looked closer to 0.47uf with smaller spikes (maybe an indication that switching was closer to zero current but still before). Slightly lower coupling and moving the primary down still seemed to help with arc appearance. Fres was now around ~350kHz. Straight arcs were having smaller arcs shoot off 90degrees around midway from the main arc (weird shape), still erratic.

A flashover from the ground end of the coil to primary killed IGBTs, even though it was not raised anywhere near where I typically experience flashovers... I have one more pair of IGBTs, working on ordering more from Mouser after this. I can take some scope shots later of to provide more information with the current configuration before I make any more changes.

I suspect the 0.1uf caps I am using for the bridge are not ideal. I ordered them awhile back and am not sure of their specs. I am going to order some WIMA caps that have seemed to work well before.




Do you have any insight into why lower primary height/lower coupling seems to promote straighter arcs?

I would like to wind an identical coil just with fewer turns to see if increasing the Fres helps with arc straightness as well.

Edit: Could the self oscillation frequency be set too high causing issues and potentially the flash over? Since the arc loading is detuning the coil so much?

[davekni]

By slewing rapidly, do you mean the spikes after switching? If so, do larger spikes indicate switching well before zero current? Like the IGBTs are switching off while current is flowing hence the spikes?

I'm talking about just before the vertical transitions you marked in yellow.  On the right plot (0.47uF), slope after transition is close to zero as you marked in red.  Before transition, slope is reasonably steep.  Slope is starting to decrease before transition.  Glitch and ringing after transition makes it a bit harder to figure out exactly what occurs then, but roughly flat as you have shown in red.

Thank you for this. I am a bit confused by what you are suggesting - are you saying to use a single 2.5W 3k resistor across C18 or 3k total across all caps (C18/19/20) or something else?

I'm suggesting a single 3k resistor rated for 2.5W or higher across C18.  (Or could be a bit lower and higher power, such as 2.5k rated for 3W or higher.)

2.5W being too much as in 3k ohm is too low allowing too much current to flow, right?

3k is fine as long as you have a resistor capable of at least 2.5W, and the warm resistor doesn't cause thermal issues for any neighboring components.

Could you elaborate on 'current source and series inductance would help'?

Only reason to consider these is if 2.5W is too much heat.  These two alternatives can discharge C18 while generating less heat.  Down side is complexity.

Edit: Could the self oscillation frequency be set too high causing issues and potentially the flash over? Since the arc loading is detuning the coil so much?

Certainly a possibility. I was wondering about that as soon as you mentioned running at 350kHz.

All this experimenting with relatively low primary capacitance is getting you close to dual-resonant coils.  There are always two resonances, even in normal SSTC designs.  Lower resonance (called lower pole) is normally far below normal SSTC operating frequency of upper pole.  Reduced primary capacitance moves lower pole closer to upper.  Boosts primary current and therefore performance as you have seen.  However, also makes it possible to lock to either frequency.  Rather than continuing with semi-random experimentation, I'd suggest more modeling.  Use JavaTC to estimate inductances and coupling factor.  (Include primary lead length, as that is often significant in high-frequency coils.)  Then use analog simulation (AC frequency sweep) to see the two poles and how close they are together.  Then check actual operation to make sure it is at upper of the two frequencies.  (Pole frequencies can be calculated with math rather than simulated.  Simulation is easier and graphical results more intuitive to understand.)  There are several free simulators, on-line ones and downloadable programs.  My favorite is LTSpice.  Repeat JavaTC and simulation for different primary heights and different cap values.

BTW, I just made my first QCW coil.  Still experimenting with it.  Straight arcs appear to be as much art as science.  Smooth voltage ramp is certainly a requirement.  I read here that humidity and surrounding object placement affect arc straightness too.

[ZakW]

I'm suggesting a single 3k resistor rated for 2.5W or higher across C18.  (Or could be a bit lower and higher power, such as 2.5k rated for 3W or higher.)
3k is fine as long as you have a resistor capable of at least 2.5W, and the warm resistor doesn't cause thermal issues for any neighboring components.

Ordered some 3.6k 5W SMD resistors before reading this. That should do the trick with room to spare.

Certainly a possibility. I was wondering about that as soon as you mentioned running at 350kHz.

I will try adjusting this again to be more in the range of the coil. The coil oscillation SHOULD take over but incase it doesn't is it better to be slightly under the res frequency or above it?

All this experimenting with relatively low primary capacitance is getting you close to dual-resonant coils.  There are always two resonances, even in normal SSTC designs.  Lower resonance (called lower pole) is normally far below normal SSTC operating frequency of upper pole.  Reduced primary capacitance moves lower pole closer to upper.  Boosts primary current and therefore performance as you have seen.  However, also makes it possible to lock to either frequency. 

That makes a lot more sense now, thanks for the info.

Rather than continuing with semi-random experimentation, I'd suggest more modeling.  Use JavaTC to estimate inductances and coupling factor.  (Include primary lead length, as that is often significant in high-frequency coils.)  Then use analog simulation (AC frequency sweep) to see the two poles and how close they are together.  Then check actual operation to make sure it is at upper of the two frequencies.  (Pole frequencies can be calculated with math rather than simulated.  Simulation is easier and graphical results more intuitive to understand.)  There are several free simulators, on-line ones and downloadable programs.  My favorite is LTSpice.  Repeat JavaTC and simulation for different primary heights and different cap values.

I'll take some measurements and plug some stuff into JavaTC. I have LTspice and was working on building a half bridge in it but there is a bit of a learning curve with it.

I am going to go back to using 0.47uf since that resulted in a nice boost to the output while keeping me far enough away from ZCS conditions. Dual resonance is a bit out of the scope for this project. While I know there is additional output to gain I want to keep this as a SSTC and eventually wrap up the project. I might try a DRSSTC or QCW next which will be a big next step for me.

That should give me some additional time to get more familiar with LTSpice and other simulation software. I know you advocate a lot for learning/testing via simulation. 

[davekni]

I will try adjusting this again to be more in the range of the coil. The coil oscillation SHOULD take over but incase it doesn't is it better to be slightly under the res frequency or above it?

I normally recommend slightly under res frequency.  Usually slightly above is fine too.  In your case, risk of hitting second mode of secondary is higher.  That rarely occurs.  However, with primary above bottom of secondary (creating capacitance to middle of secondary), risk is higher.  Any if I'm not mixing up threads, your previously failure experience suggesting secondary second mode as a likely cause.  ("Modes" is to distinguish from "poles".  Modes are analogous to higher modes of a wind instrument with multiple half-cycles along tube.)

I'll take some measurements and plug some stuff into JavaTC. I have LTspice and was working on building a half bridge in it but there is a bit of a learning curve with it.

AC sweeps are in many ways easier than more common transient analysis.  Model just the TC coils (with coupling factor) and capacitors (top load for secondary, C18, C19, and C20 for primary.  Use a voltage source and series resistor to model half-bridge.

I am going to go back to using 0.47uf since that resulted in a nice boost to the output while keeping me far enough away from ZCS conditions. Dual resonance is a bit out of the scope for this project. While I know there is additional output to gain I want to keep this as a SSTC and eventually wrap up the project. I might try a DRSSTC or QCW next which will be a big next step for me.

Sounds like a wise decision.

[ZakW]

I'm not quite clear.  Is the semi-transparent red showing showing what could be added, or what is already there?  Semi-transparent red area on Vbus+ (top of C18 and C19) is definitely helpful.  Output to primary coil is less important.  That current is already sine-wave due to primary inductance.  What does matter more is the path from high-side emitter to low-side collector.  This part of the one primary output node is important.  Current rapidly swaps from one IGBT to the other through the portion of this trace connecting the two IGBTs.  Some of that trace could be widened, the portion across C20 and C19.

Hopefully I implemented your advice correctly. I do have a few questions though.

Output to primary coil is less important.  That current is already sine-wave due to primary inductance.

I kept the primary output connector on the edge of the board and used the extra space to widen the copper pour. I don't think the extra distance to edge of the board will be an issue.

What does matter more is the path from high-side emitter to low-side collector.

I brought the IGBTs closer together than previous versions and also widened the copper pour to lower inductance. I hope there is plenty of copper now to reduce the stray inductance as much as possible.



PCB is 80mmx80mm to match the mounting points on an 80mm PC fan for cooling. Heatsink is located under PCB to save precious board space.

A few questions:

• Should the neutral plane overlap all power (red) planes? I have most areas mirrored but wasn't sure if it was best practice to have the whole area poured like a typical ground plane.
• Is there a problem with having the output power planes physically close together? They are space by 0.6mm. I was not sure if that might introduce crosstalk or result in other signal issues.
• How do the GDT traces look? Q1 are nice and short while Q2 are ran sort or parallel (if that helps) but are a bit longer. Pin 5 is routed via trace to Q2 Pin 3 instead of going through the neutral plane per your advice

Any other design tips or best practices I can implement before ordering these new boards?

Red = Front copper power plane
Blue = back side logic GND or AC neutral

[davekni]

Hopefully I implemented your advice correctly.

Overall the layout looks really good.

Should the neutral plane overlap all power (red) planes? I have most areas mirrored but wasn't sure if it was best practice to have the whole area poured like a typical ground plane.

I'd increase it a bit, up to the center of pins of Q1 and Q2 across the rest of the ECB.  Keep it higher as you have it now from Q2 emitter to the right, at least to the left edge of primary output connector.  (Likely would be fine to extend neutral plane up further under gate circuitry, but that does add unwanted capacitance to high-side gate.)

Is there a problem with having the output power planes physically close together? They are space by 0.6mm. I was not sure if that might introduce crosstalk or result in other signal issues.

Most signals are 170V peak, max of 120Vac.  0.6mm is fine for all these.  The one exception is the lower output node (center of C19 and C20).  Voltage can be higher depending on values of C19 and C20 and how close your operating frequency is to resonance of C19+C20 with primary inductance.  Might be wise to increase clearance around all edges of that one shape, perhaps to 1mm or 1.5mm total instead of only 0.6mm.

How do the GDT traces look? Q1 are nice and short while Q2 are ran sort or parallel (if that helps) but are a bit longer. Pin 5 is routed via trace to Q2 Pin 3 instead of going through the neutral plane per your advice

Looks great!  That bit of trace length difference should not cause any problems.

Any other design tips or best practices I can implement before ordering these new boards?

Looking great!  Just the above small suggestions.

[ZakW]

You're awesome Dave, thank you!

I moved the neutral plane up to the midpoint of Q1/Q2 and added a bit more clearance to the center of C19 & C20 per your suggestion.

Going to do a final pass tomorrow and get the files sent off to have them manufactured.

[davekni]

You're awesome Dave, thank you!

You are certainly welcome!  Glad to be of use.

added a bit more clearance to the center of C19 & C20 per your suggestion.

That entire node (entire copper shape) could use 1-1.5mm clearance, including at the secondary output connector lower pin (both layers), around mounting hole, and gap to other copper shapes.  Probably OK at 0.6mm, but a bit more would be good.  If connector pads themselves are closer together, may not be worth the work of editing the connector part itself in layout tool.  Even 0.6mm will typically handle 500V fine for the limited total use time of hobby equipment.

[ZakW]

That entire node (entire copper shape) could use 1-1.5mm clearance, including at the secondary output connector lower pin (both layers), around mounting hole, and gap to other copper shapes.  Probably OK at 0.6mm, but a bit more would be good.  If connector pads themselves are closer together, may not be worth the work of editing the connector part itself in layout tool.  Even 0.6mm will typically handle 500V fine for the limited total use time of hobby equipment.

I was wondering if the additional spacing would make a difference in the end if the connector pins were only 0.6m anyway. To be safe I updated the part to a 1x3 header and just wont use the middle pin. That should give me 1.2mm total spacing without too much effort. I also increased the pad clearance to the neutral plane to 1mm for C19/C20 as well as the output pins.

[davekni]

I was wondering if the additional spacing would make a difference in the end if the connector pins were only 0.6m anyway.  To be safe I updated the part to a 1x3 header and just wont use the middle pin.

Are the pins themselves 0.6mm spaced?  That would imply a very tiny connector.  Or is it the ECB pads of the connector part that are 0.6mm?  If it is the latter and it became a problem, the pad copper could be cut back a bit with a knife.  That's why I was a bit less concerned about that one location.  However, your 3-pin connector is a great solution.  I've seen that used in commercial equipment for the same reason.

[ZakW]

Are the pins themselves 0.6mm spaced?  That would imply a very tiny connector.  Or is it the ECB pads of the connector part that are 0.6mm?  If it is the latter and it became a problem, the pad copper could be cut back a bit with a knife.  That's why I was a bit less concerned about that one location.  However, your 3-pin connector is a great solution.  I've seen that used in commercial equipment for the same reason.

The connector pins are 3.96mm on center with a 0.8mm gap between them. The thin yellow line is the 1mm clearance I added.





Updated front and back copper planes:

[davekni]

The connector pins are 3.96mm on center with a 0.8mm gap between them. The thin yellow line is the 1mm clearance I added.

Thank you for the images.  Looks great!

I was misunderstanding just due to semantics.  To me "pins" refer to the metal pins of the physical connector.  The ECB shape for the connector includes pads to which the pins are soldered.  It is those ECB pads that have 0.8mm space between them.

[ZakW]

Thank you for the images.  Looks great!

Thanks! Excited to get them ordered. While I wait I can start working on the enclosure.

I was misunderstanding just due to semantics.  To me "pins" refer to the metal pins of the physical connector.  The ECB shape for the connector includes pads to which the pins are soldered.  It is those ECB pads that have 0.8mm space between them.

Good point, apologies for the confusion.

[ZakW]

V7 is up and running!!





Aside from KiCads footprint for the SOT-23 version of the 3904 transistor pins being flipped I haven't found any other mistakes.

So far this seems the most stable of all the builds with a very consistent output.

---------------------------------------------------------------

I am still testing and making adjustments. I wound a new smaller coil to get closer to ~450-420kHz range. I am hitting 20in (50.8cm) but I still need to fine tune C19/C20 as well as the primary placement.

The new coil measures in at 1.73in (44mm) using 44awg wire = Output is roughly 11.6x the length of the secondary 





---------------------------------------------------------------

Couple questions regarding the spikes I am seeing on Q1 & Q2 Vgs:

• Using my differential probe set to 50x, are the spikes really there or just an artifact of scoping/long probe length?
• If they are really present, are they harmful to the IGBTs and how might I reduce them?

Here is a close up of Q1 Vgs




I noticed that the spikes develop over time before eventually going away as the coil ramps back down... Here is the beginning, middle, and end.




Here is the Vgs waveform without power to the bridge:


Finally, here is a comparison of the output of the UCC vs Q1 Vgs over time.
Purple = UCC output (after capacitor)
Blue = Q1 Vgs





Again you can see the spikes get worse in the middle of the on period before disappearing again.

---------------------------------------------------------------

Other than that the performance seems great and I am very happy with it. I just want to make sure this issue isn't a design flaw or something else that could hurt the IGBTs.

[davekni]

Is your positive differential probe connected to gate and negative to emitter?  Or reversed (ie. scoping Veg instead of Vge)?  As shown, Vge rise time is shorter than fall time, which would suggest that D6 and D7 are backwards.
Are you scoping low-side IGBT (one with emitter to neutral)?  That reduces affect of differential probe leads.  Twisting leads together from probe up to clips also helps.

[ZakW]

Hello Dave,

Well I feel silly, D6 & D7 were backwards (+1 PCB error). Thank you!. But the scope shots still look really similar, maybe slightly faster turn off.

I confirmed I am probing positive lead to pin 1 of Q1 and negative lead to pin 3.

For the UCC if it matters I am probing output B after the 3.3uf cap, other lead is going to ground.

Here are some more captures after correcting D6/7.

Q1 Vge without powering the bridge


Q1 Vge Beginning vs Middle



Q2 vs UCC output



I overlaid Q1 and Q2 and made one transparent. If I understand correctly, it looks like Q1 (brighter blue) is turning off before Q2 is turning on. Which is what I am after right?



Here I overlaid Q1 before and after fixing D6/7. Green trace is with the diodes fixed showing the IGBTs turning off more quickly, right?

[davekni]

But the scope shots still look really similar, maybe slightly faster turn off.

Yes, looks to me like slower turn-on too, as it should be.

I overlaid Q1 and Q2 and made one transparent. If I understand correctly, it looks like Q1 (brighter blue) is turning off before Q2 is turning on. Which is what I am after right?

Yes.  Gate signals by themselves don't guarantee turn-off is before turn-on.  Most IGBTs have less turn-on delay than turn-off delay.  I'd guess you have sufficient dead-time (enough extra gate time to accommodate difference in IGBT delays).

The spikes in Vge are most likely caused by Cdg (Miller charge).  They likely occur simultaneously with half-bridge output transition.  That is one reason they are worse at the middle where half-bridge voltage transitions are largest.  I doubt the spikes are causing any problems.

[ZakW]

Yes, looks to me like slower turn-on too, as it should be.

Thanks for confirming that. I am glad it looks correct now.

The spikes in Vge are most likely caused by Cdg (Miller charge).  They likely occur simultaneously with half-bridge output transition.  That is one reason they are worse at the middle where half-bridge voltage transitions are largest.  I doubt the spikes are causing any problems.

I appreciate the context! I tried googling the Miller charge and waveform examples to see if I could find something similar before asking but I was unsuccessful. Glad to know they are not causing any issues.

[davekni]

I tried googling the Miller charge and waveform examples to see if I could find something similar before asking but I was unsuccessful.

Yes, classic Miller charge effect is a flat spot (plateau) in Vge.  However, in cases where Vge has finished transitioning before IGBT output switches, the Miller charge becomes a glitch in Vge.

[ZakW]

Question on grounding my coil while I am working on the enclosure for it. 

For all of my testing I have been using the mains ground via a cheap line filter I picked up on Amazon - https://www.amazon.com/Comimark-Filter-CW1D-10A-T-Suppressor-Power/dp/B07X3C345J/ref=sr_1_3_pp?keywords=ac+line+filter&qid=1694802673&sr=8-3 wiring my Secondary directly to it. I have tested a quick one sheet tinfoil counterpoise but I noticed the arcs were louder (snapping and popping). I assume the overall size was not providing sufficient capacitance to sink currents from the coil?

Ideally I would like to run this coil in several environments, home and work (standard commercial office) mainly. But I am concerned with EMI/RFI in the office area especially using the mains ground (I understand this is not suggested). What are my options? I was going to try testing a larger counterpoise, spray gluing tinfoil to a 18x18in piece of thin plywood to see if the arc appearance and sound matches what I see when I am connected to my mains ground.

Is my basic line filter doing much at these frequencies (~400kHz)?

For running it at the office it would be in a large open conference room not near computers or other office equipment... I also run the coil at very low BPS. I did see a nicer line filter with 2 inductor stages as opposed to the single stage mine has, might that be a better choice?
 

[davekni]

I have tested a quick one sheet tinfoil counterpoise but I noticed the arcs were louder (snapping and popping). I assume the overall size was not providing sufficient capacitance to sink currents from the coil?

Perhaps.  If too small, maybe snapping and popping is due to some small arcs between bottom of secondary and primary, if secondary bottom isn't close to ground potential?  I don't have personal experience here.  I always ground everything to both a large counterpoise and to line ground.

Is my basic line filter doing much at these frequencies (~400kHz)?

I'd guess so.  400kHz is above many power supply PFC input stage switching frequencies.  Filtering applies to line and neutral.  Ground is not filtered.

[ZakW]

Perhaps.  If too small, maybe snapping and popping is due to some small arcs between bottom of secondary and primary, if secondary bottom isn't close to ground potential?  I don't have personal experience here.  I always ground everything to both a large counterpoise and to line ground.

I haven't noticed any arcing from secondary to primary when using the counterpoise. The counterpoise was about 1.5in below the coil base. The popping sounded more like when I add too much smoothing capacitance and I flatten the mains half cycle ramp.

I'd guess so.  400kHz is above many power supply PFC input stage switching frequencies.  Filtering applies to line and neutral.  Ground is not filtered.

I was thinking about making a custom filter for my frequencies but I am not sure at this point. My coil is physically small and relatively low powered compared to other builds I have seen. Will grounding it to mains really be an issue?

Also, is using a counterpoise inconjunction with grounding to mains better than not using a couterpoise? Grouding is still a confusing topic to me.

[ZakW]

After 1year and 10months this project has come to an end! I still plan on providing lessons learned and my final schematic complete with all my edits and reasons for specific part values etc.

I will post a video as soon as I can. Boxing it up has caused some issues with stability. Still working through the problem but for now I am out of IGBTs 

Here are some pictures of the final build as well as a snip from Java TC.

• 44awg secondary 2in (53mm) tall 1.9 (48.26mm) diameter
• 2 turn primary 16awg
• Current output 18.75in (47.6cm)!! After boxing it up I cant seem to get it to hit 20 again, still working on it.
• Little over 9x the secondary length!
• Case dimensions (LxWxH) 4inx4inx4.5in (10.16cmx10.16cmx11.43cm)
• Secondary is sealed inside larger PCV pipe with Epoxy for insulation
• 1 Protoboard and 6 PCB iterations, Number 7 being the final version

Not sure why some images are rotated.




















HUGE thanks to Dave! Without your continued assistance and knowledge I would have not been able to complete this project, thank you.

Also want to give Magneticitist a shoutout. Your RSSTC YouTube videos are what inspired me to start this project in the first place, thank you.

I have learned a lot in the last couple years and I plan on continuing my efforts and start on a DRSSTC next.

[davekni]

Great little coil!  Much nicer packaging than anything I build.

HUGE thanks to Dave! Without your continued assistance and knowledge I would have not been able to complete this project, thank you.

You are certainly welcome!  I'm happy that my advice was useful.

Also, is using a counterpoise inconjunction with grounding to mains better than not using a couterpoise? Grouding is still a confusing topic to me.

I haven't noticed any arcing from secondary to primary when using the counterpoise. The counterpoise was about 1.5in below the coil base. The popping sounded more like when I add too much smoothing capacitance and I flatten the mains half cycle ramp.

Yes, grounding can be confusing.  There are typically grounded objects scattered around a room along with dielectric materials such as walls.  For a small coil away from walls etc., shouldn't be too problematic.  Counterpoise and line ground together should work fine.  Only other possibility that comes to mind for popping sound is that line zero crossing detect circuit is unintentionally sensitive to surrounding fields.  You could check this with scope a ways away from coil, so coil remains packaged.  Scope line voltage with one probe and use the other as an antenna to pick up secondary voltage.  See if bursts continue to start at line zero crossing when popping sound occurs.

[ZakW]

Great little coil!  Much nicer packaging than anything I build.

Thanks! I am really happy with how it turned out. I think I shrunk it down just about as much as possible.

I am also happy with the arc output length compared to the secondary coil.

[ZakW]

Finally, here is a video of it running. I will eventually post my schematic and additional information about the build when I have time.

Thank you, enjoy!

/>