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.