Bridge output

[flyingperson23]

I'm trying to tune zvs on my tesla coil (bsm150 full bridge, 300A ocd, ud2.9), and the way I read to do this was to minimize ringing on the bridge output. Scoping the output got me the attached image, which seems off to me. Shouldn't it more or less resemble a square wave? The signals to the gates were all normal. Any ideas?

It looks like maybe the transistors are turning on too slowly? Which doesn't really make sense because the design is based on franzoli electronics' alphaDRs, and I used the same igbts and 5.1ohm gate resistors he did.

[Mads Barnkob]

Are you testing at a low DC bus voltage? The IGBT output capacitance is constant, which makes it a major contributor to switching spikes at low voltages. Get up around 100VDC and the spikes should percentage wise to inverter output, be much smaller.

Also, please use edit when adding new information, instead of double posting.

[davekni]

Shouldn't it more or less resemble a square wave?

Parasitic inductance of interconnect (wiring and/or ECB and/or bus bars) is causing the non-square wave shape.  Connecting the scope probe and ground clip as close as possible to low-side IGBT collector and emitter will help.  Also helps to keep scope probe ground lead as short as possible and tight against probe body to avoid inductive loops that couple to H-bridge current.  There are tutorials somewhere here about scoping techniques.
If problems persist, images of H-bridge layout and probe placement will help.  And some scope images zoomed in to one or two cycles for better viewing of details.

[Manz]

It would be useful to see primary current and inverter output at the same time to ceck if the ZCS is correct, my coil's inverter output looks almost the same when the phaselead is way off.
Also as Mads mentioned the bus voltage could be a problem if too low

[flyingperson23]

Image 1 shows the gate of one of the upper transistors in yellow and the bridge in blue. Image 2 is c-e of the same transistor, but all the transistors look similar. Image 3 shows current in yellow and bridge output in blue. I have 7m3-223 and 233 inductors and sweeping through both their ranges does nothing noticeable except reduce the spikes seen in image 4, which is what it looks like with no phase lead. Image 5 is a close up of one of the transitions and definitely shows something, but I'm not sure what. All pictures were taken with a bus voltage >100V, and the problem seems to get worse with higher voltage.

[Manz]

Image 3 makes me think there is some kind of phasing problem
The inverter is switching at peak current, it's probably 90° out of phase. What are you using for your CT? A wrong core material and permeability might be the cause of that phaseshift

[flyingperson23]

I used EVR's CT kit with their ferrites.

[davekni]

Image 1 shows the gate of one of the upper transistors in yellow and the bridge in blue.

Do you have a differential scope probe to measure upper IGBT Vge?

Image 3 shows current in yellow and bridge output in blue.

How is current measured?  Do you have a 3rd CT (or 3rd CT second-stage)?  What burden resistors are on each CT?  Is UD2.9 feedback resistance 51 ohms as default?  What resistance is on OCD feedback after diode bridge in UD2.9?  What resistance is on current measurement CT?  What are the CT ratios?  Are the EVR parts just cores or pre-wound CTs?

The inverter is switching at peak current, it's probably 90° out of phase.

Yes, something is quite wrong.  More details are needed to determine if this is a measurement issue or performance issue.  I'm guessing measurement.  Besides above questions, images of probing will help, including where scope probe ground leads are connected and how differential probe wires are routed.

[flyingperson23]

I don't have a differential probe, I just leave my scope ungrounded and only measure 1 non-floating thing at a time. The current is measured with a third 1:100 CT with .25ohm burden. The feedback resistance is 51ohm. The OCD burden is 5.1 ohms, and seems to work fine. The main CTs are cascaded 1:1024, and I'm fairly certain that's not the issue as I tried the inverter with a known working CT from a different coil and it gave the same output. The EVR parts were cores with wire to wind them, but not prewound. It's a little hard to see but I have the scope ground connected to one side of the bridge output and probe connected to the other, and the second probe is connected straight to the output of the CT.

[davekni]

It's a little hard to see but I have the scope ground connected to one side of the bridge output and probe connected to the other, and the second probe is connected straight to the output of the CT.

Certainly accurate to say "hard to see".  Is the upper probe measuring bridge output?  If so, is the probe ground connection around the probe ring normally used for ground lead, or at the probe tip?

Either way, large loops between scope probe tip and ground on lower probe adjacent high current (primary current) wiring (also with large loop area) is likely making measurements way off.

Best to have primary coil wires paired as close together as voltage allows.  Twisted together is great if primary lead wire insulation is thick enough to handle primary voltage.

Scope measurements are best made from opposite side of H-bridge, or at least kept away from primary current wiring (away from MMC and leads to primary coil) and away from primary coil itself.

For measuring H-bridge output, best to connect scope probe ground lead to Vbus- (low side of H-bridge power) and measure one side of H-bridge output.  Avoids high-frequency signal on scope case, which couples to other measurements.  Make a second measurement of other output if you suspect a problem with bridge causing outputs to not be complimentary.

Measuring CT voltage across 0.25 ohm resistor would be best with coax all the way from 0.25 ohm resistor to scope input.  That relatively-low-voltage signal will be sensitive to any scope probe tip-to-ground loops.

[flyingperson23]

I got better results switching measurement positions to be closer. That was borrowing a primary from a different coil, which was far away. Now with the actual primary right on top of it, no measurements positions seem normal, even directly on the IGBT outputs. The image attached shows the output of one transistor in blue and a floating probe next to the primary in yellow, so I'm going to assume it's just the high current throwing everything off.

[davekni]

a floating probe next to the primary in yellow

Around high voltage, floating probes will always pick up significant signal.

Reasonable scope measurements are possible.  Requires minimizing wire loop areas for both high current signals (entire H-bridge output through MMC and primary coil) and for measurement connections (scope probe and scope ground connection.  If primary coil is close to h-bridge, may also require a metal shield plate (or foil) between primary and h-bridge.  Needs to be a minimum of primary-coil-radius below primary coil to minimize blocking primary field necessary for TC operation.

[flyingperson23]

I unfortunately wasn't the brightest designing this. The primary is right above the bridge, so I'll try making extensions and moving it away. I routed the hv cables to try to minimize loop area and there's not much from the mmc to the primary but there is a good bit between the inverter and mmc pointing a field right at all possible measurement points. I'll see if I can move the gdt and reroute stuff but idk if it'll be possible. Assuming I got phase lead and primary tuned with minimal interference, would there be any long term problems running the coil as it is with the interference?

[davekni]

Assuming I got phase lead and primary tuned with minimal interference, would there be any long term problems running the coil as it is with the interference?

If your driver circuitry is on an ECB with good ground planes, probably no issue.  If driver is hand-wired breadboard, may cause issues.  Shielding around driver would help if there are issues.

[Felix M.]

Hi all,
I have some similar problems with my bridge. I'm using a SKM200GB218D fullbridge in combination with a 324nF MMC. The operating frequency is at 80kHz. The driver is an UD 2.7C THT. I have gate resistors with 5.1Ω, a diode for faster discharge and some 33V zener diodes for protection.

After doing some static load test with 64V on the bus, I got this for a bridge output:

(Notice how the maximum voltage reaches double the Vbus)

I'm not too sure how valid the next shot is, because channel 1 on my scope malfunctions and the sine wave is slightly behind to where it should be. It still confirms that the coil is pretty much hardswitching:

Tuning the phase lead with the inductor doesn't work at all


I think gate signal's rise and fall times are also not as fast as they are supposed to be:

It's a 15:15 GDT on a N30 ferrite core. (Here is a link: https://www.mouser.de/ProductDetail/871-B64290A0040X830 )

Here is a shot of the UD 2.7C output with the GDT attached:



Back to the brigde output: Is there any explanation for such a funky looking signal? Any idea on how to fix it?

[davekni]

After doing some static load test with 64V on the bus, I got this for a bridge output:

Is this measurement using a differential scope probe?  Or with either the scope or 64V power supply isolated from line ground?  Even with scope or supply floating, connecting scope probe ground clip to an active signal (bridge output) can cause problems.  Scope or supply internal Y caps pass high frequency current to line.  Unless you have a differential probe, I'd recommend connecting scope ground to Vbus- and measuring one output of the bridge at a time.

If not a scoping artifact, I'd guess the ~400kHz ringing is due to wiring inductance between H-bridge and bulk capacitors resonating with film cap(s) close to IGBTs.  Pictures of your setup are helpful, including showing scope probe connections.

The end of this scope capture also looks odd, as if Vbus were ramping towards 0.  Possible GDT output phasing (lead reversal) issue?

I think gate signal's rise and fall times are also not as fast as they are supposed to be:

It's a 15:15 GDT on a N30 ferrite core.

Probably fine for an 80kHz coil.  Pictures of GDT and wiring would help here.  Reducing GDT leakage inductance (including lead wire inductance) would improve Vge fall time.  Rise time is intentionally a bit longer due to 5.1 ohm gate resistors.  Added delay avoids IGBT cross-conduction (provides dead-time).

[Felix M.]

Thanks Dave

I do have an isolated power supply and was just scoping the output with ground attached to one and the probe head to the other terminal at the bridge. There are also some WIMA FKP sunbber caps on the bridge. I'll do some further test tomorrow and post them along with some pictures of my setup.

[Felix M.]

First of all, here is a picture of my setup:

The inverter bridge is quite huge which is likely the reason for such a wobbly looking signal because of the large stray inductance. A whole redesign of the busbars would be quite a project but if it's the issue, I guess there is no real way around it.

Doing the output scope shots with the ground to the negative rail gives the same signal.


The shots from yesterday were done with 4x 0.47uF of snubber capacitance (2 caps on each IGBT).

I've put the Siemens MKV cap along with 2 more snubbers back in place today and did the shots with them in the setup (not sure if it changed anything)

[davekni]

Doing the output scope shots with the ground to the negative rail gives the same signal.

Looks better to me.  Does the other H-bridge output look similar?   Presuming so, I think you'll be fine with this build.  Output swing is not dropping to half near the end as with previous capture.
What is the H-bridge output connected to for this test?  Is it MMC and primary only?  Or is secondary in place with top load?  Or is there a metal pot or other dummy load on primary instead of secondary?  I don't see the typical higher-frequency ring at the end of the normal frequency ring-down square wave (caused by IGBT output capacitance and primary inductance).  Perhaps induction load is dampening that ring.

The shots from yesterday were done with 4x 0.47uF of snubber capacitance (2 caps on each IGBT).

I've put the Siemens MKV cap along with 2 more snubbers back in place today and did the shots with them in the setup (not sure if it changed anything)

That explains some of the improvement.  With 4x 0.47uF it appears that bus/snubber resonance was at ~400kHz, fifth harmonic of TC frequency.  With more capacitance, bus/snubber frequency is no longer hitting a harmonic of primary frequency.

[Felix M.]

Looks better to me.  Does the other H-bridge output look similar?

The output of the second brick looks almost the same.

The test were performed with the tank circuit at the output of the inverter. The static load was mainly a large isolation transformer along with a mot.


There is some ring-down accuring, I just missed it on the scope shot. I'm really not too sure if more snubber capacitance fixed the wobbly signal, since the scope can now only capture the positive half of the squarewave and you can't see as mutch as in the shot from yesterday. It would be great however 

The next problem is the phaselead:


A friend mentioned that the CTs could cause the phaseshift. However if I lay the sine wave from the current over the gate signal of the IGBTs, it's almost perfectly in the zero crossing. Unfortunately the ferrite peace inside my slot 7 inductor broke which makes tuning impossible right now. Tuning the phase with the inductor before when it was still functional didn't really work. I was never able to get it to the zero point. Bypassing it would also not make a huge difference. If you have any suggestion on the right slot 7 inductor, please let me know.

[davekni]

I'm not too sure how valid the next shot is, because channel 1 on my scope malfunctions and the sine wave is slightly behind to where it should be.

What leads you to believe channel 1 malfunctions?  How do you know that the malfunction is smaller than measured phase lead error?
If channel 1 can't be trusted, then use it as trigger only.  One side of UD2.7 output (one GDT primary wire) is a good trigger source.  Presuming coil behavior is repeatable from one burst to the next, measure other signals sequentially using channel 2:  One H-bridge output, other H-bridge output, current, low-side Vge, GDT input (one side).  If your scope can save and display reference signals, save each channel 2 measurement and display together.
How are you measuring current?  Do you have a separate CT?  What turns ratio?  What core?  What burden resistor?

Inductance can be calculated once phase is more accurately measured.

[Felix M.]

The signal on Ch1 is off by mby 1-5°. The current is measured with an external N87 CT with something around 10 windings and 1,5 ohms. It however doesn't matter if the current is probed with an external CT or across the 51ohm resistor on the feedback.

[davekni]

The signal on Ch1 is off by mby 1-5°.

Even 1 degree is ~35ns.  Are you using a 10x probe?  If so, have you checked compensation (square wave ~1kHz output from scope)?  Could be probe compensation issue (on one probe or the other).  Either that or scope's probe delay settings are way off, if your scope has that ability.

The current is measured with an external N87 CT with something around 10 windings and 1,5 ohms.

10:1 CT into 1.5 ohms will become a problem at higher currents.  Also, if 1,5 ohm resistor is wire-wound, resistor inductance may be causing phase error in current measurement.

It however doesn't matter if the current is probed with an external CT or across the 51ohm resistor on the feedback.

This works well as long as scope is directly across 51 ohm resistor, not across resistor + inductor (for phase lead).  I'd suggest staying with this method, with probe directly across resistor, and with probe ground lead running adjacent probe body/tip to avoid inductive loops that can pick up stray magnetic fields.  Some people wrap ground wire around probe to keep it close.  Or use tape or twist ties or ... to keep ground lead as close as possible to probe body.

[Felix M.]

The one scope probe on x10 mode and the channel 1 got fried because they saw too much voltage (there's some more information in the ud2.7c "sawtooth" output thread). Sine signals are fine to measure and the probe can only be used in x1 mode. (The scope reads just a few milivolts on a 7Vpp signal from my function generator).

I scoped the 51ohm resistor directly with the probe ground at the resistor's ground and the head on the other side. No matter what you try with shielding or reducing probe inductance, the current is still way off. I'll probably buy a new scope soon but until then I'm stuck with the broken channel.

I don't think that the measurement is uncorrect although the channel malfunctions under some conditions because I do have zcs with the Vge signals. For some reason Vce is lagging behind.

[davekni]

The one scope probe on x10 mode and the channel 1 got fried because they saw too much voltage (there's some more information in the ud2.7c "sawtooth" output thread.

OK, now I understand.

I probed the 51ohm resistor directly with the probe ground at the resistor's ground and the head on the other side. No matter what you try with shielding or reducing probe inductance, the current is still way off. I'll probably buy a new scope soon but until then I'm stuck with the broken channel.

Until you get a new scope, use channel 1 for trigger only.  Just because sine waves look like sine waves doesn't mean any useful data can be deciphered.  Use channel 2 to sequentially measure signals including current.

[Felix M.]

Is there any way to do phase lead adjustments with just one channel? (If there would be, I‘m not aware of it)

And what about the slot 7 inductor. I have to order a new one and it‘s better to order the correct one right away.

[flyingperson23]

I (kinda) did phase lead with effectively one channel 2 ways:
1 - look at the bridge output and adjust until theres minimal voltage spikes
2 - look at the primary current under high bus voltage/current, and there are noticeable pulses when the bridge switches. That may be an artifact of my bad probing and/or circuit design, but still it kinda worked

Likely neither of these got it perfect but I've been running the coil hard and it hasnt blown up yet, so better than nothing

[davekni]

Is there any way to do phase lead adjustments with just one channel? (If there would be, I‘m not aware of it)

I've been tried to describe this in two posts above, but apparently not successfully.  Easier in your case because you can use channel 1 for trigger only.  (Or, if your scope has an external trigger input, use that.)  Trigger on a signal with fast edges (not current).  UD2.7 output (one GDT primary lead) is best.  Does not matter that it looks like a trapezoid.  Just trigger on the first edge.  Then measure other signals one at a time using channel 2 (your good channel).

Flyingperson23's ideas will work for finer adjustment once you have a good inductor ordered.  I'd recommend (1) for fine adjustment as preferred to just comparing current and voltage phase.  Initial test as above will get close enough to order an appropriate inductor.

[Felix M.]

I‘m sorry davekni, I just didn‘t get it what you ment with the trigger and how it should help me adjusting the phase. Doing it until voltage spikes at the bridge are minimal would be an option but in comparison to my setup without any snubbers, there are not that many spikes left. Tuning the phase was reducing the Vmax by something around 2V at best. And I can‘t do much in terms of tuning it with the inductor. I‘d be glad to get some more information on what you mean davekni.

[davekni]

Does your scope have the ability to save and display reference waveforms?  Here's a youtube video showing how to do that on a Rigol DS1054z scope:
   

/>Useful part starts at about 3 minutes into video.
Will be a bit harder to measure if your scope doesn't have such ability.  I'll describe one or the other way in more explicit detail depending on your answer.

[Felix M.]

Fortunately my scope has this function. Here are some shots of the inverter's output and the primary current:



Doesn't look that different to me than the shots I did with the broken channel measuring the current but there's probably some difference.

And like before, the gate signal and current appear to match each other quite closely:


Is there any formula or table for the right Slot 7 inductor for specific phase shifts?

[flyingperson23]

https://kaizerpowerelectronics.dk/calculators/ud2-x-phase-lead-calculator/

[Felix M.]

The current is off by almost half of one half cycle. I'm not sure if this wasn't a problem with the CTs all the time.

[Felix M.]

Fortunately my scope has this function. Here are some shots of the inverter's output and the primary current

Yeah, I made the mistake to have my old broken inductor in the system. After re doing the test with the inductor shortet, the phaseshift is at around 1us which makes a lot more sence. I ordered a bunch of new ones but 60uH should be the value I approximately need.
It's still strange to me that it shifts the phase towards hardswitching when there's more inductance added.

[davekni]

I ordered a bunch of new ones but 60uH should be the value I approximately need.

Yes, that value sounds good.  Looks like about 1us of delay, so phase lead needs to shift about -1us (1us earlier).

It's still strange to me that it shifts the phase towards hardswitching when there's more inductance added.

UD2.7 inductance (inductor in series with 51 ohm feedback CT burden resistor) moves H-bridge switching earlier.  I haven't seen any evidence in your posted scope images of different behavior.

[Felix M.]

I did two shots, one with the inductor (at max 42uH) and one with the inductor shorted.

With the inductor:


Without the inductor:


I probably missread something in the shots but without any inductor, it almost couldn't switch off more current. I'd need something around 200uH for it so shift enough. Is this really an issure with the inductor or rather something else? I got a good hint to check my resistors and if they are wire wound but they are all metal film. I ordered such a large inductor with up to 200uH but it seems odd to me that the current and gate signal match but current and inverter is so much off.

[davekni]

I did two shots, one with the inductor (at max 42uH) and one with the inductor shorted.
With the inductor:

This capture looks fine.  However, looks like ~1us of phase lag even with 42uH.  If this is really with 42uH, then you will likely need over 100uH to get proper phase lead.  100-150uH may work fine.  Other option is to try reducing delay through H-Bridge (reducing GDT leakage inductance and perhaps reducing gate series resistor value).

Phase lag can be seen here in H-bridge output voltage.  Voltage increases slightly as current reverses.  IGBT voltage is changing from Vce forward drop to diode forward drop (opposite polarity).  I've circled these steps of the final cycle in your capture.  Easier to see at ground w/o Vbus ringing.
 [ You are not allowed to view attachments ]

Without the inductor:

Second capture is clearly not correct.  That is a risk of sequentially probing.  Either trigger didn't catch same edge or coil behavior changed between the two captures.  If you are using channel 1 for trigger, may be useful to display that channel in spite of distortion.  Would make it clear if trigger is same for the two captures being compared.
Another way to check:  Notice on the first image that current starts at roughly the same time as voltage.  If you panned to the left on the second image, presumably current and voltage would start at different times due to trigger change.  If they do start at the same time, then coil operation must have changed for some unknown reason.
Notice that steps on voltage waveform look roughly the same as on upper image.  Thus actual phase didn't change very much between the two.

[Felix M.]

I didn't have time to do mutch today. The inductors arrived and the big one with the 200uH range shifts the bridge output too mutch. Tomorrow I'll try a smaller one and also rework the GDT connections. Is it important that the cables to the pair of the high gates and the pair of the low gates are the same lenght to have the same inductance or delay? While doing test with the big inductor, I noticed some spikes in the gate signals that reached up to over 40V every time it turned on and off. That's probably a bad thing and I'd like to avoid them.

[davekni]

Is it important that the cables to the pair of the high gates and the pair of the low gates are the same lenght to have the same inductance or delay?

Yes, it is good to have matched GDT lead lengths.  Otherwise either rising or falling edge of H-bridge output will have more phase lag.  Correcting for that (adding phase lead) will cause other edge to have slightly too much phase lead.  (Too much is better than too little phase lead.)

While doing test with the big inductor, I noticed some spikes in the gate signals that reached up to over 40V every time it turned on and off.

I'm presuming Vge spikes are negative.  Negative spikes are common since diode is bypassing gate series resistor.  40V is not likely damaging gates.  However, I'd lower the value of your 33V zener diodes and perhaps change to TVS diodes (zener diodes intended for high peak current).  Other option to reducing negative spikes is adding some series resistance to diode across gate resistor.

BTW, spec for SKM200GB218D shows 3.8ohm internal gate resistance.  Combined with ~40nF of Cge for negative voltages, short Vge spikes external to brick will be filtered to lower voltage internally, so less likely to cause problems.

[Felix M.]

They are positive spikes. I do also have some negative spikes but they are not that big.

[davekni]

They are positive spikes. I do also have smoe negative spikes but they are not that big.

Positive spikes are less common.  Perhaps a scoping artifact (Vce switching coupling into scope probe/ground loop)?  If real, increasing gate series resistance should fix them.  When you get a chance, a scope capture and a picture of probe connection might be helpful.

[Felix M.]

Yeah, scope shots and pics of the GDT will come tomorrow.

[Felix M.]

Well, no scope shots today. After seeing that sine signals on ch1 are in fact shiftet by smth like 60° (they were not a few days before), I stopped doing test since the current also shifts slightly when I adjust the phase and it's pretty pointless doing it with the ref function. New equipment will arrive next week so I'll probably won't do much until then. I did however take some pictures of my setup.

GDT:


Wiring of the GDT between driver box and gates:


Feedback CTs:


Overview:


Overview of the setup with the MMC and the Siemens MKV capacitor in place:


If you see some flaws in the pics, just let me know. I also cut out some unnecessary cable length between driver and GDT to reduce innductance. The spikes I saw yesterday disappeared so that's good.

[davekni]

I did however take some pictures of my setup.

Those detailed images are great.  Leads to two suggestions:

Is your GDT wound from CAT5 cable or something similar containing 8 wires as 4 twisted pairs, with one wire of each pair used for GDT secondary and other wire of each pair for GDT primary?  Presuming so, it is best to keep each of the four primary winding leads as twisted pairs all the way back to driver, paralleling them at driver.  Makes a significant reduction in leakage inductance.  Primary lead inductance counts 4x because it has current of all four secondaries added together.  Here's my little tutorial on GDT winding.  It's for a half-bridge, but easy to extend to full H-bridge:
    https://highvoltageforum.net/index.php?topic=1854.msg13949#msg13949

Only other possible concern is routing of MMC output lead.  Looks like it passes across MMC.  Node between MMC and primary coil is much higher voltage than H-bridge output.  May be OK if wire insulation is good.

[Felix M.]

The wire is some kind of wire (I guess its cat 5) I found at my local junkyard. It has 6 internal wires with 4x2 going to the gates and emitters and 2x2 going to the driver allthough I changed it to 1x2 and just cut one of the 6 wires off so it won't do anything. All cables of the same colour are twisted together after leaving the wire mantle at the core.

Don't worry the cable does not acually reach over the MMC. I guess a normal car jumping cable like the one used in my build can handle up to 5kV because I had a similar one under these conditions in a larger pulse capacitor but normally the primary would be mounted on top of the wooden frame. There is a little feed-through hole in the wooden base plate of the primary where it's fed through. Here's a picture:


It'll look like this later with no coables or conductive materials near the aluminium rails of the MMC. Even the mouting screws are plastic to have no contacts you can accidentally touch.


(Those LEDs are somewhat shieldet in an aluminium profile. I'm curious to see how long they can stand those fields  )

[davekni]

The wire is some kind of wire (I guess its cat 5) I found at my local junkyard. It has 6 internal wires with 4x2 going to the gates and emitters and 2x2 going to the driver allthough I changed it to 1x2 and just cut one of the 6 wires off so it won't do anything. All cables of the same colour are twisted together after leaving the wire mantle at the core.

That isn't CAT5 cable.  Sounds like old wired telephone cable.  You will have significantly lower GDT leakage inductance if you rewind with CAT5 cable (or any such ethernet cable, which has 4 twisted pairs, 8 wires total) per tutorial.  Ethernet cable is common in scrap wire recycling around my area.  Not that expensive to purchase new if necessary.

(Those LEDs are somewhat shieldet in an aluminium profile. I'm curious to see how long they can stand those fields  )

That will depend on how LEDs are driven.  If in series strings, best to add an anti-parallel diode across each LED (or a capacitor across each LED to limit AC pick-up).  Reverse voltage is the most common cause of LED failure.

[Felix M.]

I‘ll try to get my hand on some cat5 cable, in fact I might even have some at home. What causes such a big difference between ordenary cables and cat5? How would you reccomend wiring it? Keep two in paralel or just one cable per gate/emmiter?

It‘ll be quite hard to put a reverse diode at each LED because it‘s just such an LED string with a silicone cover you can put in your rooms but I can add one at the input. 

[davekni]

What causes such a big difference between ordenary cables and cat5?

With 4 twisted pairs, each secondary winding is paired physically-close to one primary winding.  Makes less loop area for leakage inductance.

How would you reccomend wiring it? Keep two in paralel or just one cable per gate/emmiter?

If the tutorial isn't clear, please post a question on that thread.  Then answers will help anyone looking at the tutorial.

It‘ll be quite hard to put a reverse diode at each LED because it‘s just such an LED string with a silicone cover you can put in your rooms but I can add one at the input.

If this is a 12Vdc string, the one diode might be useful.  If for line voltage, diode would short half-cycles, so be problematic.

[Felix M.]

It's been a while. After having problems with the driver's PCB (the contacts for the inductor got bad after changing the inductor that often). I redid it and it's finally running again.

Meanwhile I remade my GDT:

I think it should work just fine. The phasing (high and low sides) are also how they should be.

The bridge also received some changes. The old design measured over half a meter in width. I was able to immensely reduce this size and thus the stray inductance

Old bridge (without the Siemens MKV film cap and the 5th and 6th snubber):


View on the bottom of the new busbars:


Overview of the whole finished bridge:


I'm going to do some test tomorrow after work with my new scope and differential probe to see which Slot7 inductor I need.
Please let me know your thoughts about the new bridge and GDT.

[flyingperson23]

For the driver board inductor pads - desoldering the inductor every time you want to change it is a huge pain. Assuming you're using a slot 7 inductor, you can use these: https://www.digikey.com/en/products/detail/mill-max-manufacturing-corp/0555-0-15-01-30-27-10-0/158204
They can be a little tricky to solder but once they're installed it's super easy to change inductors.
You don't even need to solder the case tabs; the 5 pins in those sockets are enough friction to keep it solidly in place.

[Felix M.]

I‘ll keep them in mind but hopefully this will be the last time soldering in a slot 7 inductor. I‘ll measure my offset between current and bridge out in order to calculate the right inductor directly.

[davekni]

Please let me know your thoughts about the new bridge and GDT.

Both look great!  Hope your testing goes well.

[Felix M.]

Good news and bad news. Good one first: The whole system is running again although there are some things to change like some wiring in the tank circuit as well as at the GDT.

(Channel 3 is just to give me a 0 reference line)
I captured this shot with the slot7 inductor shorted (no phase lead) in order to see how much delay I have:


After then putting in the right inductor I got it almost to zero. It's probably just a matter of fine adjustment:



I used the 7M3-154 with a range of 113-188uH.
(Turns out it's not necessary to buy extra pins when you have old IC sockets you can take appart. It's so much easier to change the inductors so thaks for the advice flyingperson23  )

Bad news is the bridge output signal. After a complete redesign it's still that wobbly. My guess is my 64V 5A lab bench power supply because the red current limit LED lights up at even 600mA which is much less than 5A. Maybe it can't supply the pulse current the coil needs or something like that. The wave also changes when the dummy load is changed so it would mayby just disapper at larger power and load with a propper mains supply but before the thing ever sees 3 phase power, I need to finish my softstarter.

[davekni]

After then putting in the right inductor I got it almost to zero.

This second scope capture shows ~0.7us long pause in H-bridge output transitions.  Looks like the common "triple transition".  Just to be certain, does this same pause show up on both H-bridge outputs individually?  I suggest measuring the two outputs individually (even with a differential probe) just to rule out the possibility that the pause is caused by one output switching before the other.  Presuming both outputs show this pause, its 0.7us length suggests longer dead time than necessary.  Reducing gate resistor value should shorten this pause (move second half of transition earlier).  With that change, phase lead is probably good at this UD2.7 inductor value.

Bad news is the bridge output signal. After a complete redesign it's still that wobbly.

Wobble is due to inductance of bus bar between bulk caps and IGBTs resonating with snubber caps.  It is better.  Notice that the wiggle is now 7th harmonic of operating frequency rather than previous 5th harmonic.  Bus bar inductance is lower so resonant frequency is higher and impedance lower.  This amount of 7th harmonic ringing is very unlikely to cause problems.

My guess is my 64V 5A lab bench power supply because the red current limit LED lights up at even 600mA which is much less than 5A. Maybe it can't supply the pulse current the coil needs or something like that.

Pulse load is likely causing current limit LED to light up.  Supply is hitting current limit during pulse load as it should.  That is not a cause for any concern and not related to 7th harmonic ringing.

The wave also changes when the dummy load is changed

Load makes slight changes in frequency as well as damping.  Probably moving 7th harmonic closer or farther from snubber cap resonant frequency.  Interesting detail to explore if you have the interest, but very unlikely to be of any real concern for coil operation.

[Felix M.]

It's right, the PSU has nothing to do with the wobbly signal. If I charge the bus and do one puls without the psu, it stays wobbly. For the test today I grounded the negative bus rail, attached the probe ground to it and measured each output of the two bricks. This is how it looks:



I have no experience with dead times etc. so you tell me if this looks good or bad. All gates have identical gate resistors and diodes. The wires between them and the GDT are twisted and have the same length.

In addition I captured the GE signal of one IGBT:



There's a hint of the positive spike I saw a few weeks back but it doesn't exceed the zener diode's 33V, it stays at 24V. The rise time was about 730-740ns. I could try changing the gate resistors to make it rise faster.

[Felix M.]

I scoped almost everything possible on both bricks. Negative busbar was grounded, all scope grounds were attached to it. For the one gate with it's emmiter being the output terminal of the IGBT, I used my differential probe.

CH1=Vge1, CH2=Vge2 (Measured on IGBT Brick 1):


CH1=Vge1, CH2=Vge2, CH3=primary current at 51R burden resistor (Measured on IGBT Brick 1):


CH1=Vge1, CH2=Vge2, CH3=primary current at 51R burden resistor, CH4=Brick output with ground ref. (Measured on IGBT Brick 1):


CH2=Vge, CH4=Brick output with ground ref. (Measured on IGBT Brick 1):



CH1=Vge1, CH2=Vge2 (Measured on IGBT Brick 2):


CH1=Vge1, CH2=Vge2, CH3=primary current at 51R burden resistor (Measured on IGBT Brick 2):


CH1=Vge1, CH2=Vge2, CH3=primary current at 51R burden resistor, CH4=Brick output with ground ref. (Measured on IGBT Brick 2):


CH1=Vge, CH4=Brick output with ground ref. (Measured on IGBT Brick 2):


A couple of things I noticed: Why does Brick 2's output look so much smoother and less wobbly than the one from Brick 1? Are the bricks output signals supposed to lagg behind that much (t on+t off)? Are current and Vge supposed to be out of phase like that? Could the problem just be one malfunktioning brick (likely Brick 1)?

[davekni]

A couple of things I noticed: Why does Brick 2's output look so much smoother and less wobbly than the one from Brick 1? Are the bricks output signals supposed to lagg behind that much (t on+t off)? Are current and Vge supposed to be out of phase like that? Could the problem just be one malfunktioning brick (likely Brick 1)?

Yes, I'd guess that brick 1 has a problem.  It is much slower than brick 2.  Brick 1 may have more problems too, but for sure it is much slower than brick 2.
Brick 1 problems might also be causing shoot-through current when it switches, which may be exciting that 7th harmonic ring more than normal operation would.

[Felix M.]

I'll change the brick tommorow and see how it performs. Can you tell if the bricks are legit by this pic?

[davekni]

Can you tell if the bricks are legit by this pic?

I can't tell, but don't see anything that looks suspicious.

Presuming the replacement brick is appropriately fast like brick 2, you may have too much phase lead.  Previous adjustments were trying to compensate for an overly-slow brick.

[Felix M.]

Good thing I have a lot of inductors on hand and can swap them easily

[Felix M.]

I changed the slow brick 1 for a new one with no resoults. Even a second change with a second unused brick wouldn't make a difference. After then putting the gate board of brick 1 on brick 2 and the other way around for the secon board, it yet again didn't change anything. Allways the brick with its output connected to the MMC looked wobbly so I also changed that, no results. I then changed brick 2 (the faster one) and captured those shots:

Gates on brick 1:


Gates and output on brick 1:


Gates, output and current on brick 1:


Rise and fall times on brick 1:



Gates on brick 2:


Gates and output on brick 2:


Gates, output and current on brick 2:


Rise and fall times on brick 2:



Could my low busvoltage cause any wrong measurements in terms of IGBT characteristics? I was told I should go in the 100V+ region (I'm doing test at 64V). And about the brick change: could it be possible that brick 1 was actually ok the whole time and the faster brick 2 was weird for some reason?

[davekni]

Could my low busvoltage cause any wrong measurements in terms of IGBT characteristics? I was told I should go in the 100V+ region (I'm doing test at 64V).

For delay testing as you are doing, current is often more important than voltage.  You could use a longer enable pulse to let current ramp up higher.  Removing or reducing any dummy load on primary coil (ie iron pan or whatever you may be using) also helps current ramp up higher.  Looks like you are already hitting 80A or so.  Not likely to be much different at 300A.

And about the brick change: could it be possible that brick 1 was actually ok the whole time and the fatser brick 2 was weirf for some reason?

Yes, a possibility.  However, that skm200gb128d spec lists typical Td(off) as 620ns at 125C die temperature.  You are much cooler than 125C for this low power testing, so delay should be a little lower, not over twice (1.4us as you show).  Makes me wonder if you have one genuine brick (or at least a faster counterfeit) and remaining bricks are slower counterfeits.

Either way, might as well use two matching bricks and continue with your build.  Will require large phase lead to compensate.  Worst outcome (probably) would be that the bricks fail at higher power.

May be better with larger gate resistors for these slow bricks.  There may be some cross-conduction.  IGBT turn-off is usually slower than turn-on, as is the case for spec speeds on this part.  The glitches in Vce scope traces just before each Vce transition may be due to some cross-conduction (also called shoot-through).

BTW, ignore my previous comment about excessive dead time.  That was based on your earlier differential H-bridge output plot and a now-false presumption that the pause was not due to mismatch of the two half-bridges.

[Felix M.]

I‘ll continue with the two brick I have build in right now and see how they do at higher voltage and power. They are going to be replaced with SKM400 bricks in some time the future. I could break one of the three slower bricks open but I’m not sure about this idea. They were bought from a private person who kept them cold and dry in storage for smth. like 2 years.

The max. current I ever scoped with the longest on time my interrupter can provide was 191A, OCD is set at 200.