My first QCW

[Mathieu thm]

Hi all,

After making several drsstc, the next step was to make a qcw.

Build pictures : https://photos.app.goo.gl/E1QXNwidyRnMS6ri9



My qcw is based on a buck modulator with Finn Hammer ramp generator.

Specs :

• FF200R12KT3 (I use the module's internal diode for the converter diode)
• two 0078912A7 for the core (116 µH, 32 turns with 14AWG)
• 600v 20µF filtering cap (with 3.3k 7w bleeder resistor)

• UD2.7c
• 8x FGH60N60SMD in full bridge configuration
• 168:1 CT
• MMC : 11.7nF 20kV (R75UN239050H3J in 10s3p)
• 14.3x10cm secondary with 0.25mm wire (~360kHz)
• upper pole : 430khz
• 8x28cm toroid
• OCD :just under 100A if I remember correctly 150A

I have a problem with the primary frequency, I have no problem oscillating the primary around 350kHz but above 360kHz it's as if the ud27 was missing pulses, maybe I'm missing something obvious. I couldn't take a screenshot but maybe I will tomorrow.
Another weird thing is that, if i've understood correctly, qcw must operate at the upper pole, so the primary frequency must be above the secondary frequency. I've done some tests despite the fact that i can't go above 360kHz for some reason and i still get a result.

With 180V on the buck converter, with a ramp of about 10ms, I get this (40cm sparks) :



Oscilloscope measurements were made with the secondary coil.
The Ramp (10ms):


yellow : bridge output      purple : primary current


zoom in :





With a little more on-time there are quite big peaks, I don't know if it comes from my probe or from the real output of the full bridge.

[davekni]

Another weird thing is that, if i've understood correctly, qcw must operate at the upper pole, so the primary frequency must be above the secondary frequency. I've done some tests despite the fact that i can't go above 360kHz for some reason and i still get a result.

As you have demonstrated, QCW coils can run at lower pole frequency.  I think most QCW coils perform better at upper pole frequency, and that ideal is to run at upper pole frequency while primary is tuned below secondary.  That ideal requires either a PLL or self-oscillation (as can be mod'ed onto UD2.7).

I have a problem with the primary frequency, I have no problem oscillating the primary around 350kHz but above 360kHz it's as if the ud27 was missing pulses, maybe I'm missing something obvious.

With slight primary frequency increase, operation is switching from ~270kHz (per your scope traces) to perhaps 450kHz (or whatever your upper pole frequency is).  May require more phase lead, and/or increased ramp starting voltage to get oscillation to start reliably.  Also check relative times of ramp start and UD2.7 enable pulse start.

With a little more on-time there are quite big peaks, I don't know if it comes from my probe or from the real output of the full bridge.

Voltage spikes are likely real.  Current spikes are likely not real, an artifact of voltage spikes coupling through CT capacitance (or other interference path from voltage to current measurement).  Voltage spikes may be due to insufficient dead time at higher bus voltage and/or insufficient phase lead for higher bus voltage.  I suppose too much phase lead could be an issue instead, but unlikely to have such rapid onset.

[Mathieu thm]

I have a problem with the primary frequency, I have no problem oscillating the primary around 350kHz but above 360kHz it's as if the ud27 was missing pulses, maybe I'm missing something obvious. I couldn't take a screenshot but maybe I will tomorrow.

With slight primary frequency increase, operation is switching from ~270kHz (per your scope traces) to perhaps 450kHz (or whatever your upper pole frequency is).  May require more phase lead, and/or increased ramp starting voltage to get oscillation to start reliably.  Also check relative times of ramp start and UD2.7 enable pulse start.

I retested the primary circuit but without the secondary, and in “normal drsstc” mode.
I've tested several values of variable inductance on the ud 2.7 but it still doesn't work, when I lower the number of turns on the primary coil the circuit doesn't oscillate as it should and the current is completely out of phase.

Around 350kHz (9 turns on the primary) it works well :


At 8 turns :


And 7 turns :


Could it be that the c33 capacitor is the wrong value for 400kHz ? (currently it's a 1nf).

[Mathieu thm]

That ideal requires either a PLL or self-oscillation (as can be mod'ed onto UD2.7).

I looked at the self oscillation mode for the ud2.7, removed R7, put a 50k resistor between pins 3 and 7 of the TL3116 and changed C33 for a 300pF one.

Without powering the bridge, I get this at the output of ud2.7 :



And with the bridge powered up and the primary at turn 7 like last time, I now have a nice waveform as expected ! (the primary oscillates at about 415kHz) :




I have a question, I set the c33 to have 450kHz in “self oscillation”, should I set the c33 to be as close as possible to the primary frequency? Or can I leave it at 450kHz and tune the primary a few tens of kHz around it?

[Anders Mikkelsen]

Nice build!

One thing to look into is your feedback CT. With only 13 turns on the second transformer, the magnetizing inductance is a lot lower than typical builds with 30+ turns on each transformer. This can affect the feedback by modifying the phase and amplitude response of the current sense signal. Not guaranteed to be a problem, but it's good to rule it out as a cause, so I would recommend rewinding your CTs to raise the magnetizing inductance. You could for example use 6 turns for the first transformer and 28 for the second one, giving a similar ratio but more magnetizing inductance compared to 13 and 13 turns.

[Mathieu thm]

Nice build!

Thanks !

One thing to look into is your feedback CT. With only 13 turns on the second transformer, the magnetizing inductance is a lot lower than typical builds with 30+ turns on each transformer. This can affect the feedback by modifying the phase and amplitude response of the current sense signal. Not guaranteed to be a problem, but it's good to rule it out as a cause, so I would recommend rewinding your CTs to raise the magnetizing inductance. You could for example use 6 turns for the first transformer and 28 for the second one, giving a similar ratio but more magnetizing inductance compared to 13 and 13 turns.

I'll look into it thanks

did you say it like that ? (I did 7:24:1) :

[davekni]

I retested the primary circuit but without the secondary, and in “normal drsstc” mode.
I've tested several values of variable inductance on the ud 2.7 but it still doesn't work, when I lower the number of turns on the primary coil the circuit doesn't oscillate as it should and the current is completely out of phase.

Scope traces look like too little phase lead and, for conventional non-self-oscillating UD2.7, too large a C33 value for this higher frequency.

I looked at the self oscillation mode for the ud2.7, removed R7, put a 50k resistor between pins 3 and 7 of the TL3116 and changed C33 for a 300nF one.

Without powering the bridge, I get this at the output of ud2.7 :

Looks ideal!  I recommend staying with self-oscillation.

And with the bridge powered up and the primary at turn 7 like last time, I now have a nice waveform as expected ! (the primary oscillates at about 415kHz) :

Looks good.  Phase lead may be insufficient.  Or phase lead may be OK and current measurement setup is causing phase shift in apparent current waveform.  How are you measuring current?  Across UD2.7 51-ohm resistor?  Or with separate CT?

I have a question, I set the c33 to have 450kHz in “self oscillation”, should I set the c33 to be as close as possible to the primary frequency? Or can I leave it at 450kHz and tune the primary a few tens of kHz around it?

450kHz is ideal.  Goal for QCW is to be close to upper pole frequency which is above primary frequency.  Your upper pole frequency is likely around 450kHz.  Self-oscillation frequency can be a little above upper pole to further encourage starting at upper and not lower pole frequency.
With self-oscillation, hopefully you can leave primary tuned at or a little below secondary frequency and still get upper-pole operation.  That tuning will give best results (as long as operation does not change to lower pole frequency).  Just check scope traces after adding secondary and top load to verify operating frequency.

One thing to look into is your feedback CT. With only 13 turns on the second transformer, the magnetizing inductance is a lot lower than typical builds with 30+ turns on each transformer.

Good point, Anders.  I hadn't looked at CT construction.  Could be the cause of apparent insufficient phase lead while actually contributing to phase lead.

did you say it like that ? (I did 7:24:1) :

Sorry Anders for the bad etiquette of answering for you   Please respond yourself.  But, yes, this CT construction is preferred.
If CT was contributing significantly to phase lead (and apparent lack of sufficient lead), you will need to increase L1 inductance after replacing CT in order to maintain actual phase lead.

[Mathieu thm]

Looks good.  Phase lead may be insufficient.  Or phase lead may be OK and current measurement setup is causing phase shift in apparent current waveform.  How are you measuring current?  Across UD2.7 51-ohm resistor?  Or with separate CT?

I measure the current with a separate CT with a 10ohms burden resistor.

Sorry Anders for the bad etiquette of answering for you   Please respond yourself.  But, yes, this CT construction is preferred.
If CT was contributing significantly to phase lead (and apparent lack of sufficient lead), you will need to increase L1 inductance after replacing CT in order to maintain actual phase lead.

Perfect thanks, I increase L1 and I no longer have any phase shift problems. (the inductance comes from my 160mm drsstc, so I'll see if I can easily make a viriable inductance by rewinding recuperation inductances of the same size as the 7m3.)

450kHz is ideal.  Goal for QCW is to be close to upper pole frequency which is above primary frequency.  Your upper pole frequency is likely around 450kHz.  Self-oscillation frequency can be a little above upper pole to further encourage starting at upper and not lower pole frequency.
With self-oscillation, hopefully you can leave primary tuned at or a little below secondary frequency and still get upper-pole operation.  That tuning will give best results (as long as operation does not change to lower pole frequency).  Just check scope traces after adding secondary and top load to verify operating frequency.

Perfect, thanks for the explanation !
I did some tests with the secondary (and the new CTs) and now the primary oscillates well on the upper pole (about 430kHz at the moment).

I get this with about 150V at the buck input :



The current measurement is very noisy because I use a ct I had on hand of 900:1, I will make other measurements with an appropriate CT.

[davekni]

I get this with about 150V at the buck input :

Looking good!

The current measurement is very noisy because I use a ct I had on hand of 900:1, I will make other measurements with an appropriate CT.

Lower CT ratio will increase signal so improve signal-to-noise ratio.  Keeping CT and CT secondary leads away from H-Bridge and its outputs is good too (except that of course one output wire passes through first CT core).  Shielded wire if you want to get yet cleaner signal.  Bits of foam or other spacers to keep CT primary wire in center of first toroid core also helps as that reduces capacitance from primary wire to secondary wires and cores.

There's some transition occurring around 8ms into ramp.  No voltage spikes before, spikes after.  Would be worth zooming in to waveforms shortly before transition and again after to compare waveforms.  Those spikes aren't likely to be enough to damage anything.  Would still be good to understand cause.  Perhaps either dead time or phase lead isn't quite sufficient for higher bus voltage.

[Mathieu thm]

Lower CT ratio will increase signal so improve signal-to-noise ratio.  Keeping CT and CT secondary leads away from H-Bridge and its outputs is good too (except that of course one output wire passes through first CT core).  Shielded wire if you want to get yet cleaner signal.  Bits of foam or other spacers to keep CT primary wire in center of first toroid core also helps as that reduces capacitance from primary wire to secondary wires and cores.

Thanks for the advice, I'll do it next time.

There's some transition occurring around 8ms into ramp.  No voltage spikes before, spikes after.  Would be worth zooming in to waveforms shortly before transition and again after to compare waveforms.  Those spikes aren't likely to be enough to damage anything.  Would still be good to understand cause.  Perhaps either dead time or phase lead isn't quite sufficient for higher bus voltage.

Yes, I did some tests to see what was causing these peaks, and from what I could test, when I increased the value of L1, the peaks appeared later in the ramp and were smaller.
I'd do more phase lead testing next time and scope captures before and after this transision

[Anders Mikkelsen]

did you say it like that ? (I did 7:24:1) :

Sorry Anders for the bad etiquette of answering for you   Please respond yourself.  But, yes, this CT construction is preferred.
If CT was contributing significantly to phase lead (and apparent lack of sufficient lead), you will need to increase L1 inductance after replacing CT in order to maintain actual phase lead.

No problem, fully agree.

There's some transition occurring around 8ms into ramp.  No voltage spikes before, spikes after.  Would be worth zooming in to waveforms shortly before transition and again after to compare waveforms.  Those spikes aren't likely to be enough to damage anything.  Would still be good to understand cause.  Perhaps either dead time or phase lead isn't quite sufficient for higher bus voltage.

Yes, I did some tests to see what was causing these peaks, and from what I could test, when I increased the value of L1, the peaks appeared later in the ramp and were smaller.
I'd do more phase lead testing next time and scope captures before and after this transision

Can you capture some waveforms before and after this point showing the switching waveforms up close? It could potentially be pole switching or loss of ZVS for some other reason.

[Mathieu thm]

Can you capture some waveforms before and after this point showing the switching waveforms up close? It could potentially be pole switching or loss of ZVS for some other reason.

Yes, of course I'll take scope captures next time (I can only work on the qcw again in 1 or 2 weeks when I'm back home).

[Mathieu thm]

I was able to make some measurements :



before the transition :

after the transition :


I tested with a larger L1 value (I can raise the bus voltage higher before the peaks appear.) :


before the transition :

after the transition :



And with the primary at the turn 9, there is a transition at 10ms on the current :



With the primary at the turn 10.5, there's no more transition on the current (and I can raise the input voltage higher, 200Vac here without hitting the OCD (100 A) :



Regarding the transition on the bridge output, when I increase the value of L1 on UD2.7, the peaks become smaller and appear later.
I couldn't test with a larger L1 value because I don't have one, could it be that the L1 value is too small?

[davekni]

I couldn't test with a larger L1 value because I don't have one, could it be that the L1 value is too small?

Yes, most likely need more phase lead (higher L1 value).  What is your existing L1 inductance?  If it is already ~20uH or more, phase lead is already ~50 degrees or more (at your ~440kHz operating frequency).  90 degrees would be infinite inductance.  Further inductance increases generate diminishing increases in phase lead.  May require figuring out how to reduce delay through H-Bridge, such as reduced gate series resistance (reducing dead time).  And a small improvement can be made by changing HC08 to AHC08 on UD2.7 board.

Thank you for posting detailed well-labeled scope captures.

[Mathieu thm]

Yes, most likely need more phase lead (higher L1 value).  What is your existing L1 inductance?  If it is already ~20uH or more, phase lead is already ~50 degrees or more (at your ~440kHz operating frequency).  90 degrees would be infinite inductance.  Further inductance increases generate diminishing increases in phase lead.  May require figuring out how to reduce delay through H-Bridge, such as reduced gate series resistance (reducing dead time).  And a small improvement can be made by changing HC08 to AHC08 on UD2.7 board.

I replaced the 10ohms gate resistors with 5.6ohms.
I've tested with two different L1, 7m3 123 (9-15µH) and 7m3 333 (25-40µH) but there's almost no difference, peaks are always present.

[Mathieu thm]

Good news :

The problem is half solved.
I decided to set the OCD to 150A and with the primary on turn 9, the peaks are less present and the arc sizes are really big.
I completely forgot to take scope captures, but I do have a pic of the coil in action:
The ramp is about 20ms and end at something like 300V (230Vac buck input). For now, the largest arcs is around 167cm, which is 11.5 times the size of the secondary.




There was a small flashover, but I think it's due to all the primary taps I did during the tests, so I'm going to put new wire with a single primary tap on turn 9 hoping that the flashovers won't be a problem.

[davekni]

I replaced the 10ohms gate resistors with 5.6ohms.

Is that value per pair of FGH60N60SMD, or for each one?  If per pair, 5.6 ohms is likely a good value.  If per device, you may be getting too much overshoot.  Depends on GDT leakage inductance.  Some overshoot may be fine, especially if clamped by TVS diodes on gates.

The problem is half solved.
I decided to set the OCD to 150A and with the primary on turn 9, the peaks are less present and the arc sizes are really big.

450kHz is fast for IGBTs.  Tricky to get clean switching at that frequency.  Sounds like you may be close enough.

I completely forgot to take scope captures, but I do have a pic of the coil in action:
The ramp is about 20ms and end at something like 300V (230Vac buck input). For now, the largest arcs is around 167cm, which is 11.5 times the size of the secondary.

Looks great!

[Mathieu thm]

Is that value per pair of FGH60N60SMD, or for each one?  If per pair, 5.6 ohms is likely a good value.  If per device, you may be getting too much overshoot.  Depends on GDT leakage inductance.  Some overshoot may be fine, especially if clamped by TVS diodes on gates.

There are 5.6ohms for each gate with 33V TVS, the overshoot reached 30v if i remember correctly.

Looks great!

Thanks !

[Mathieu thm]

I haven't managed to completely remove the peaks but it's still reasonable, it doesn't exceed 400-450V.
I was able to test it outside, and I'm very happy with the result. The largest arc reaches 13.6x the size of the secondary (193cm).
Video of the QCW in action : https://www.youtube.com/shorts/3Xm8OgM7NB4
Thanks for the help!

[Mathieu thm]

I ran the coil this evening, and everything went perfectly, except for one problem.
The toroid heated up so much that it collapsed by a few millimeters (toroid is 3d printed). I managed to put it back the way it was before.
At the end of the video you can see that the ocd tripped all the time, I hope it's nothing serious, I hope it's due to a problem with the toroid.
I knew the toroid was heating up from the previous tests, but I didn't think it would heat up so much !

/>