QCW questions

[flyglas]

I love the idea to use the feedback comparator as self oscillation circuit. Nice idea!

[Rafft]

David

initial results:
MAX913 set to 50/50% ratio and 470KHz self-osc

yellow-bridge output blue-Iprim. wick or no-wick, doesnt matter. adding wick though lengthens the ramp-up time.





starting freq. im unsure why it never started at 470KHz. but it does start from the free-run freq because qcw doesnt ignite w/o wick.


and here is end of ramp-up. horrible waveforms.  I wonder though if the feedback did take-over. will try to lower more the MMC to 19.5nF(currently its on 16.3nF, from 14.2nF). and I need to go back to 24uH+50R for the lowpass.


Ralph

[davekni]

done 50/50% duty ratio and adjustable(to some extent) frequency & set to 470KHz

I'm guessing that you reduced R4 in your schematic to get 50% duty cycle.  That is a good tweak.  To make it less sensitive to input bias current, you could reduce values of R1 and R5 proportionately, even down to 600 and 5k.  That lowers impedance, so reduces voltage induced by bias current.  (Might then need to tweak R4, since it's value is compensating for both bias current and output voltage.)

my feedback CT is 81:1 (9:9:1).

My simulated circuit was for 500:1 CT ratio.  81:1 will swamp the self-oscillation too soon.  Reducing R6 to about 1.6 ohms would get feedback amplitude about the same with your 81:1 CT as in my simulation using 500:1 and 10 ohms.

I wonder though if the feedback did take-over. will try to lower more the MMC to 19.5nF(currently its on 16.3nF, from 14.2nF). and I need to go back to 24uH+50R for the lowpass.

I recommend staying with 16.3nF until everything is working there.  Further reduction makes upper-pole starting more difficult.
Your thought about 24uH+50R is along the right lines.  My low 0.1uF inductance was to match zero-delay H-bridge simulation.  You need phase lead still, the same amount as before, since you are using a real H-bridge.  The lack of phase lead is why waveforms look ugly.  Glad IGBTs haven't fried due to that insufficient phase lead.  However, to get the same phase lead with 1.6 ohms requires 24uH*1R6/50R=0.77uH.  If inductor resistance is significant compared to 1.6 ohms, then the 1.6 ohm resistor (1R6) needs to be reduced further so that total resistance of L1+R6 is 1.6 ohms.

starting freq. im unsure why it never started at 470KHz. but it does start from the free-run freq because qcw doesnt ignite w/o wick.

Insufficient phase lead may explain this too.  I'm not certain.

Ooops forgot to mention, that trace was at bridge output, not on igbt gate/s. sorry.

I forgot to look at scope scale too, so didn't guess correctly.  Likely cause of the curved/sloped tops and bottoms is the same - parasitic wiring and/or scope ground lead inductance.

kinda expecting that especially that they seem 'çheap', about 10USD for 5pcs. the other 2 IGBTs 4PF50WD(this is genuine) & 4PC50FD(bought locally YEARS ago, perhaps genuine as well(?)).

Noticed one other think that points to fake:  Spec thermal resistance values suggest a diode die about 25% of the area of the IGBT die.  Picture appears to have a diode over 50% of IGBT area.  That die size ratio would fit the slower MagnaChip part MBQ50T65.  Capacitance measurements also fit MBQ50T65.  Since there are no worst-case AC parameters specified, there is no way to prove counterfeit to get your money back.  (I was surprised how many IGBT and FET specifications have worst-case values for only DC parameters, and just typical values for AC parameters such as switching times and losses, gate capacitance and charge.  Saves cost in production test.  DC parameters are easier to measure.  Makes counterfeiting easy, however.  Designing fast IGBTs requires trade-offs with DC parameters.  It is easy to make a fake part that meets guaranteed DC values but is much slower.)

I love the idea to use the feedback comparator as self oscillation circuit. Nice idea!

Thank you for the compliment.  All my coils are self-oscillating.  Two use this circuit idea, as do many of my other (non-Tesla coil) resonant H-bridge circuits.  When I found this forum 3 years ago, was surprised to find that none of the standard UD1/2 drivers were self-oscillating.

Good luck experimenting!  I'm thrilled to see upper-pole starting tested this way.  Satisfies my curiosity without needing to change my future QCW design plans.

[Rafft]

0.78uH & 1R8 2watt latest mod only



ramp start.sometimes it starts at lower freq like 438KHz.


end of ramp-up.concered of those 100v spikes though


did another zoomed-in capture of ramp-up end(to check Iprim=blue)


spark length are short. MMC value perhaps? AFAIR  10.7uH/14nF had the best spark length(feedback CT, no self-osc). compared to 16.5nF

I was thinking maybe use 24uH/50R for lowpass and a resistor voltage divider after D1&D2.. and/or change C2 to something smaller like 1nF. phase lead still terrible

AND

I think I will change R1 to 600R(as you suggested) for a stronger free-running start-up.


edit#2
I have been playing with the sim just to confirm parts value interaction. and sure enough THEY DO by quiet a LOT! and is a bit finicky to tune... my main problem for today is HOW I would like my feedback to not be strong enough during start-up. here on the waveform can be seen the taking of feedback(430KHz) w/c I have set it on a 50uS delay(on sim sine generator) before going to the comparator(w/c Ive set the freq HIGHER, so it can bee seen clearly on plot). here for sim schematic, Ive taken from UD2.1b. btw , this is a realy elegant solution for having a free-running + feedback switchover.

[davekni]

0.78uH & 1R8 2watt latest mod only

Not sure why, but this image is not showing up.  Just a blank rectangle where the image belongs.  Just tried opening image in a new tab and then it shows up.  Perhaps some issue with the place you are hosting images.  You can post JPG format images directly to this forum.  That is what I do.  Your method is generally working, so no pressure to change.

end of ramp-up.concered of those 100v spikes though

Yes, those are concerning.  Looks like too much phase lead.  However, hard to tell given noise in current trace.  Did something change in the way you are scoping current?  Or in H-bridge output scoping (ie. how scope ground is connected to bridge)?

did another zoomed-in capture of ramp-up end(to check Iprim=blue)

Lower voltage, current, and spikes in this capture.  Still guessing a bit too much phase-lead.  (Was this a slower ramp, or a capture earlier in the ramp when voltage was lower?)

I was thinking maybe use 24uH/50R for lowpass and a resistor voltage divider after D1&D2.. and/or change C2 to something smaller like 1nF. phase lead still terrible

Voltage divider is a fine idea, but needs to be before D1&D2.  Otherwise diodes will clamp feedback signal too low to ever overcome self-oscillation signal.

I have been playing with the sim just to confirm parts value interaction.

Great!  Simulation is very helpful.  However, I'd suggest staying with the topology where AC coupling (C2 in your sketch, C3 in latest simulation) is after clamp diodes.  With this latest simulation schematic, if 1k and/or 470R values need to change to hit 50% duty cycle, both dividers need to change together, adding complexity.

I need to update my simulation too.  Not enough time today.  Realized that my shortcut of simulating a zero-delay H-bridge and almost-no lead compensation misses some details.  For one, phase-lead inductance increases feedback voltage (increases impedance of R+L burden on CT), so causes feedback to kick in sooner.  Separately, I didn't look at phase lead during startup and transition.  To get phase lead at startup, self-oscillation frequency needs to be set slightly higher than upper pole frequency.

In case it is of interest, here's my zipped ltspice source for my previously-posted simulation.  Not updated yet with simulated H-bridge delay and corresponding phase lead:
sstc_ac1.zip

[Rafft]

David

this is just a quick reply. will update later.

did you say 'phase lead too much'? hmmmn havent thought of that. indeed the Iprim trace has 'shifted' more to the right(of the bridge output-yellow trace).

refering to my sketch, I will make a guess its R5(6k) that adds MORE  phase delay to this. how about we make lower like 1k?

I know for sure this will again alter the main comparator self-osc freq and therefore need readjustment.

can you recalculate that part with 0.77uH/1R9 and 81:1 ratio? with using 1k for R5.


================================

Yes, those are concerning.  Looks like too much phase lead.  However, hard to tell given noise in current trace.  Did something change in the way you are scoping current?  Or in H-bridge output scoping (ie. how scope ground is connected to bridge)?

nothing is changed, same all connections..same probe, same scope settings etc. the only thing connected to mains is the scope. my coil is battery-powered. and is it too much? or lack of phase lead though.. maybe its time you include a phase lead in the sims  me im not sure how Id do that in sim.

Lower voltage, current, and spikes in this capture.  Still guessing a bit too much phase-lead.  (Was this a slower ramp, or a capture earlier in the ramp when voltage was lower?)

yes lower bridge output. different capture. I just wanted to see another capture of it, Primary Current.

 

Voltage divider is a fine idea, but needs to be before D1&D2.  Otherwise diodes will clamp feedback signal too low to ever overcome self-oscillation signal.

based on your schematic, changing R5 value(6k) also influences the free-running comparator oscillation. as I have mentioned before, MAX913 R1 R2 C1 R3 R4 alone oscillates on 2.4MHz. adding the rest of the components C2 D1 D2 R5 L1 R6 =  lowers the comparator freq (little board only, not connected to coil) now I will re=check little board AND ADD the CT, and check again the free-running freq(still not connected to coil, but output#7 #8 only)

 

I need to update my simulation too.  Not enough time today.  Realized that my shortcut of simulating a zero-delay H-bridge and almost-no lead compensation misses some details.

yes, if at all possible, good to include phase delay(lead)

To get phase lead at startup, self-oscillation frequency needs to be set slightly higher than upper pole frequency.

by how much higher? my 1st test was on 470KHz, now I removed little board again and did some more mod, and adjusted freq to 485KHz.. would that be enough for higher-than-upper-pole? or maybe go 500KHz?

[davekni]

With delay and phase-lead added, simulation is starting at upper pole even with original 24uH+50R CT burden and 81:1 CT ratio.  Simulation attached at end, and image here:



This is with 270ns simulated gate-driver and H-bridge total delay.  That's what looked about right for 24uH+50R.  Not sure how well that matches reality for your circuit.  I haven't had time to play with values much.  Increasing self-oscillation frequency didn't seem to matter.  (I slightly reduced self-oscillation strength by dropping from 6k to 5k.  That didn't seem to matter either.)

nothing is changed, same all connections..same probe, same scope settings etc. the only thing connected to mains is the scope. my coil is battery-powered. and is it too much?

Hmm.  Increased current noise still seems unexpected.  Could you try scoping the other H-bridge output to see if it matches (opposite phase of course)?  I'm wondering if something partially-fried or whatever, some explanation for the noise increase.  Could you run a coax cable from scope-measurement CT back to scope to minimize noise coupling?

did you say 'phase lead too much'? hmmmn havent thought of that. indeed the Iprim trace has 'shifted' more to the right(of the bridge output-yellow trace).

Yes, that is the shift I was seeing.  Also, the spike makes sense for switch-off at higher current (farther before zero-current).  Parasitic inductance of H-bridge (even just what is inside IGBT leads) will cause a voltage spike as current suddenly transitions from lower IGBT to upper IGBT (and visa versa).

udx2.zip

[Rafft]

David

Thanks again for the simulation

This is with 270ns simulated gate-driver and H-bridge total delay.  That's what looked about right for 24uH+50R.  Not sure how well that matches reality for your circuit.  I haven't had time to play with values much.  Increasing self-oscillation frequency didn't seem to matter.  (I slightly reduced self-oscillation strength by dropping from 6k to 5k.  That didn't seem to matter either.)

Im using 510+510 for R5, so had to retweak R2 C1, for 480KHz & R3 R4 for 50% duty

Hmm.  Increased current noise still seems unexpected.  Could you try scoping the other H-bridge output to see if it matches (opposite phase of course)?  I'm wondering if something partially-fried or whatever, some explanation for the noise increase.  Could you run a coax cable from scope-measurement CT back to scope to minimize noise coupling?

kindly check attached. Im not sure if there is a PC app that can open this. before 'going back to stock circuit', I checked my IGBTs(diode check), all are ok. I dont have a coax cable for I prim, but will make some. need to buy sheilded wire and plug/jack.

Yes, that is the shift I was seeing.  Also, the spike makes sense for switch-off at higher current (farther before zero-current).  Parasitic inductance of H-bridge (even just what is inside IGBT leads) will cause a voltage spike as current suddenly transitions from lower IGBT to upper IGBT (and visa versa).

again, pls check waveforms. pls tell me if it is indeed too much phase lead(how does your 10R/0.1uH compare? sgould I use this values instead?).

going back to my stock is using 22-24uH/51R. our mini-board is 0.77uH/1R9 (and R5=1k). what to do next? I think I see a small difference with using R5=1k(current ckt config) versus R5=6k . R5-1k looks like Iprim cleaned up a bit.

meanwhile here are scope waveforms:
NewFile1 = with the self-oscillationg, R5=1k 1R9/0.77uH and re-tweaked value R2 C1, changing R5 value affects oscillating freq of MAX913
NewFile2 = went back to feedback only , still same 16.5nF on PRIMARY
power supply used for testing both , is 50Vdc
still the same wiring, nothing changed



edit:
oh sh.. I made a mistake on my little board! R5 was connected between L1 and R6! fixxing error

[Rafft]

yet another (corrected)UPDATE 

R5 - 6k
L1&R6 - 0.77uH 1R9(recalculated from 24uH 50R)
free-running 485KHz
MMC 16.5nF

it finally cleaned up(Iprim). my initial question, did it work? (freq takeover)

does L1 R6 value need re-adjusting?  final tweaks?

kindly check attachments. shown are start of ramp, ramp-up peak, and overall ramp

again, same probe hookup

[davekni]

ZIP files didn't open on my computer.  Images are coming through fine.  Yes, scope traces look good now.  (Of course, sometimes things change at full voltage, especially the optimum amount of phase lead.  IGBT delay depends on voltage and current to some degree.)

Appears to be working well.  If I recall correctly, lower pole was ~330kHz.  So I think you are starting at upper pole as intended.

[Rafft]

hi David

they are RAR files. just renamed to ZIP so I could attach it here. inside is .wfm file(about 6Mb). from scope. I think tektronix has a pc app to open that kind of file. there you could zoom-in on the scope waveform. and way better than just a screen grab.
-dont mind downloading PC apps for viewer, they do not work!


I guess the 0.7uH/1R9 is sufficient enough phase lead? will check later with higher bus voltage

-Ralph

edit#1
scope shots with using its power supply, only about 150vdc bus(for testing). spikes on output are manageable?

AND sparks are MORE straighter than before. im not maxxing out my ramp. just wanted a clean ramp-up(and down). cranking it more results in more longer sparks as well. ramp starts from around 470KHz(from a set 485KHz) down to 416KHz. thats a big freq difference, right? I will try changing MMC values higher. lets see if I can bring it to lower pole. but in my previous attempts (w/o the phase lead working) I could not bring it to oscillate in the lower pole.. it seems its getting 'pulled' by the free-running osc to be at upper pole 430KHz..

[davekni]

they are RAR files. just renamed to ZIP so I could attach it here. inside is .wfm file(about 6Mb). from scope. I think tektronix has a pc app to open that kind of file. there you could zoom-in on the scope waveform. and way better than just a screen grab.
-dont mind downloading PC apps for viewer, they do not work!

Now that I know what they are, I likely could decode them.  There are two or three different versions of Tek's .wfm file format over the years.  The PC app for the oldest version doesn't appear to be available any more.  However, I've written a few PERL scripts to decode at least a few specific cases.  That's at my work, not here at home, so would be a bit of nuisance to test on your files.  Don't see any need at this point.  Your images are fine for communicating here.

I guess the 0.7uH/1R9 is sufficient enough phase lead? will check later with higher bus voltage

Looks good on your scope plots.  I think spikes are reasonable.  Might be a bit lower spikes with a bit less phase lead.  However, I suggest not changing.  Much more likely to fry IGBTs from too little phase lead than too much.  It is better to stay with slightly too much phase lead to handle increased IGBT delay when they get hot.

AND sparks are MORE straighter than before. im not maxxing out my ramp. just wanted a clean ramp-up(and down). cranking it more results in more longer sparks as well. ramp starts from around 470KHz(from a set 485KHz) down to 416KHz. thats a big freq difference, right? I will try changing MMC values higher. lets see if I can bring it to lower pole. but in my previous attempts (w/o the phase lead working) I could not bring it to oscillate in the lower pole.. it seems its getting 'pulled' by the free-running osc to be at upper pole 430KHz..

Your start-of-ramp scope capture at 470kHz may be before feedback has taken over fully.  I suspect it gets to your measured 430kHz upper-pole frequency at a bit higher Vbus level, then drops to 416kHz due to arc loading.  I think this is all fine and as expected.  Yes, try larger MMC values to see how they perform and if this driver circuit still manages to start at upper pole.  Of course, upper pole frequency will drop some as MMC increases.  So it is possible that you would need to drop self-oscillation frequency a bit to match.  Also possible that no tweak is needed.

Have fun experimenting!  Great to see that this self-oscillation concept is working in practice, at least so far.

[Rafft]

David

Might be a bit lower spikes with a bit less phase lead.

less phase lead,  meaning a slightly higher value for 0.77uH? or lesser? maybe something like an adjustable 0.1uH - 2uH?

currently the inductor is a basic(no plastic former) coil with a ferrite slug inside. had to adjust it close to 0.77uH(and glue it) prior to installing on mini board. Im planning to make another board for this, with a better adjustable inductor(plastic former). Dont worry about frying IGBTs. Ive even ran it without the phase lead.  I always test with lower Vbus.

Your start-of-ramp scope capture at 470kHz may be before feedback has taken over fully.  I suspect it gets to your measured 430kHz upper-pole frequency at a bit higher Vbus level, then drops to 416kHz due to arc loading.  I think this is all fine and as expected.

I think its already working as intended. I do have to experiment with the tuning (MMC) more. maybe get back to using higher-coupling primary and the bigger topload.

Have fun experimenting!  Great to see that this self-oscillation concept is working in practice, at least so far.

Couldnt have done it without you. once again, Thanks David.

I think this UD2.1b should be named UD2.1R (with free-running osc) 

cheers
-Ralph

[davekni]

less phase lead,  meaning a slightly higher value for 0.77uH? or lesser? maybe something like an adjustable 0.1uH - 2uH?

Less phase lead is lower inductance, perhaps 0.6 to 0.7uH.  However, be careful.  Less may look great in initial testing.  Once you run your coil harder (rapid firing at max voltage), IGBTs get warmer and slow down some.  Slower IGBTs need more phase lead to compensate.  You could test by heating your H-bridge heat sink with a heat-gun (or whatever) to 100C or a bit more, then run short bursts to measure phase lead.

Couldnt have done it without you. once again, Thanks David.

I think this UD2.1b should be named UD2.1R (with free-running osc) 

You are certainly welcome!  Please keep us all posted with your progress.  And, thank you for trying this experiment.

[Rafft]

Hi again

I have been playing with higher coupling+slightly bigger topload versus lower coupling+2stacked smaller toploads.

the goal really for the topload was the 2 stacked ones give me straighter sparks compared to the single bigger topload(slighly bendy and branchy even at 430kHz). Im not going back using the bigger one, many times Ive tried swapping that in but no straight sparks.

here are the results

lower coupling


higher coupling. and I hit my OCD limit of around 120A peak.


I already know that with higher coupling will farther the lower and upper poles. will also raise the resonant freq.

going back to the shots, it is seen that with higher coupling result in faster primary current rise.

Im not at all so much concerned about tuning(maybe?) because If I wanted to get longer arks, I just had to add more  ramp length (and Vbus as well) BY lowering C=14nF and L around 11uH or so(basically making it Low inductance)..

but I believe Im still missing something else...

primary coil impedance?

detuning?

can someone pls explain this in a more laymans term way 


edit:
I just ordered yet another cheapER IGBT. 40T60(ANF)
cancelled it. im unsure if this IGBT has the version WITHOUT body diode(?) no wonder they where so cheap(?) Im also unsure if the ANF should be separated from the part# upper part(?)


and if ever I go for full bridge 8 IGBTs(double each switch). will a single GDT be enough to drive them all? I was thinking of using a ferrite E-I or EE core (no gap of coarse).

I also have this circuit built a few years ago. dick smith lopt/fbt tester. would this suffice in testing gdt for inductance check?


thanks

[davekni]

higher coupling. and I hit my OCD limit of around 120A peak.

going back to the shots, it is seen that with higher coupling result in faster primary current rise.

Have you zoomed in to see if you are still running at upper pole frequency?  Any QCW coil owners have insight on expected effect of coupling on primary current?  I could imagine a small current increase since upper-pole frequency increases and capacitances remain the same.  However, arc voltages tend to drop as frequency increases, so the net effect might be minimal.

I'm not clear on your remaining questions.  Tuning is important.  Larger difference between primary and secondary frequencies will result in higher primary current.  (More current is required to get power to secondary.)

and if ever I go for full bridge 8 IGBTs(double each switch). will a single GDT be enough to drive them all? I was thinking of using a ferrite E-I or EE core (no gap of coarse).

Yes, a single GDT is fine.  Construction technique (minimizing parasitic inductance in GDT and leads) is important, especially at your 430kHz frequency.  E-I or EE cores work fine for GDTs too, with no gap as you know.

[Rafft]

David

yes coil starts at upper pole around 475KHz then goes lower.
I have set back my primary to be 11uH/14nF(just like before w/o the 'upper pole initial freq') w/c has given me  longer spark output.

Im not sure why but when I go higher coupling, it likes more primary current( even at a lower Vbus) compared with the lower coupling. it should have been the opposite(higher coupling lower Pri current), or so ive read on other posts.

I was hoping to understand more about 'detuning' and if would help in any way with my coil setup

anyways ill UP my ocd to maybe something like 150Apk and see if nothing explodes.

regarding impedance, this is the topic im talking about. sorry if I havent posted link before. and Im talking the sweet spot of 10uH/15nF for MY COIL. if it makes sense.

https://highvoltageforum.net/index.php?topic=113.0

[davekni]

yes coil starts at upper pole around 475KHz then goes lower.

Have you checked how much lower the frequency goes?  In light of Uspring's latest post to the other thread:
 https://highvoltageforum.net/index.php?topic=113.msg15012#msg15012
I wonder if operation is jumping to lower pole as upper pole gets too damped for ZCS.  However, higher coupling should (I think) help maintain ZCS at upper pole for any given arc load.

Other than that, I'm out of ideas here.  Higher current at higher coupling is not what I'd expect.  I don't have enough relevant experience to guess at any other possible reasons for your results.

Have fun experimenting!  Hope you end up with performance you are satisfied with.

[davekni]

In case it is of interest, I purchased some "MBQ60T65PES" parts from two different EBay suppliers.  Both were counterfeit, but not exactly the same.  They were roughly half-capacitance.  Vce reached worst-case voltage at ~30A rather than specified 60A.  Even lower current for spec'ed typical Vce.  Got refunds from both suppliers, one immediately and the other more reluctantly.

[Rafft]

too bad they are fake too. glad that it was refunded

thats why I dont overdrive much my 60T65PES 

anyways I ordered(yet another) 10pcs because they where sooo cheap.
used my yellow LCR meter to check my un-used 60T65PES(for comparison) with these cheaper igbt. checked datasheet and gate capacitance seems close to the LCR reading. its only 80A@25degC so I will 'double switch' this for a full bridge and drive this to 150Apk. ran out of igbt heatpads, ordered them. will arrive 2 weeks tops.

will see if it holds up!

just curious, how did you test it to ~30A only? id imagine you also 'heat up' the igbt prior to testing