Ramped SSTC - Power Supply Question + General Improvements

[davekni]

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

[ZakW]

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

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

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

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

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

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

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

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

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

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

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

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

______________________

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

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

[davekni]

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Great!  Good design practice.

[ZakW]

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

Excellent! Thanks for confirming that for me.

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

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

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

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

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

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

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

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

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

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

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

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

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

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


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

[davekni]

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

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

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

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

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

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

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

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

[ZakW]

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

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

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

/>

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

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

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

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

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

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

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

 

[davekni]

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

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

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

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

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

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

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

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

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

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

[ZakW]

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

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

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

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

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

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

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

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

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

Thanks for elaborating. I was just curious.

______________

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

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

[davekni]

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

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

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

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

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

[ZakW]

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

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

[ZakW]

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

[davekni]

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

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

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

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

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

[ZakW]

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

Here are some scope shots to illustrate what is happening.

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

Here is the interrupter tuned to the zero crossing.



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



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

Any thoughts or ideas would be appreciated.

[davekni]

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

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

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

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

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

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

Thank you for the clearly defined scope traces.

[ZakW]

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

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

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

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

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

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

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

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

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

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


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

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

Thank you for the clearly defined scope traces.

Happy to provide them, I appreciate the help!

[davekni]

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

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

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

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

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

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

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

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

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

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

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

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

[ZakW]

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

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

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

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

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

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

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

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

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

[davekni]

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

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

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

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

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

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

[ZakW]

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

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

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

From Page 1, my first post.


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

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

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

[davekni]

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

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

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

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