RDRSSTC - Project Build

[ZakW]

Fixed the output, the driver appears to be switching correctly now. Got a crash course on the logic used and followed it to the driver IC. I used an non invert/invert instead of a dual inverting driver IC. Must have mixed up that parts along the way and installed the wrong IC.

Looking at the bridge now!

Update:using my signal generator I was able to get a small breakout at 400kHz using my variac. The coil seems to be running CW and is not being interrupted. Still looking into it.

That "also circled in red" inverter appears to be a limit on maximum on-time and duty cycle.  Would cause trouble for a coil running 5ms on-time.

Can you explain a bit more about this? I don't see the equivalent in the 1.3 driver.

[davekni]

Can you explain a bit more about this? I don't see the equivalent in the 1.3 driver.

The R/C network on the inverter input is a delay that monitors enable signal.  If on too long, the inverter input goes high, output low, which removes enable.  After a longer off-time delay, inverter output goes back high, allowing enable to pass through the AND gate again.  This helps prevent coil burn-out if enable is accidentally stuck on or set to too-high a duty cycle.  That feature appears to have been added after UD1.3.

[ZakW]

The R/C network on the inverter input is a delay that monitors enable signal.  If on too long, the inverter input goes high, output low, which removes enable.  After a longer off-time delay, inverter output goes back high, allowing enable to pass through the AND gate again.  This helps prevent coil burn-out if enable is accidentally stuck on or set to too-high a duty cycle.  That feature appears to have been added after UD1.3.

Thanks for the clarification.

So after scouring post after post it sounded like the typical fiber connector inverts the interrupter signal. I cut the trace from pin 2 of the 74hc14 and wired it directly to pin 3 of the 74hc74 and it is working!!! The driver only runs during the interrupter pulse, current consumption is way lower (of course). Now I just cant get it to work off of CT feedback.

I can get it to work using my signal generator but that is it. Using a 400kHz signal I am getting about 3in arcs. I tried flipping the phase but scoping the bridge and driver I am not seeing the coil trying to oscillate.

First core (white wire) has 1 turn through both blue and green wire cores. 32 turns on all cores.



Update: Right now I am just trying to get the primary to oscillate so I removed the secondary. I have a chunk of metal inside the primary to absorb some energy.

Primary: 7 turns or 18awg wire, 6.8nF cap. JavaTC says it should oscillate around 877kHz. That is about twice as high as I want so I will try increasing the capacitance to 20nF, that should get me to 511kHz. Still though, I should be seeing the primary oscillate even if it is around 880kHz, right?

I also tested at what voltage is too low for using the signal generator as feedback. Anything under 2v and the output of the driver becomes unstable, cutting out completely at around 1.3v. I rewound a new CT using 2 N87 cores with 22T on each for ~500:1. Flipped phasing around, still cannot get it to oscillate on its own.


Update #2: Referencing Dave's suggestion here - https://highvoltageforum.net/index.php?topic=1373.msg10197#msg10197

My favorite way to fix startup issues is to make the driver self-oscillating near the coil's resonant frequency.  That can be done to UD2.7 by adding a ~50k resistor from positive comparitor output (IC8 pin 7) to its inverting input (IC8 pin 3), removing R7, and adjusting the value of C33 to get a reasonable frequency.  This works only when jumper SV1 is inserted.  Adding the 50k resistor may be easier soldering by using the open pad for R7 as a connection to IC8-7 and and one end of D1, D2, or R2 for a connection to IC8-3.

50k did not work for me though. I added a 100k pot and tried different values. I still need to measure the resistance but at a certain point I can get the driver to oscillate and subsequently the primary does as well. This is with CT feedback. I added the secondary in there to see if I could get any output. The variac was thumping hard, I tried removing the ferrite rod from my DIY inductor and the thumping was less loud, however when I removed the jumper (no phase lead) the coil came alive! I referenced Loneoceans suggested inductor size but I likely need to fine tune it.

So it seems to be working with self-oscillation as well as no phase lead.

Update #3: Correction, it is working intermittently with phase lead. The self oscillation seems unstable and very finicky. I also switched to a larger secondary I had on hand. I was using one of my mini 2in coils but the larger coil is easier to test on. Managed to get ~10in arcs at around 75v but they were not very straight.

New coil Fres (unloaded) is 500kHz.

JavaTC says 5 turns @ 20nF should put me around 457kHz with 0.324 coupling.
 

[davekni]

I cut the trace from pin 2 of the 74hc14 and wired it directly to pin 3 of the 74hc74 and it is working!!!

I'm not quite following this mod, but if you are bypassing an HC14 stage, you may be missing out on HC14 hysteresis.  Slow rise and/or fall times from optical receiver may trigger subsequent logic at different voltages and therefore different times if not cleaned up by HC14 hysteresis.  Generally best to add an extra stage of HC14 inversion rather than removing one.  Though I'm not certain exactly whether your mod has this issue or not.  Obviously working for now.  Just a risk of possible issues in the future.

50k did not work for me though. I added a 100k pot and tried different values. I still need to measure the resistance but at a certain point I can get the driver to oscillate and subsequently the primary does as well. This is with CT feedback. I added the secondary in there to see if I could get any output. The variac was thumping hard, I tried removing the ferrite rod from my DIY inductor and the thumping was less loud, however when I removed the jumper (no phase lead) the coil came alive! I referenced Loneoceans suggested inductor size but I likely need to fine tune it.

You may need to decrease size of C6 to get your relatively high frequency with a reasonable self-oscillation feedback resistor value.
BTW, QCW and ramped coils are particularly prone to startup problems, since bus voltage is low when oscillation needs to start.  QCW coils often use FPGAs or other logic to force some number of startup cycles, or use a PLL to have oscillation running prior to feedback being sufficient.  My self-oscillation modification is an alternative, one I use in my QCW coil.

[ZakW]

I'm not quite following this mod, but if you are bypassing an HC14 stage, you may be missing out on HC14 hysteresis.  Slow rise and/or fall times from optical receiver may trigger subsequent logic at different voltages and therefore different times if not cleaned up by HC14 hysteresis.  Generally best to add an extra stage of HC14 inversion rather than removing one.  Though I'm not certain exactly whether your mod has this issue or not.  Obviously working for now.  Just a risk of possible issues in the future.

I skipped IC1F pins 5 & 6 and wired my interrupter directly to pin 1 of IC2A (74HC08). My signal was being inverted one too many times, that seemed to have fixed the issue. Might not be the best solution but at least it got it working for now.

You may need to decrease size of C6 to get your relatively high frequency with a reasonable self-oscillation feedback resistor value.
BTW, QCW and ramped coils are particularly prone to startup problems, since bus voltage is low when oscillation needs to start.  QCW coils often use FPGAs or other logic to force some number of startup cycles, or use a PLL to have oscillation running prior to feedback being sufficient.  My self-oscillation modification is an alternative, one I use in my QCW coil.

I tried adjusting this but I cant seem to get consistent results. Tuning this does impact the look and sound of the arcs, sort of like my previous coil. Once adjusted correctly the popping and snapping is minimal.

Speaking of which, I cant seem to get good sword arcs with a topload attached. I took a short video and some scope captures. My secondary Fres is 500kHz unloaded so my primary is still out of tune at around 550kHz. I know the topload is lowering the Fres even more but is that why the arc sounds snappy and loud? In the pictures below you can see the blue trace (bridge output) how the sword arcs make a really uniform smooth output but when the arcs are snapping it is more trumpet shape.

Blue = Bridge output
Yellow = TL3116 output (pin 7)
purple = CT feedback (SV1 jumper)

No topload



Blue = Vge


Notice how this waveform is uniform compared to when a topload is used.


Blue= bridge output


Topload on






Do the 'no topload' waveforms look alright? I haven't tried tuning phase lead yet.

[davekni]

I skipped IC1F pins 5 & 6 and wired my interrupter directly to pin 1 of IC2A (74HC08). My signal was being inverted one too many times, that seemed to have fixed the issue. Might not be the best solution but at least it got it working for now.

Yes, not the best solution.  Missing hysteresis of HC14.  Would be better to patch in the spare HC14 instead.  Might stay working fine as is.  I haven't analyzed in detail what issues could be caused by this specific circuit missing hysteresis.

purple = CT feedback (SV1 jumper)

Triangle-wave shape looks like running in SSTC mode rather than DRSSTC.  Would expect closer to a sine wave.  Even SSTCs with reasonable coupling factor look more sine wave like than your captures, which are all triangle-wave shaped.  Is MMC way too large or perhaps failed shorted?

Speaking of which, I cant seem to get good sword arcs with a topload attached. I took a short video and some scope captures.

I think the issue is too short a breakout on top of the top load.  Top load shields breakout tip, reducing electric field at tip.  Takes higher secondary voltage before breakout starts.  Breakout is already well up line quarter-cycle ramp.  Once breakout occurs, voltage is high enough to make arc grow too fast for sword sparks, fast like normal DRSSTC arcs.

My secondary Fres is 500kHz unloaded so my primary is still out of tune at around 550kHz. I know the topload is lowering the Fres even more but is that why the arc sounds snappy and loud?

Even farther out of tune may contribute to arc behavior.  However, QCW coils (and I presume RDRSSTC coils too) work best operating at upper pole.  Two ways to get upper-pole operation.  One is to keep primary frequency a bit above secondary (unlike normal DRSSTC tuning).  Other is to have pre-startup oscillation at a higher frequency (upper pole frequency).  Latter approach works only if coupling is high enough and secondary frequency is not too far above primary frequency.

[ZakW]

Yes, not the best solution.  Missing hysteresis of HC14.  Would be better to patch in the spare HC14 instead.  Might stay working fine as is.  I haven't analyzed in detail what issues could be caused by this specific circuit missing hysteresis.

I will make an update to my schematic and PCB file to include an extra stage.

Triangle-wave shape looks like running in SSTC mode rather than DRSSTC.  Would expect closer to a sine wave.  Even SSTCs with reasonable coupling factor look more sine wave like than your captures, which are all triangle-wave shaped.  Is MMC way too large or perhaps failed shorted?

I am having issues getting the coil to run at different frequencies. It seems to always oscillate much higher than what javaTC predicts, I know it is just an estimation but I cannot seem to change the frequency much regardless of MMC value or primary. It tends to always be around 500kHz even though I aim for 400kHz or a little lower.

I did notice that after increasing coupling, changing the MMC amount and turn count the feedback signal (purple) was a lot more sinusoidal.

I think the issue is too short a breakout on top of the top load.  Top load shields breakout tip, reducing electric field at tip.  Takes higher secondary voltage before breakout starts.  Breakout is already well up line quarter-cycle ramp.  Once breakout occurs, voltage is high enough to make arc grow too fast for sword sparks, fast like normal DRSSTC arcs.

You were right, I increased the length and the snapping went away! 

Even farther out of tune may contribute to arc behavior.  However, QCW coils (and I presume RDRSSTC coils too) work best operating at upper pole.  Two ways to get upper-pole operation.  One is to keep primary frequency a bit above secondary (unlike normal DRSSTC tuning).  Other is to have pre-startup oscillation at a higher frequency (upper pole frequency).  Latter approach works only if coupling is high enough and secondary frequency is not too far above primary frequency.

Sounds good, I have been reading as much as I can about tuning.

Question regarding checking secondary and primary tuning with a signal generator. I found this video https://www.youtube.com/watch?v=GLyW1zRZymk and they show this diagram for how everything is all connected:



Is that correct? I always see the MMC in series with the primary unlike the diagram. I followed the diagram and was able to find the resonant frequency of my current MMC & primary as well as the secondary with a 20in wire to simulate loading. Secondary with topload and wire was 360kHz and the primary was 310kHz  but again when I ran the coil it was quite a bit higher than what was measured. I am at a loss for this.

Edit: I see your post here https://highvoltageforum.net/index.php?topic=783.msg5109#msg5109

What values do you have for L1, C32, and C33?  Also, what IGBTs are being used?

I am using the standard values per the schematic, 150nF 150pF for C32 and 220pF for C33. What impact does adjusting C32 have?

2) C33 should be a bit lower for your relatively-high frequency operation of 280kHz.  I'd suggest 470pF, or 680pF at most.

Per Loneoceans notes, I think 220pF for C33 is correct. Ideally I want to operate somewhere in the range of 350-450kHz, depending on the secondary used. I am using a place holder until I get everything running correctly.

Update: Seems like it is the phase adjustment portion of the driver that is causing the issues, copied from the 2.1b and added to 1.3b. When the SV1 jumper is in place the primary fres is around 555kHz, hardly changes with added capacitance. If I remove the jumper fres drops to ~320kHz and responds when MMC is adjusted.

[davekni]

I did notice that after increasing coupling, changing the MMC amount and turn count the feedback signal (purple) was a lot more sinusoidal.

Good, sounds like progress.

Is that correct? I always see the MMC in series with the primary unlike the diagram. I followed the diagram and was able to find the resonant frequency of my current MMC & primary as well as the secondary with a 20in wire to simulate loading. Secondary with topload and wire was 360kHz and the primary was 310kHz  but again when I ran the coil it was quite a bit higher than what was measured. I am at a loss for this.

That video is measuring frequency of a parallel-resonant circuit, seeing where impedance peaks (maximum impedance frequency).  If measuring series resonant, drive impedance will be minimum at resonance.  Either works for primary.  However, was secondary removed for primary measurement?  Presence of secondary changes primary frequency due to coupling.  For secondary, series resonant mode is the only reasonable way to check.  Any connection to top will change capacitance.  For measuring secondary, it can be in place, but primary coil must be open-circuit to not affect secondary.
Any pair of coupled resonant circuits has two resonant frequencies, neither of which are same as individual frequencies.  Upper pole and lower pole frequencies.  Most QCW and ramped DRSSTC coils run at upper pole frequency.  You can measure those two frequencies by measuring with both coils in place, secondary bottom grounded, and MMC connected.

I am using the standard values per the schematic, 150nF for C32 and 220pF for C33. What impact does adjusting C32 have?

C32 is a small cap just to filter out any high-frequency noise (such as coupled from bridge output switching events) that might cause unintended comparitor switching.  Presuming we are looking at the same schematic, C32 is the capacitor across CT secondary.  Value is usually 150pF.  If you are using 150nF that could explain problems!  C33 value of 220pF should be good for your frequencies.

When the SV1 jumper is in place the primary fres is around 555kHz, hardly changes with added capacitance. If I remove the jumper fres drops to ~320kHz and responds when MMC is adjusted.

Operating with SV1 removed is likely to damage circuitry with excess CT output voltage.  I'd check R2 to see if it is still 1k.  Although if you really have 150nF for C32, that would avoid the damage, providing sufficient load to CT output.

[ZakW]

Good, sounds like progress.

Maybe a little, but tuning it today was frustrating.

That video is measuring frequency of a parallel-resonant circuit, seeing where impedance peaks (maximum impedance frequency).  If measuring series resonant, drive impedance will be minimum at resonance.  Either works for primary.

Thanks for the clarification. If I understand you correctly, I can have the MMC series or parallel? Are there any benefits or drawback to either configuration? When I google DRSSTC schematics I only see MMC in series with the primary.

However, was secondary removed for primary measurement?  Presence of secondary changes primary frequency due to coupling.  For secondary, series resonant mode is the only reasonable way to check.  Any connection to top will change capacitance.  For measuring secondary, it can be in place, but primary coil must be open-circuit to not affect secondary.

I took several measurements today and noted all of the values. I also measured under different configurations: with and without primary in place, different toploads, finally a 20in wire to simulate arc.

For the secondary testing I just connect my signal gen to the base and hang a probe near by. Tested various size toploads and noted the Fres.

For the primary I set it up like the schematic suggested so I guess I was measuring parallel-resonance instead. I took measurements with and without the primary around the secondary. Another thing I noticed was the impact of grounding the secondary. I wasn't sure which was correct so I noted those values as well. Grounding the secondary made the Fres quite a bit lower.

primary coil must be open-circuit

As in not being connected to the MMC?

Any pair of coupled resonant circuits has two resonant frequencies, neither of which are same as individual frequencies.  Upper pole and lower pole frequencies.  Most QCW and ramped DRSSTC coils run at upper pole frequency.  You can measure those two frequencies by measuring with both coils in place, secondary bottom grounded, and MMC connected.

So if I know the Fres of the secondary is say 500kHz but when the primary is in place and the MMC is connected (secondary is grounded) I can remeasure and determine the upper pole and lower pole? Will the Fres peak higher than 500kHz for upper and the same for a Fres below 500kHz for lower? I was reading Loneoceans DRSSTC and QCW pages but he never explained how he took these measurements.

C32 is a small cap just to filter out any high-frequency noise (such as coupled from bridge output switching events) that might cause unintended comparitor switching.  Presuming we are looking at the same schematic, C32 is the capacitor across CT secondary.  Value is usually 150pF.  If you are using 150nF that could explain problems!  C33 value of 220pF should be good for your frequencies.

That was a typo  I updated my post to correct it. I did install a 150pF cap like the schematic suggests.

Operating with SV1 removed is likely to damage circuitry with excess CT output voltage.  I'd check R2 to see if it is still 1k.  Although if you really have 150nF for C32, that would avoid the damage, providing sufficient load to CT output.

I did notice without SV1 that feedback voltage was high, I think around 6.4V at times. I only removed it out of frustration. I will check R2 to verify its value and run it with the jumper installed going forward.

Final note on testing: I am using a multi turn primary with 8 turns. I have several caps soldered in place to be able to easily add and remove them. I use two 10nF caps as well as several small 6.8nF for smaller increments. I measured the primary Fres which came out to 310kHz but when the coil was running , bridge output and feedback both showed it running over 500kHz... what could cause this? The secondary Fres without a topload was 530kHz, with a topload and 20in wire, it dropped to 360kHz. So the primary seems to be running far off from resonance.

Sometimes I get what looks like a good primary tap & MMC but then OCD trips at or near120v. So I make a slight adjustment but then OCD trips at a much lower voltage. If it doesn't trip then the arcs are loud, snappy and branched.   

How might I assume a safe primary current for my IGBTs? I know Loneoceans has a section on how OCD CT voltage correlates to current and how to adjust that voltage via R10 but I am wondering how you can assume this part can run a 250A or 500A? I can double check the datasheet if the answer is there but I assumed since the IGBTs are being driven harder we assume they can handle higher peak currents.

[davekni]

Thanks for the clarification. If I understand you correctly, I can have the MMC series or parallel? Are there any benefits or drawback to either configuration? When I google DRSSTC schematics I only see MMC in series with the primary.

Any given inductor and capacitor together can form a resonant circuit.  Frequency is the same whether driven in series or parallel.  Most DRSSTC primaries are driven in parallel.  (Only exception I'm aware of is ZVS driven coils.)  Measuring primary frequency can be done in parallel or series configuration.  Doesn't matter.  Frequency is the same.

As in not being connected to the MMC?

OK to have one primary lead connected to MMC as long as other lead is open or other side of MMC is open.  Just avoid any closed circuit on primary.  Break that closed circuit at any one point, where ever is convenient.

So if I know the Fres of the secondary is say 500kHz but when the primary is in place and the MMC is connected (secondary is grounded) I can remeasure and determine the upper pole and lower pole? Will the Fres peak higher than 500kHz for upper and the same for a Fres below 500kHz for lower? I was reading Loneoceans DRSSTC and QCW pages but he never explained how he took these measurements.

Yes for the two pole frequencies.  Probably easiest to measure the assembled system through the primary.  Have secondary in place, bottom grounded and top to top load (and arc simulation wire if you want loaded frequencies).  Then measure frequencies the same way you measured the single isolated primary frequency.

I did notice without SV1 that feedback voltage was high, I think around 6.4V at times. I only removed it out of frustration. I will check R2 to verify its value and run it with the jumper installed going forward.

SV1 can be installed in either position (with or without phase lead), just not left open.  UD2.7 CT (SV1) voltages are intended to be high, often peak at 50 to 100V.  Without SV1 installed, voltage can go even higher, so fry resistors.  However, in your case, perhaps CT secondary current is too low for reliable US2.7 feedback.  Perhaps CT ratio is too high for a relatively low primary current coil.

Final note on testing: I am using a multi turn primary with 8 turns. I have several caps soldered in place to be able to easily add and remove them. I use two 10nF caps as well as several small 6.8nF for smaller increments. I measured the primary Fres which came out to 310kHz but when the coil was running , bridge output and feedback both showed it running over 500kHz... what could cause this? The secondary Fres without a topload was 530kHz, with a topload and 20in wire, it dropped to 360kHz. So the primary seems to be running far off from resonance.

Once you measure upper pole frequency, see if that is close.  For initial experiments I'd include top load but no arc-simulation wire.  If operating frequency doesn't match upper pole, then perhaps self-oscillation component values are such that it oscillates at too high a frequency.  Might help to post your schematic as it is now including any self-oscillation modifications.

Sometimes I get what looks like a good primary tap & MMC but then OCD trips at or near120v. So I make a slight adjustment but then OCD trips at a much lower voltage. If it doesn't trip then the arcs are loud, snappy and branched.   

Scope current during these conditions to see what actual values are, along with frequency and shape.  Scoping across 51 ohm CT burden resistor is a good option (doesn't require another CT just for scoping).  If SV1 is set for phase lead, scope across just 51 ohms, not including phase-lead inductor.

How might I assume a safe primary current for my IGBTs? I know Loneoceans has a section on how OCD CT voltage correlates to current and how to adjust that voltage via R10 but I am wondering how you can assume this part can run a 250A or 500A?

Ramped coils have long on-times (quarter or half line cycle).  Rated peak currents are generally for 1ms, 225A I think for your part.  I wouldn't go above 250A for a ramped coil.

[ZakW]

Yes for the two pole frequencies.  Probably easiest to measure the assembled system through the primary.  Have secondary in place, bottom grounded and top to top load (and arc simulation wire if you want loaded frequencies).  Then measure frequencies the same way you measured the single isolated primary frequency.

I'll give that a shot.

SV1 can be installed in either position (with or without phase lead), just not left open.  UD2.7 CT (SV1) voltages are intended to be high, often peak at 50 to 100V.  Without SV1 installed, voltage can go even higher, so fry resistors.  However, in your case, perhaps CT secondary current is too low for reliable US2.7 feedback.  Perhaps CT ratio is too high for a relatively low primary current coil.

Being completely unfamiliar with the UD, I only installed a 2 pin header instead of a 3 pin. I see now why it is a 3 pin header. My only option is to run SV1 closed unless I remove the inductor and short the connection.

I still need to check the 1K resistor (R6) to make sure it is not fried.

For the inductor I took apart a misc part that I had on hand and added a couple layers and measured with LC meter until I was around "9 - 15uH; works well with TO247 IGBTs" - loneoceans. I can move the slug up or down to vary the inductance. I think that should be fine.

Scope current during these conditions to see what actual values are, along with frequency and shape.  Scoping across 51 ohm CT burden resistor is a good option (doesn't require another CT just for scoping).  If SV1 is set for phase lead, scope across just 51 ohms, not including phase-lead inductor.

ground lead would be connected to TP1, but are you saying to then scope at TP2 or 3?



Updated schematic using logic symbols instead of IC's. Added notes about things I need to update but this schematic is accurate for the modifications I have made so far.

However, in your case, perhaps CT secondary current is too low for reliable US2.7 feedback.  Perhaps CT ratio is too high for a relatively low primary current coil.

Ramped coils have long on-times (quarter or half line cycle).  Rated peak currents are generally for 1ms, 225A I think for your part.  I wouldn't go above 250A for a ramped coil.

My CT is 33:1:25, so 825:1

I'll have to check loneoceans notes on calculating the output voltage given my turns ratio in order to figure out how to set the OCD to the correct value. So far nothing as died so I am grateful for that. Wish my last SSTC had OCD, would have save so much $$$!

Thanks for the all the information!

[davekni]

ground lead would be connected to TP1, but are you saying to then scope at TP2 or 3?

Yes, ground to TP1 and probe to TP2.  That is scoping directly across 51 ohm resistor.  Represents current that way.  (TP3 includes phase lead, so would be leading actual current phase and a bit higher amplitude too.)

Updated schematic using logic symbols instead of IC's. Added notes about things I need to update but this schematic is accurate for the modifications I have made so far.

Thank you.  Makes discussion much clearer.  BTW, self oscillation may work better if R6 is increased from 1k to ~5k as mentioned near the end of this post:
    https://highvoltageforum.net/index.php?topic=1914.msg14854#msg14854
Above post recommends changing R2, the label from common UD2.7 schematics.  This is R6 for your schematic.  Increasing R6 value should help make self-oscillation frequency more stable and easier to adjust.

Without self-oscillation, using 4148 diodes increases CT feedback voltage necessary to start oscillation, relative to UD2.7 schematic that uses schottky diodes.  (Some UD2.7 versions use 1N4148 instead.)  However, 1N4148 or equivalent is good for a self-oscillating version as you now have.

My CT is 33:1:25, so 825:1

Thus for feedback CT input, +-250A primary becomes 250/825=0.303A, which is +-15.45V across 51 ohms.  Should be fine.

I'll have to check loneoceans notes on calculating the output voltage given my turns ratio in order to figure out how to set the OCD to the correct value.

Hope I'm not encouraging you to become too lazy by calculating for you.  Fairly simple.  0.303V becomes 3.03V across 10 ohm OCD burden resistor.  Because shutdown takes a cycle (where current can continue to increase slightly), set trip point slightly under 3.03V.

For the inductor I took apart a misc part that I had on hand and added a couple layers and measured with LC meter until I was around "9 - 15uH; works well with TO247 IGBTs" - loneoceans. I can move the slug up or down to vary the inductance. I think that should be fine.

For a crude approximation, 9uH/51ohms=176ns.  15uH/51ohms=294ns.  This approximation only works when frequency is low enough where 294ns is a small fraction of 90 degrees (of 1/4 cycle).  In your high frequency use, time lead will be less than this simple calculation.  So you may need more inductance depending on both IGBT speed and delay through driver.  Driver delay can be reduced a little by using AC08 instead of HC08.  But AC08 requires good ground plane and supply bypassing due to fast switching.  Use only if ECB layout is good and AC08 is not in a socket.

Thanks for the all the information!

You're welcome.  Glad it's been useful.

[ZakW]

Thank you.  Makes discussion much clearer.  BTW, self oscillation may work better if R6 is increased from 1k to ~5k as mentioned near the end of this post:

I missed that recommendation. I will give that a try.

Without self-oscillation, using 4148 diodes increases CT feedback voltage necessary to start oscillation

Not quite sure I follow. Regardless of the self-oscillation mod are you suggesting to swap out D1 & D2 with 4148 type diodes?

Hope I'm not encouraging you to become too lazy by calculating for you.  Fairly simple.  0.303V becomes 3.03V across 10 ohm OCD burden resistor.  Because shutdown takes a cycle (where current can continue to increase slightly), set trip point slightly under 3.03V.

Not at all. On the contrary, I was trying not to be a burden by asking you for step by step instructions but I appreciate you doing it. I am adding as much as I can to my notes so I can reference these calculations in the future. Growing up I always copied schematics and never invested time into learning theory. So now I am trying to increase my understanding via projects but it can be tough.

For a crude approximation, 9uH/51ohms=176ns.  15uH/51ohms=294ns.  This approximation only works when frequency is low enough where 294ns is a small fraction of 90 degrees (of 1/4 cycle).  In your high frequency use, time lead will be less than this simple calculation.  So you may need more inductance depending on both IGBT speed and delay through driver.  Driver delay can be reduced a little by using AC08 instead of HC08.  But AC08 requires good ground plane and supply bypassing due to fast switching.  Use only if ECB layout is good and AC08 is not in a socket.

That makes sense, I guess I can make that determination after I scope across the CT resistor. I am comparing the CT feedback to the primary output, right? A lot of DRSSTC scope shots I see dont always label the traces.

[davekni]

Not quite sure I follow. If I wasn't using the self-oscillation mod are you suggesting to swap out D1 & D2 with 4148 type diodes or add diodes like the OCD CT input?

The schematic you show in above post shows MBR0530, a schottky diode as used in most UD2.7 schematics.  This is best for standard UD2.7 without self-oscillating mods.  This post explains normal (not self-oscillating) startup a bit, though for an SSTC circuit with HC14 feedback:
    https://highvoltageforum.net/index.php?topic=840.msg5667#msg5667
UD2.7 input startup is similar.  CT feedback voltage required from first half-cycle is the forward voltage drop of D1 with ~14uA of current flowing through it (quiescent current through 100k R27).  Schottky diodes have much lower Vf, especially at low current, making startup easier.  (D1 is from above post, which is D6 in your schematic.  R27 is R3 in your schematic.)
However, for self-oscillation, schottky diodes clamp the self-oscillation feedback voltage to a low value, reducing frequency stability.  1N4148 diodes allow more voltage there for cleaner self-oscillation.  Because voltage is already oscillating at close to operating frequency, little additional CT feedback voltage is needed to lock to coil frequency.  The higher Vf doesn't increase required startup CT feedback voltage in self-oscillating configuration.

Not at all. On the contrary, I was trying not to be a burden by asking you for step by step instructions but I appreciate you doing it. I am adding as much as I can to my notes so I can reference these calculations in the future. Growing up I always copied schematics and never invested time into learning theory. So now I am trying to increase my understanding via projects but it can be tough.

An excellent way to increase understanding is with analog simulation.  Many free options available.  My favorite is LTSpice.  Analyzing self oscillation of just the comparitor circuit would be a good place to start.  Doesn't require simulating entire coil.

I am comparing the CT feedback to the primary output, right?

Yes,  CT feedback current across 51 ohm resistor only (TP2 to TP1 from previous post).  Or some people use a third CT with separate burden resistor for scoping.

A lot of DRSSTC scope shots I see dont always label the traces.

Yes, that is a common problem that I sometimes complain about.  Hard to respond appropriately to unlabeled scope traces.

[ZakW]

The schematic you show in above post shows MBR0530, a schottky diode as used in most UD2.7 schematics.  This is best for standard UD2.7 without self-oscillating mods.  This post explains normal (not self-oscillating) startup a bit, though for an SSTC circuit with HC14 feedback:
    https://highvoltageforum.net/index.php?topic=840.msg5667#msg5667
UD2.7 input startup is similar.  CT feedback voltage required from first half-cycle is the forward voltage drop of D1 with ~14uA of current flowing through it (quiescent current through 100k R27).  Schottky diodes have much lower Vf, especially at low current, making startup easier.  (D1 is from above post, which is D6 in your schematic.  R27 is R3 in your schematic.)
However, for self-oscillation, schottky diodes clamp the self-oscillation feedback voltage to a low value, reducing frequency stability.  1N4148 diodes allow more voltage there for cleaner self-oscillation.  Because voltage is already oscillating at close to operating frequency, little additional CT feedback voltage is needed to lock to coil frequency.  The higher Vf doesn't increase required startup CT feedback voltage in self-oscillating configuration.

Thanks for that, I will take a look at that post. I copied the 2.1b schematic which uses MCL4148 diodes instead of the MBR0530. I have some regular 1N4148W diodes I can try in their place.

An excellent way to increase understanding is with analog simulation.  Many free options available.  My favorite is LTSpice.  Analyzing self oscillation of just the comparitor circuit would be a good place to start.  Doesn't require simulating entire coil.

I have been meaning to get more into simulation. It is on my list to start practicing.

[davekni]

Thanks for that, I will take a look at that post. I copied the 2.1b schematic which uses MCL4148 diodes instead of the MBR0530. I have some regular 1N4148W diodes I can try in their place.

I expect any diode with "4148" in part number has similar characteristics to the original 1N4148.  No reason to change parts.

[ZakW]

I decreased R6 (normally a 1K (R2 in 2,1b)) to 560ohms and now I am getting a much better closer to operating frequency output range. Before I swapped out the 1K resistor with a 5K and the TL3116 was oscillating at most several mHz anything lower than 1mHz was not stable. I am using a 50k POT for adjustment.

I will take the other measurements and share my findings, probably tomorrow.

[davekni]

I decreased R6 (normally a 1K (R2 in 2,1b)) to 560ohms and now I am getting a much better closer to operating frequency output range. Before I swapped out the 1K resistor with a 5K and the TL3116 was oscillating at most several mHz anything lower than 1mHz was not stable. I am using a 50k POT for adjustment.

Just looked back at your updated schematic from a few posts ago.  C15 will need to be larger than 220pF, ~1nF, to go along with 5K for R6.  I'd recommend changing C15 rather than going so low for R6 value.

[ZakW]

Here are scope captures showing phase lag/lead.

Purple = probe across 51 ohm resistor
Blue = bridge output
Yellow = TL3116 output (I don't think it applies, just had the input on during captures)

This is with the ferrite slug fully seated





Switching looks good to me, not a lot of spikes.


This is with the slug removed





Switching looks worse, is that too much lead?

Edit: Looking at the first picture I am trying to calculate the primary current.

I_secondary=V_burden/R_burden

4.56V/51ohms = 0.089A

I_primary = I_secondary * Turns ratio

0.089A * 825 turns = 73.43A flowing through the primary?

[davekni]

This is with the ferrite slug fully seated

Switching looks good to me, not a lot of spikes.

Looks reasonable, but would be better with slightly more phase lead.  You'll need a bit higher inductance to get enough.

Switching looks worse, is that too much lead?

No, this is far too little phase lead.  Slug removed is minimum inductance so minimum phase lead.

0.089A * 825 turns = 73.43A flowing through the primary?

Yes.

BTW, for a bit of detail, look at the little bumps in bridge out just prior to switching in your first two slug-in scope captures.  In the second of those first two (at 50ns/div), the bump lasts roughly from division 3 to 4, with the main switching starting at division 4.6.  The beginning of that bump at division 3 is when one set of IGBTs turns off.  The end of the bump at division 4 is when current reverses, pulling bridge output back to it's previous value.  Then opposite set of IGBTs turns on starting at division 4.6.  (Current sense trace appears to be ~70ns delayed from reality.  Not sure exactly why, but seems somewhat common.  Perhaps due to parasitics of CT.)  With ideal phase lead, that bump becomes the real output transition and opposite IGBTs turn on just as current is passing through zero, around 100ns more phase lead than you have now.