High Voltage Forum

Tesla coils => Dual Resonant Solid State Tesla coils => Topic started by: bozidar on August 05, 2019, 10:04:50 PM

Title: Problems with my first DRSSTC
Post by: bozidar on August 05, 2019, 10:04:50 PM
Hello.
I'm making DRSSTC and I'm having some problems, so I hope one of you can help me.
First problem is Vge overshoot.In attached images you can see gate-emitter voltage (yellow signal), primary current (blue signal) and Vce (green signal).


This is core that I use for GDT and CT https://www.aliexpress.com/item/32857367831.html?spm=2114.search0104.3.15.968a6e5f9nOtT4&ws_ab_test=searchweb0_0%2Csearchweb201602_9_10065_10068_319_10059_10884_317_10887_10696_321_322_10084_453_10083_454_10103_10618_10307_537_536%2Csearchweb201603_53%2CppcSwitch_0&algo_expid=1e5a0c1a-4811-4a8e-926b-6c1f0134c046-2&algo_pvid=1e5a0c1a-4811-4a8e-926b-6c1f0134c046 (https://www.aliexpress.com/item/32857367831.html?spm=2114.search0104.3.15.968a6e5f9nOtT4&ws_ab_test=searchweb0_0%2Csearchweb201602_9_10065_10068_319_10059_10884_317_10887_10696_321_322_10084_453_10083_454_10103_10618_10307_537_536%2Csearchweb201603_53%2CppcSwitch_0&algo_expid=1e5a0c1a-4811-4a8e-926b-6c1f0134c046-2&algo_pvid=1e5a0c1a-4811-4a8e-926b-6c1f0134c046) .
I'm using calculations for GDT from this site: http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-i/ (http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-i/).
For my GDT, I put only 4 turns of 5 twisted wires .I use 5 twisted wires  that are wrapped with aluminum foil and all this is put into the shrink tube (handmade) to minimise leakage inductance.
That is explaned here http://thedatastream.4hv.org/gdt_practical.htm (http://thedatastream.4hv.org/gdt_practical.htm)
I tried to reduce voltage oscillations on gate-emitter by lowering leakage indutcance of GDT , but without much succes.
Then I bought this core https://export.farnell.com/epcos/b64290l0048x830/ferrite-core-cylindrical-34mm/dp/2355017?ost=B64290L0048X830&ddkey=https%3Aen-EX%2FElement14_Export%2Fsearch (https://export.farnell.com/epcos/b64290l0048x830/ferrite-core-cylindrical-34mm/dp/2355017?ost=B64290L0048X830&ddkey=https%3Aen-EX%2FElement14_Export%2Fsearch) to try to get better results, but the results were the same.
I use 5.1 Ohm resistors with reverse diode and P6KE30CA diodes across the gates.

I put on an 11 Ohm resistor today and I got a lot better waveforms , but I don`t know is this resistor too big because rise and fall times are little slower(around 300nS) .



My second problem is that all the time i have had problems to make the coil  work properly. When I set the freq to 3kHz and 200us on time the coil did not want to start to work, and if i decrease freq or on time everything returns to normal. That happens on low voltage on Vbus ( 18 V ) and when I put 310V on Vbus the coil  will not work at all (of course ,with low on time, 10us ).When I connect a small cap (100pF) between the ground and one GDT of the primary output connector, the coil works again as it should ( 18V -Vbus ). I really don't know how to fix this and what can cause this?
I check all connections,igbts,caps etc and everythig is allright. That happens with low voltage (18VDC) on Vbus and also with full voltage ( 310 VDC ).(I just decrease supply voltage between 15V and 18V to keep Vge peak under 30 v).
I have already connected the coil to 220VAC twice and it worked properly but now it will not work properly again, only if I put 100pF cap as I mentioned before(with 18V on Vbus).
Therefore sometimes the coil works properly and sometimes it does not.

I made almost identical Loneoceans UD2.7, just with through hole components. In place of optic fiber receiver I'm using arduino nano with optocoupler . Arduino comunicate with my phone over bluetooth. I wrote code for arduino that generate impulses with variable frequency (1 Hz - 20kHz) and on time (1 - 200 us).
Specs:
Bridge : 4x IXGN60N60C2D1 (full bridge configuration)
Primary resonant frequency : 149 kHz
Secondary resonant frequency : around 160kHz
MMC: 4 strings in parallel of 4 in series WIMA FKP1 capacitors for 68nF at 8000VDC
The coil is not big , secondary is around 23 cm tall and 7,5 cm in diameter. I plan to set current limiter to 300A.
Sorry for my bad English and thanks for the help.
Title: Re: Problems with my first DRSSTC
Post by: davekni on September 23, 2019, 01:45:37 AM
Not sure I can help with all the issues, but a couple things come to mind:

1) The gate-drive overshoot looks quite symmetric between rising and falling edges.  That's surprising with diodes across the series resistors, which usually make the waveforms asymmetric.  If most of the gate-drive-transformer leakage inductance is on the primary side, that would make a bit more sense.  With only 5 wires on your transformer, is the primary a single wire?  I'd suggest 8 wires, 4 twisted pairs, with one half of each twisted pair going to the primary gate drive.  (I saw that suggestion somewhere here - using Cat5 cable.)  Either of your cores should be fine.
2) I tried WIMA FKP1 capacitors initially on my latest coil.  They failed quite quickly.  These were 22nF 6kV parts running at +-4kV peak at 80kHz.  They degraded quickly until a couple parts ruptured their cases.  I rebuilt my entire MMC with Chinese induction cooker capacitors, 0.33uF 1200VDC, 630VAC.  Being wary, I started with an abuse test of a pair of parts, running at +-1200V 75kHz, 1% duty cycle bursts.  After this ran fine for 5 days straight, I increased the voltage/current to +-1700V.  Ran a couple more days with little sign of degradation!  Increased again to +-2050V.  After a day there, their capacitance had dropped about 5%, which is likely getting close to catastrophic runaway failure.  That's at 170% of rated peak voltage!
3) I'd suggest experimenting at MUCH lower than 3kHz pulse rate.  3Hz would be better.
4) The 100pF capacitor to ground suggests some noise coupling issue.  Grounded shielding around the control circuitry and common-mode chokes (large ferrite beads) around all connections from control electronics to power electronics helps too.  Even though everything is transformer-isolated, transformer capacitance causes high-frequency current pulses.  Common-mode chokes make those slower with lower peak current.
Title: Re: Problems with my first DRSSTC
Post by: Phoenix on September 23, 2019, 07:37:55 AM
Hello

Are you only using 4 turns for your GDT? I would use way more turns, the GDT of my DRSSTC has 14 turns for example. Your second core should be fine, I am using the same N30 material.

Greetings
Phoenix
Title: Re: Problems with my first DRSSTC
Post by: Hydron on September 24, 2019, 12:21:49 AM
4 turns sounds a little low, but remember it's at 150kHz or so - it probably doesn't need many to work well and you don't want to go overboard, as it makes leakage inductance worse. The issues I can see in the waveforms are not what I'd expect from too few turns - note that there isn't any droop in the tops of the waveform.

As for the OP's issue - I haven't seen quite as large an overshoot before, and certainly not one symmetrical like that when the speedup diode is included. Maybe try a small resistance in series that is not bypassed by the diode? Also could be worth playing around in ltspice or the simulator of your choice to try and recreate the waveform and then play around trying to solve it.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on September 24, 2019, 10:03:02 PM
Thanks for answers,
Today I tried to wind the GDT again , so I put 9 turns of CAT 5 cable on N30 core , replace 11 ohm res with 15 ohm and replace uf4007 diodes with STPS2H100 schottky diodes and I got a waveforms like this.(first img)
This is with signal gen connected to the feedback input but with feedback CT connected the coil will not work. So I guess that the feedback CT is problem. How I can fix this? I`m using 1:625 turn ratio for feedback.(second img) Is this good or not?
Title: Re: Problems with my first DRSSTC
Post by: Fumeaux on September 24, 2019, 10:40:36 PM
Isn't that a 1:1 Transformer? The white wire has 1 Turn on each other Core. You would need the brown wire to have a single turn on the middle core and the white wire one turn on the blue wire core.

That might be the source of your problems

Turns out it is not, my bad.
Title: Re: Problems with my first DRSSTC
Post by: Hydron on September 24, 2019, 11:17:24 PM
There is a crucial bit of info missing from the picture - the primary is a single turn (not shown) through the central toroid, and the twisted wires seen on the outer toroids are two ~independent secondary outputs.

With this known the 1:625 makes sense, however I can certainly see why it could also be interpreted as a 1:1 from the pic shown!
Title: Re: Problems with my first DRSSTC
Post by: bozidar on September 25, 2019, 12:36:44 AM
Sorry,
I forgot to mention that I use brown wire for OCD (24 turns) and blue for feedback(25 turns)  and the primary wire (one turn , not shown in the picture) passes through the core with white wire (25 turns) .
Title: Re: Problems with my first DRSSTC
Post by: davekni on September 25, 2019, 04:19:22 AM
Several thoughts:

Are you certain that the CT is connected with the correct polarity?  At low power, it shouldn't hurt to try reversing it just to check.

Looking closely at the last working scope image, the duty cycle is asymmetric at the start.  I'm not sure if this is typical for UD2.7 or not.  To me, it implies there is some DC offset.  I'd suggest switching back to your signal generator, and dial down the amplitude to 50-100mV P-P.  That should be enough to generate gate signals.  If it doesn't, then scope through the UD2.7 signal path to find the issue.  For asymmetry, a couple possibilities are that D1 and D2 aren't opposite directions, or that the quiescent voltages on the + and - comparitor inputs (IC8 pins 2 and 3) aren't close to the same (issue with the values of associated resistors).

At 155kHz, 4 turns should be fine for gate-drive with the cores you selected, but 9 won't hurt either.

The gate drive voltage is lower than most people run, but should be fine until you push primary current to the IGBT's peak rating or past.  If you are worried about the negative undershoot getting too large, either decrease the TVS voltage for the GE clamps or add a smaller resistors (2-5 ohms) in series with the gate resistor shunt diodes.

Good luck!
Title: Re: Problems with my first DRSSTC
Post by: davekni on September 25, 2019, 04:55:26 AM
One more thought on possible issues with CT feedback not functioning:

(BTW, hopefully continuing my last reply isn't what is banned as double-posting.  I can't figure out how to edit my last reply.)

Do you have a bleed resistor across your primary MMC array?  If not, the UD2.7 could be driving the IGBTs to the state that has the same voltage as is left on the MMC array previously.  That would induce no current, so no signal to reverse the gate drive. (Which is one reason to prefer self-oscillating controller circuitry.)

David Knierim
Title: Re: Problems with my first DRSSTC
Post by: Mads Barnkob on September 25, 2019, 08:40:33 AM
One more thought on possible issues with CT feedback not functioning:

(BTW, hopefully continuing my last reply isn't what is banned as double-posting.  I can't figure out how to edit my last reply.)

Do you have a bleed resistor across your primary MMC array?  If not, the UD2.7 could be driving the IGBTs to the state that has the same voltage as is left on the MMC array previously.  That would induce no current, so no signal to reverse the gate drive. (Which is one reason to prefer self-oscillating controller circuitry.)

David Knierim

A simpler and quicker fix for the MMC residual charge, is to place a 5-10K power resistor across the output of the inverter, this removed instability problems with my large DRSSTC.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on September 25, 2019, 09:02:33 AM
I only use signal gen as feedback input because when I connect feedback CT nothing works( also with reversed wires ). I used 4 turns on my previous core with high AL value (11400 nH/turn²). I'm using power supply with adjustable voltage (14v to 24 v) , so I adjust voltage to somewhere around 17 v to keep Vge overshoots under 30v and I use the same power supply for Vbus. ( 1.5KE27CA TVS is connected across gate emitter). I connect 10Mohm res across every capacitor( 4 strings of 4 in series).
Title: Re: Problems with my first DRSSTC
Post by: davekni on September 26, 2019, 04:54:35 AM
10Meg on 68nF is a 0.68 second time constant.  That may not be discharging in time.  Mad's suggestion of a 5-10k power resistor across the bridge output would be a good idea.  Perhaps that is the only issue - residual charge on the primary capacitors.  1-2 watts should be enough for the resistor - would cover 310V bus at 10% duty cycle.

If that doesn't fix things, then I'd suggest my previous signal-generator experiment. I understand (I think) what you did previously with the signal generator.  I'm suggesting a new test to debug the UD2.7 controller with signal-generator input, if residual charge isn't the only issue.  Feeding the UD2.7 current sense input with a sine-wave from the signal generator, ramp the voltage from 0 until barely seeing gate-drive signals.  A signal should start appearing around 25mV P-P.  By 50mV, the gate drive should be reasonably symmetric.  If significantly more signal-generator output voltage is required, then I'd set it to barely-enough and scope through the UD2.7 signal path from current-sense input to gate-drive output.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 01, 2019, 10:42:31 AM
At yours suggestion I put a 6 kohm resistor ( I found a few 3w and one 5w and connected them in parallel ) across bridge output and I continued testing with signal gen connected to the feedback input. Unfortunately  the signal began to appear just over 200mV P-P and the signal at the gates was not symetrical. Than I connect CT feedback instead of signal gen and increased the supply voltage to 24V ( also 24V on Vbus ) and the coil works just with not symetrical gate voltages as you can see in attached images. I noticed that the  signal is symetrical at the end where the current (feedback voltage ) is higer. So can i just reduce the number of windings on CT so that I get more voltage at the feedback input?
Are these negative peaks dangerous for transistors?
Title: Re: Problems with my first DRSSTC
Post by: Hydron on October 01, 2019, 01:39:15 PM
If you are using a diode to bypass the gate resistor on turn-off then a larger over-shoot of gate voltage in the negative direction is expected. If you ignore the over/undershoot (only consider the flat portion of the waveform) then the voltage should be symmetrical. Finally, fewer turns will give a stronger feedback signal which may be useful if oscillation is a bit iffy at lower currents, but you do need to be careful not to over-do it and cook the driver!
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 02, 2019, 05:07:29 AM
After reading your results, I examined the UD2.7 schematic some more, especially the biasing of comparator IC8 (TL3116).  Your 200mV result is actually reasonable for the circuit.  The negative input will be pulled down by 16uA across R27 into the forward drop of diode D1.  (I hadn't previously noticed that R27 bias current, so was using the R3/R4 voltage divider as the presumed DC quiescent level of IC8 negative input.)  I think that's less than ideal, but appears to be inherent in the UD2.7 design.

One option is as you suggested, lower the current-transformer turns ratio.  As Hydron already noted, make sure the output current doesn't get above ~1A when you ramp up power.  (At higher bus voltage, this 200mV issue may be insignificant.  The coil current will ramp up quickly, so soon overdrive the 200mV level enough to make symmetric 50% duty-cycle gate waveforms.)

Another option is to remove 100k resistor R27 and place it in parallel with D1 and D2.  That will make the UD2.7 respond to current sense signals less than 200mV, and improve gate drive duty cycle symmetry.

Others here have more experience with gate-drive undershoot.  From what they say, I gather it's not too concerning.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 08, 2019, 09:53:21 AM
After my last post  when everything was worked , yesterday when I turned it on again nothing worked as it should. So I just don't know what to do  :(
That's happening from beginning , one day it works and when I try the other day it doesn't work properly. Only works if I connect signal gen.Than I reduced feedback CT turn ratio to 1:450 but results were the same.
In attached img can be seen gate signal ( yellow trace )with CT  connected to the feedback input. ( blue trace - current) It's weird that things get a little better when I touch one of the feedback inputs with a screwdriver  :o
What can cause this behaviour , driver , CT , GDT or something else?
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 09, 2019, 04:57:19 AM
It looks like the first gate-drive half-cycle isn't always generating enough primary current to start the oscillation.  One of your earlier scope images that zoomed in to the start of a burst (when it was working) showed about 1V of current sense on the first gate-drive half-cycle.  That should be sufficient, especially with your test showing that 0.2V P-P was enough to get some gate signal.  It's hard to tell on this latest overview scope image what the difference is between the bursts that function vs. the ones that don't.  If it is possible to zoom in on the start of a failed burst, that would be great.  Especially helpful if compared to a good burst (as most in your latest scope image), and if the two H-Bridge outputs can be included in the traces.

For fixes, the only thing that comes to mind is to change the resistors you added (at Mads suggestion) from across the H-Bridge output to two separate ~5-10k resistors, one from one output to the VBus + supply and the other from the other H_Bridge output to the VBus - supply.  You'll need to try both ways (swapping which H-Bridge output is puilled up and which one down).  One connection should make it not work at all, and the other will double the current on the first half-cycle.  Hopefully double-current will get it above the threshold for the UD2.7 circuit and cause it to reliably start oscillation.
Title: Re: Problems with my first DRSSTC
Post by: Mads Barnkob on October 09, 2019, 09:09:19 AM
It's weird that things get a little better when I touch one of the feedback inputs with a screwdriver  :o
What can cause this behaviour , driver , CT , GDT or something else?

It is clear that you have a too weak feedback when you can get better performance when using yourself as a capacitively coupled antenna on the feedback.

I would look at all components from feedback op-amp -> diodes -> CT

Did you rule out phase error? Reversing primary connections (remember to place CT on the low voltage end of LC circuit), phase of CT, phase of GDT. It is possible to get unstable operation when driving things off-off or out-of-phase just from the leakage energy through the system.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 10, 2019, 08:42:21 PM
I checked the diodes and replaced chinese chip LT1016 with MAX9013EUA (from farnell). The results are still the same. :(
I also checked phase of CT and phase of GDT.
In attached img can be seen some waveforms (filename of first 7 img is the frequency and on time).
At 1.9kHz and 196us everything is almost good and in attached img you can see what happens at 1.9kHz and 200us.
In last 4 images you can see bridge , driver etc.
Hope this is not too many pictures  ;D
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 11, 2019, 06:19:09 AM
Thank you for all those pictures!  It's much easier to analyze issues when there is actual information rather than guessing how to fill "between the lines".  (One question, though:  What bus voltage was used for the scope images?)

I'm presuming the third image is of the current feedback, 450:1 on 51-ohm load.  It has a particular anomaly where the current reverses 3V to -3V very rapidly (~+26A to -26A).  If that condition is repeatable, it would be a good place to focus.

Looking at the wiring pictures, I see one important issue:  The IGBTs have double emitter terminals.  The intended use is as a Kelvin connection - one terminal is used for high current (C-E current), and the other only for gate drive.  That avoids inductive and resistive voltage drop from the high current affecting the gate signal.  The pictures show the collector current from the low-side IGBTs passing through the high-side IGBTs, in one emitter terminal and out the other.  That definitely needs fixing.  The gates may not be damaged yet, but that is likely contributing to your operational issues, and will likely fry the gates eventually.  For each of the four IGBTs, one emitter terminal should be used for the gate-drive signal (and nothing else), and the other for all power connections.  Most bricks have small emitter terminals for gate-drive return.  This part happens to have two large interchangeably connections, but it's still necessary to pick one for gate-drive return use only.

Some less critical suggestions:
The gate-drive transformer will have lower leakage-inductance if the four paralleled primary wires of the CAT5 cable are individually twisted and ran back to the driver board, then patched to a single pair of wires at the driver.  That reduces the inductance of the relatively-long wiring from the driver-to-transformer by 4x.  That's an improvement tweak - not the cause of your man issues.
Also, I'd periodically measure the capacitance and leakage current of each quarter of your MMC.  As I'd mentioned earlier, my initial DRSSTC MMC used WIMA FKP1 capacitors, and they failed quite quickly.
Finally, the local VBus bypass capacitors (snubbers I see them called, although I use that term differently, for R+C series networks added across switching voltages) have fairly high wiring inductance for the fast IGBTs you are using.  Multiple parallel (smaller if necessary) caps will lower inductance, or copper plane/foil connections for the Vbus + and - rails.

Good luck!  Thank you for the bountiful scope traces and pictures.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 11, 2019, 11:34:59 AM
Vbus is 24VDC.
Yes , the third picture is of the current feedback (450:1 on 51 ohm load) and yes , anomaly where the current reverses is repeatable. I think that's what happens when the gate signal have small duty cycle ( second picture).
I didn't know that for double emitter terminals. I'll fix it as soon as I can.
I'm using four paralleled primary wires but i patched them to a single pair of wires right next to the GDT.I shielded all wires with aluminium foil and put everything in shrinking tube.
In my local shop I can only find these caps for MMC so I hope that they ( 68nF 8kV) can handle  4.5 kV peak at 150kHz. I checked all the capacitors separately , with RLC meter , and they are all good.
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 12, 2019, 04:41:19 AM
I'm guessing there's another issue besides just the emitter wiring.  At 24V and ~26A, the emitter connections seem unlikely to cause the anomalous behavior (fast current reversal event and resulting short gate pulse).  The other things I'd suggested are in the minor-improvement category, not causes of the issue either.  Still suggest fixing the emitter wiring before any higher-power testing.

Is the current-limit circuit fully wired on your board?  If there were to be an open (no load) on the current-limit half of the current transformer, that core is likely to saturate, causing odd signals on the other half of the current sense transformer.  Not sure that could explain exactly what you are seeing, but worth a check.

Would it be possible to capture that anomalous event at 24Vbus and ~26A with the following four scope signals:
    Current sense voltage (as you had before)
    Both H-Bridge outputs
    High side of primary caps (connection to the primary coil).
If you have only two scope channels available, make three captures.  The current sense signal would be common to all three captures, with the other probe monitoring each of the three other signals sequentially.  At ~26A, the primary cap high-side should be ~400V, so within capability of normal scope probes.

This is interesting debugging remotely, where I can't just move a scope probe from one point to the next to see where an issue lies.  If my probing requests become too tedious, please feel free to say so.  If you don't mind continuing to capture signals, I'm confident it will be possible to find and fix the issue.  (If someone else manages to recognize the likely cause from the symptoms, that will be great.  If not, then walking through signals with your scope will get there.)
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 17, 2019, 05:53:09 PM
I found out there was a problem in my oscilloscope (settings) that caused the signal not to display properly  ;D
I have a USB oscilloscope ISDS205B (20MHz, 48Ms) that can measure up to +-60V so I can`t measure voltage at  high side of primary caps  :-\
I fixed the emitter wiring and now everything is almost fine.When I turned on the coil for the first 2-3 minutes everything worked again as before, some impulses were skipped, but after that everything slowly returned to normal.
Here is some scope images after that.
First two pictures- yellow trace is feedback input from CT and blue trace is primary current (1V=1Amp)
Pictures 202,203,204 -yellow trace is Vce from one transistor and blue trace is primary current
Pictures 205,206,207 -yellow trace is Vce from other transistor and blue trace is primary current
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 18, 2019, 06:30:06 AM
Great!  Nice to see scope traces that look as expected.

The occasional startup issues at 24Vbus are reasonable for the UD2.7.  The 1.3A current of the first half-cycle fits calculations (15.7 ohms primary L and C at 149kHz on about 21V bridge output).  1.3A through 625:1 CT into 51 ohms gives 0.106V, barely enough to trigger the UD2.7.  This issue should completely disappear at higher Vbus.  (I'd previously suggested some UD2.7 patches if you want to get more reliable operation at low Vbus, but no need to change anything with Vbus above 24V.)

How are you scoping primary current on the blue channel?  (Cyan channel in my color-science world:)  Do you have a current probe, or a resistor in series?  I'm asking because the CT output appears to have 20-25 degrees of phase lead relative to that current.  Do you have L1 in the circuit now?  That would explain the phase lead of CT signal.  (The parasitic inductance of the CT would account for only 1 degree or so based on your N30 core and 25 turns for the second stage.)

Given that you have primary current scoped, primary voltage isn't necessary.  If you want to scope such in the future, inexpensive 100x scope probes are available, such as this P4060:
https://www.ebay.com/itm/P4060-1-100-High-Voltage-2KV-2000V-60MHz-Oscilloscope-Scope-Probe-100X/372463710732?hash=item56b891320c:g:eLIAAOSw8UZaLlMi

The occasional short positive spikes on the H-Bridge outputs might be due to not quite enough phase lead.  If the IGBTs switch after zero-current, the reverse diodes of a pair of parts are conducting.  The other pair then switches on, rapidly removing the diodes stored charge, then overshooting as the diodes snap-off.

The longer spikes when the drive pulse ends may be on the edge of problematic at 310Vbus.  This is caused by an L/C circuit ring, where C is the smaller Vbus capacitors at the H-Bridge and L is the wiring inductance back to the larger bulk Vbus capacitors.  That ring will grow in proportion to primary current at the turn-off point.  Over-current shutdown, by definition, occurs at high primary current, causing high voltage spike on Vbus at the IGBTs.  (This issue caused a rework in my Vbus wiring, as I'm using 600V IGBTs at 450Vbus.)  At your 310 Vbus, extrapolating the spikes, they should be a little below 600V, but without a lot of margin.

To reduce that Vbus spike at the end of high-current pulses, either increase the capacitance local to the IGBTs or reduce wiring inductance back to the bulk Vbus capacitors.  Using a few wire-ties to force the Vbus wires together will help.  Multiple pairs of Vbus wires, each pair twisted or tied to remain adjacent, will help even more.  That's what I did, as I had no room for more local-to-IGBT caps.

Have fun - nice progress!
Title: Re: Problems with my first DRSSTC
Post by: bozidar on October 22, 2019, 07:44:55 PM
I`m scoping primary current with another separate CT (41 turns connected to 41Ohm resistor).
I've already adjusted phase lead before with proper oscilloscope as best as I can, so now I`m afraid to adjust it again (unfortunately I had limited access to that oscilloscope). First picture that I posted is taken by that oscilloscope.
After some experimenting at low voltages I decide to power bridge at full voltage (310VDC) and the coil didn't work at all (only some small sparks at 7KHz and 9us on time).
Then I replaced switching power supply which powers the driver (24VDC 10A ),because that's the only thing I didn't check, with classic iron core transformer with greatz , some capacitors and 24V regulator and finally the coil works (still skipping some pulses at low frequencies below 100Hz but it does not matter)  :D :D :D
I guess there was some interference between driver and switching power supply  :o
The coil produce around 30cm sparks at 200Apk , 1.5KHz and 30us on time. Given that the secondary is only 23cm tall I think this is a good result.
I'll post a video soon.
Thanks for all your help, especially davekni for the great tips and explanations.
Title: Re: Problems with my first DRSSTC
Post by: davekni on October 23, 2019, 04:33:46 AM
Yes, I can imagine a switching supply being troubled by the nearby electromagnetic fields.  The cheap DMMs get completely confused by stray fields.  On the other hand, I have good luck with old Dell laptop 19V supplies, which are switching.

Sparks might be longer with longer pulse width and lower repeat frequency (to keep average power constant).  Other's likely have better intuition on that.

Looking forward to the video!  Glad I could assist - thank you for the mention.
Title: Re: Problems with my first DRSSTC
Post by: Mads Barnkob on October 28, 2019, 08:41:19 AM
The longer spikes when the drive pulse ends may be on the edge of problematic at 310Vbus.  This is caused by an L/C circuit ring, where C is the smaller Vbus capacitors at the H-Bridge and L is the wiring inductance back to the larger bulk Vbus capacitors.  That ring will grow in proportion to primary current at the turn-off point.  Over-current shutdown, by definition, occurs at high primary current, causing high voltage spike on Vbus at the IGBTs.  (This issue caused a rework in my Vbus wiring, as I'm using 600V IGBTs at 450Vbus.)  At your 310 Vbus, extrapolating the spikes, they should be a little below 600V, but without a lot of margin.

To reduce that Vbus spike at the end of high-current pulses, either increase the capacitance local to the IGBTs or reduce wiring inductance back to the bulk Vbus capacitors.  Using a few wire-ties to force the Vbus wires together will help.  Multiple pairs of Vbus wires, each pair twisted or tied to remain adjacent, will help even more.  That's what I did, as I had no room for more local-to-IGBT caps.

Low voltage testing is notoriously known to be misleading when it comes to switching spikes for IGBTs in a resonant circuit.

Quote from: From a Powerex application note
Coes – Output Capacitance
This is the output capacitance measured between the
collector and emitter terminals with the gate shorted to
the emitter for AC voltages. Coes is made up of the
collector to emitter capacitance (CCE) in parallel with
the gate to collector capacitance (CGC), or

Coes = Cce + Cgc

For soft switching applications, Coes is important
because it can affect the resonance of the circuit.

 [ Invalid Attachment ]

Kizmo explained this: As you can see the IGBT open state capacitance is largely depending on the voltage that is applied over Collector-emitter! And imagine what happens at your bridge when for example high side transistor turns on and low side transistor turns off. The emitter of the low side brick is tied to negative bus rail, and collector is quickly whacked to the full bus voltage which means the parasitic capacitance inside of the brick will be charged to your dc bus voltage at very high voltage rise time. And that charge current must travel through all parasitic inductances that your bridge has.

As you increase your bus voltage, your parasitic capacitance will decrease and amplitude of these transients related to the bus voltage will go down. My big coil does that too, at 100V bus the transients are almost 200% above bus voltage. But at 650V they are only maybe 20% over.
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 01, 2019, 04:39:28 AM
Mads,

Great point about IGBT capacitance vs. voltage.  That applies to the fast spikes that show up occasionally at the switching transitions.  (I had a similar issue on my resonant PFC at full voltage, but it was caused by diode recovery charge rather than capacitance, when switching after current zero-crossing.)

The lower-frequency ring at the end of the drive pulse, the spike that lasts a couple microseconds, isn't due to IGBT capacitance.  It's rather the local Vbus capacitance at the bridge, resonating with the wiring inductance from the bridge back to the bulk capacitor.  That ring voltage will be a relatively linear function of current at the end of the drive pulse.  (There is a small relatively-constant part due the the IGBT's diode forward voltage drop.  Most is due to the wiring inductance and local Vbus capacitor.)  This is the problem that caused me to redo wiring from bulk capacitors to H-Bridge on my 40-IGBT DRSSTC drive, to lower inductance.  This spike drops by that resonant impedance, so proportional to sqrt(wiring inductance) for given local H-Bridge VBus capacitance (360uF in my case - 12 30uF PP caps local on the H-Bridge).
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 09, 2019, 08:14:08 PM
Unfortunately, the switching power supply was not a problem, because when I turned on the coil again at 310VDC two IGBTs exploded  :(
Luckily i had 2 spare IGBTs, and now my problem is the same as before. Sometimes the coil works properly and sometimes it does not. When I touch the feedback input with a screwdriver everything works as it should (Vbus 24VDC). I noticed when I disconnect 24VDC from bridge and touch the feedback input (with a screwdriver) I can hear audible buzz on GDT. And when I remove a screwdriver there is no more buzz on GDT. Then I connected only GDT (without OCD and feedback CT) and the same thing with a screwdriver happens. In attached images you can see gate-emitter voltages with (11.jpg) and without (10.jpg) a screwdriver.


This is written on the loneoceans site:
www.loneoceans.com/labs/ud27 (http://www.loneoceans.com/labs/ud27)
Quote
Note that if you have no input signal in your FB jacks, the driver will not see any feedback, causing the drive to lock up. Here's a scope shot of a GDT output with a feedback input signal (left). But if you have a GDT attached with no signal generator input on the feedback, this will make a signal on the output of the GDT look like the above (right - where the driver doesn't know what to do), and will cause quite a loud audible buzz from the GDT with this associated output per pulse. This sound should diminish greatly with a feedback input signal.


So without feedback , the signal should look like last attached image.  Right?
Maybe my driver board is bad?
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 11, 2019, 02:01:12 AM
If I'm understanding the scope trace context correctly, then yes there does appear to be some issue with the driver board.
For these scope images, are you sending roughly 15us wide enable pulses every 266us to the optical input?  I'm guessing that from the scope plots.

The gate drive appears to be low amplitude, just under +-7V for the largest pulses.  Also, it appears that the gate drive is not being shorted between enable pulses as it should be.  Between enable pulses, the NFET halves of U1, U2, U3, and U4 should all be turned-on (high output on pin 5 of IC9 and IC10).  When the two IGBTs failed, it's quite possible that they sent a spike back through the CT large enough to fry U1-U4.  Or, it could be the reverse, where something failed on the driver first, causing low gate voltages, which triggered the IGBTs to fry.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 11, 2019, 10:19:52 AM
Here is some scope images with and without a screwdriver and circuit of my driver (Vge - blue trace). 1kHz 200us on time - input signal. I connected only GDT ,without OCD and feedback CT .
A similar thing with a screwdriver happened before IGBTs exploded. I wrote this in one of my past posts.
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 12, 2019, 06:20:41 AM
Thank you for sharing the schematic.  Looks like there's a signal inversion issue relative to the original UD2.7 schematic.  IC3 needs to be an inverting driver, UCC27423.  With the non-inverting driver shown in your schematic, the gate drive is floating between enable pulses, causing probably all of your issues.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 12, 2019, 06:44:36 PM
When buying parts for the driver I couldn't find the UCC27423 chip, so I thought I could use the UCC27424 chip.

How the gate drive can float between enable pulses when enable pins are connected to 9V?

After some experimenting with LTSpice I added a small circuit  (Picture1.png) to improve impulses at the beginning of the input pulse(from interrupter). In attached images you can see gate-emitter voltage (blue trace) and output pulse from AND gate (yellow trace) (second image) , input (blue trace) and output signal (yellow trace) of added circuit (1.png and 2.png) (Vbus 0V). After that I connected 24VDV on the bridge and the coil works  :D
The only problem is that sometimes some pulses are skipping at some frequencies, much like the one in the picture from the earlier post (https://highvoltageforum.net/index.php?action=dlattach;topic=722.0;attach=4990;image (https://highvoltageforum.net/index.php?action=dlattach;topic=722.0;attach=4990;image)). I don't think it should be able to damage IGBTs at full Vbus. What do you think?
Another question, it may be stupid but I have to be sure. Should I connect the ground to the driver board when it is mounted in a metal housing?
I connected the other end of the secondary to the main ground as well as the IGBT coolers, metal case for the driver board and strike rail.
Title: Re: Problems with my first DRSSTC
Post by: dexter on November 12, 2019, 11:08:44 PM
When buying parts for the driver I couldn't find the UCC27423 chip, so I thought I could use the UCC27424 chip.

i made the same mistake on my first DRSSTC and couldn't get why my coil didn't start

the gates of the p channel mosfets are capacitively coupled so they need a high to low transition in order for one of the p's to conduct

if you use a non inverting driver they'll get a low to high transition and won't turn on

with an external signal the coil will work but you still loose the 1st cycle
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 13, 2019, 12:35:34 AM
Can I just invert the input signals (2 and 4 pins of ucc27424) with 74hc14?
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 13, 2019, 06:36:57 AM
Besides what dexter pointed out, there's another issue with non-inverting gate drive.  At the end of each pulse, the driver outputs are both low, leaving NFETs Q1 and Q2 off.  PFETs Q3 and Q4 will be initially on, but will turn off after 50-100us when C15 and C16 charge.  That leaves the gate-drive transformer input floating (no FETs on).  The residual gate transformer flux will then cause a gate drive transient, which can be seen in your previous scope plots.

Inverting inputs (pins 2 and 4) with HC14's should work.  It does add some extra delay, so you'll need more phase-lead (higher L1 inductance) to keep the gate signal leading the zero-current point.  (That's less than ideal, more delay and more compensation for the delay, but workable.)

Recommend not running at 310V until the IC3 inversion issue is fixed.  Without fixing that, the occasional half-voltage gate signals will likely fry IGBTs.

In general, I'd say yes, ground the driver board to its box.  That's the way my coil runs.  It's best to avoid ground currents running through the board, so I'd suggest connecting to the board ground near the power input connector.  The other connections to the driver board are all optically or transformer coupled, so shouldn't inject ground currents.

Your added circuit looks workable for kicking the oscillator at startup with low Vbus.  Have you looked at a simpler alternative:  Switch the right side of R18 (100k) from pin 7 to pin 8 of IC10.  That should make it self-oscillating at a frequency determined by C20.  (My DRSSTC has a discrete-transistor comparator implementation, but uses that topology to be self-oscillating.)  The CT feedback will override the self-oscillation frequency, so it still locks in fine.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 13, 2019, 09:08:45 AM
Ok then I will buy the UCC27423 chip, it`s better than with 74HC14 and UCC27424.
I have GBU808 bridge rectifier (800V 8A). Is that enough for a 230AC or should I buy another that can handle more current?
I had a FB2510 (1000V 25A) but it burned out when the IGBTs exploded.
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 14, 2019, 05:35:19 AM
The input bridge current depends on how hard you are going to run your coil - frequency and width of the enable pulses.  If you are measuring line current, just keep it under 8A or buy higher current diode bridge.  BTW, I recently bought some cheap bridges from China on EBay, supposedly 50A at 1000V, but they were worthless, perhaps good for 5A.  Years ago I'd gotten some of these that worked OK, at least at 15A.

Once the driver is swapped to non-inverting, it will require reversing either the CT feedback or GDT terminals to get the feedback phase back to correct.
Title: Re: Problems with my first DRSSTC
Post by: bozidar on November 24, 2019, 07:26:29 PM
Since my chip (UCC27423) didn't arrive in time I couldn't wait any longer and just inverted the UCC27424 inputs with 74HC14 and also adjust a phase lead a little bit , and now DRSSTC works perfectly. :D
Thanks again for all your help.
Here is some videos.
Title: Re: Problems with my first DRSSTC
Post by: davekni on November 24, 2019, 10:10:06 PM
Great!  All your hard work to get to this point will seem less burdensome now that you have results.
Title: Re: Problems with my first DRSSTC
Post by: Mads Barnkob on November 25, 2019, 10:16:22 AM
Congratulations on getting it a working state!

Your topload seems very uneven on the surface, since you have at-least two other places with breakout than from your break out point. You will get longer sparks when they just come from one place, to get that topload much more smooth and maybe use a material with larger diameter for the break out point to avoid corona on the sides of it.