Ok now I'm stuck ¯\_(ツ)_/¯

[dj.cosmo.esq]

Ok here is my first technical post so please be gentle. Ok I jumped into the drsstc thing from spark gaps. Although a lot of things are similar there's still a hell of a learning curve but if I can figure out these little quirks i know I'll be ok. So im running the cloud leopard driver from China which was on a whim from banggood for like 30 bucks with the interrupter. The GDT is humming along switching a half bridge of 1MBI600PX-140 igbts both have 2kvdc 2.5uF fabmika caps sitting on the bricks. Two 350v 4200mfd electrolytic caps in series run voltage doubled.  Of course TVS to block any surge. My MMC is 9 series 2 parallel of Cornell Dubilier UPP10W3P3K 3.3mfd 1050vdc caps. So 9450v at .7333uF and I have 9 more if needed to do 3 strings but they have never even gotten warm when just running 1 string of 6. Here's now my dilemma, I can't get this coil to go be past 2 foot arcs. I have tapped and tuned to the best of my (limited) knowledge and it's maxed out. I'm assuming with this setup I should be able to get some good arcs? I used a scope and got my fres of about 58khz and the adjusting it to about 49khz I get 2 foot arcs. Now here are the noob questions(sorry ahead of time). For the life of me I can't get feedback to drive the damn coil. So I'm using a signal generator to run it but even with as much as I'm pushing into this coil 2 feet seems really small... Any input or silly oversights on my part are highly appreciated.
Thanks, Courtney

[Mads Barnkob]

Running a DRSSTC from a signal generator = hard switching = dead IGBTs

You need to fix the feedback before trying to make bigger sparks.

I always recommend people to use a Steve Ward 1.3b or 2.1 Universal Driver for their first DRSSTC. It works, it is simple and many people knows how it works. The Chinese drivers are hit and miss on documentation due to loss in translation, if there is a translation.

Here I kill a bridge with a signal generator

[dj.cosmo.esq]

Thanks for the reply Mads. I feared that was  going to be the answer. So this is what I know the input for the feedback and ocd do work because I can trigger them artificially. Is there a size limit to the core or the feedback? Like if the core is too big it won't pickup the current? I have some small ft-140-75 and some larger ft-240-77. I've tried cascading but nothing I've only observed feedback once but somehow the wire melted going through the toroid.  I'd there a easier way to get feedback like maybe 2 or 3 turned of primary? Or is it possible it's my startup procedure? I start the driver , then Interrupter, then the bridge with a variac set to zero and slowly ramp up. In order for the feedback to work it has to get started from somewhere and once the power is up enough on the bridge if it hasn't switched the gate at all how does start? Do I have to then turn the driver on and off to Kickstart maybe?  Is it the whole cascading the trick because I've seen the switch to the three toroid recently where the larger wraps 1 turn around the two smaller cores? More turns or less turns? I'm so close yet this feedback seems like a unicorn I'm trying to catch.
Thanks again,
Courtney

[davekni]

To repeat one of Mad's suggestions from other posts, do you have bleed resistors across the H-Bridge output?  Most drivers are not self-oscillating without feedback.  The first half-cycle after enabling needs to create enough feedback signal to trigger the next half-cycle.  Depending on the remaining charge on the MMC, the first half-cycle may be driving it to the same state it's already in, so create no feedback signal at all.  Bleed resistors across the H-Bridge output forces the MMC to neutral (no charge) state between drive bursts.

[dj.cosmo.esq]

Thanks davekni and Mads. Ok ok ok I see what you mean about the voltage ramping up.and the bleed resistor immediately draining the output to start say a swing or oscillation of sorts? In that case, Nope, no bleed resistor! I swear I've read every post on this forum but I must have misunderstood the context. I apologize about that RTFM moment. Now my question would be what rating on the resistor!?! Cause that's a nice chunk of power!

[Mads Barnkob]

Thanks davekni and Mads. Ok ok ok I see what you mean about the voltage ramping up.and the bleed resistor immediately draining the output to start say a swing or oscillation of sorts? In that case, Nope, no bleed resistor! I swear I've read every post on this forum but I must have misunderstood the context. I apologize about that RTFM moment. Now my question would be what rating on the resistor!?! Cause that's a nice chunk of power!

Luckily it is not the full charge of the MMC you need to drain, but a low voltage residual charge. I have used 10-15 kOhm 50 Watt power resistor, but you can just calculate it from the inverter output voltage and resistor resistance to get the power rating needed. It is even also a very low duty cycles (2-5% for a DRSSTC).

[dj.cosmo.esq]

Thanks Mads! And just to double check I would be putting the resistor from the collector to emitter, parallel with the output of the MMC and NOT from the output of the MMC to the other side say bypassing the primary in essence shorting the MMC bank with the resistor? Correct?
Thanks, Courtney

[Mads Barnkob]

Thanks Mads! And just to double check I would be putting the resistor from the collector to emitter, parallel with the output of the MMC and NOT from the output of the MMC to the other side say bypassing the primary in essence shorting the MMC bank with the resistor? Correct?
Thanks, Courtney

Yes, it goes directly on the inverter output and across the LC primary circuit. This is mostly for convenience, it could go across the MMC and do the same job, but you just suddenly need a 10-20 kV voltage rating instead of 1 kV.

If your MMC have individual bleed resistors, that will properly work as well, but that it not seen so often anymore, when we use large snubber capacitors

[davekni]

Since this is a half-bridge, you'll need to try the bleed resistor across one transistor's C-E, then across the other.  It needs to be across the transistor that is in the OFF state for the first half-cycle of oscillation.  Not knowing the controller or phasing of your GDT, can't say which transistor needs the bleed resistor.  If you have a scope, you could capture which IGBT is on first at the start of a burst, rather than guessing.  Put the resistor on the other IGBT (the one that's not on first).

[wuschel]

Hi,
im looking over your JavaTC Calculation. Im thinking there is something wrong with the Toroid calculation.
The Toroid from this calculating must be an Monster.
The resulting Frequency must be higher.
You can test it with the "Helicat calculator" from mad Barnkobs Homepage.
PS 2 Feet Arcs are not so Bad for an first try.


Have Fun

[dj.cosmo.esq]

I'm pretty sure it's right or very very close I'm still smoothing the darn thing out but  I must admit it was a ton of work but it's slowly coming out to a nice finish. I saw a tutorial of sorts on good ol YouTube

and they guy was saying how to get a (close to) spun aluminum toroid look. So I decided to go with his idea but tweak it just a tad. I went by Mads suggestions on his site (or tried) by going with the diameter of the coil for the minor of the toroid and the length of the coil for the major. Definitely correct me if I misunderstood because I'm here to learn as much as possible. Anyways 5 cans of spray foam later. I usually never take pics but I attached a couple. As you can see my work space has been overrun by the BMW

[Mads Barnkob]

Your JavaTC calculation is correct, I am not sure what wunchel is refering to by my helical coil calculator, since its for coils, maybe that I got a example of a toroid capacity in it: http://kaizerpowerelectronics.dk/calculators/helical-coil-calculator/ but that is 26 pF, close to yours.

Secondary coil diameter = toroid minor diameter and secondary winding length = toroid major diameter gives you a secondary system that has proportions optimized for electromagnetic field shaping, charge storage, getting distance to the primary coil out from the primary coil and it easily goes along with hitting the magic 50k impedance.

It is a very nice toroid you made, the one I made for my DRSSTC1 with epoxy was way too much work and I should have gone with foam But aluminum tape is hard to get all flat and shiny, he must have polished it up in his video, the one you embedded.

[dj.cosmo.esq]

Yes Mads it looks to be that way. I figured out a trick to actually getting the wrinkles and imperfections out.  Since it's only aluminum foil and adhesive I basically take multi oscillating tool and 3m 2000 grit wet dry sandpaper and run over all the lines which makes the lines smooth relative to the overlapping piece, then I use 3000 grit and erase it completely. The result after a light polish is like a mirror. Initial test have shown no corona discharge from anywhere except the breakout point. So much so I had to polish the breakout point because it ended up being the most jagged surface and it was to the naked eye perfectly smooth! Now it is only discharges from the very tip of the tungsten rod resulting in larger and cleaner streamers. It's a ton of work but I'm pretty sure that the roughly $35-40 in materials is way less than getting a spun aluminum toroid in the size I needed for this project.

[dj.cosmo.esq]

Ok I have just I few questions all. I would make a different post all together but it seems like everyone gets stuck at this part It seems like a magical mythological unicorn hybrid! I did a full bridge with the same setup and changed out the caps to Rubicon photo flash caps 360v 2000uf. I still can't get this damn thing to oscillate a few questions. Does the core have anything (size wise) to do with feedback? Does the type of wire matter (stranded vs solid)? I put a 10k resistor across the bridge out and I tried it on the half bridge on both sides. I tried phasing and the commercial units I have into the dot out of the dot! With a frequency generator it works but I know on time is critical. I wound 1:33:33 a 16:16 paired and crossed to make a 32:32 and finally a 1:50. In everything from 1 inch to 2 inch core with both material #75 and #77 high permiability and work awesome for GDTs. Is possible to make just a quick circuit to add into this board to just do feedback? Thanks
Courtney

[davekni]

Do you have an oscilloscope available?  I'm not sure how to help without some scope traces.

Have you found schematics for your purchased controller?  A couple possibilities come to mind at the moment.  One is that the controller needs more feedback voltage than most.  (UD2.7 is the only controller I've studied other than my own home-made one.)  To test this, you could run (at low power) from your signal generator, then lower the signal generator output amplitude until it quits working.

The other possibility is that the controller is set to run continuously rather than enabled in pulses.  Most drivers create one gate-drive edge at the start of each enable pulse.  That edge needs to create enough feedback to start the oscillation.  If either that initial edge is missing, or the generated feedback amplitude is too low, then it won't work.  (Scoping that gate-drive output, half-bridge output, and feedback signal would answer a lot of questions.)

[dj.cosmo.esq]

Thanks for the insight David. I ran it like you said and when I turn the signal generator slowly down there comes a point where it stops running and the green indicator led fades off. The weird thing is if I turn the generator off somehow there is a signal that's enough to make the circuit oscillate coming from I'm guessing 60hz line noise. If amplitude is my issue how can I bump it up? I havent found a schematic for the board cause I think a lot of it is done with one chip (FPGA) and a pretty standard in/output stage. It uses mic4423 instead of the UCC ti chips I've used previously. I'll get scope shots soon. Thanks again...

[davekni]

When you get to scoping, it would help to measure what signal-generator output amplitude is enough to barely make the driver function.  That would be reference point for scoped feedback signal.

When you say "enough to make the circuit oscillate", is this just single gate-drive pulses at ~60Hz?  If it is line noise, there may be some grounding issue.  A couple other possibilities:  Is the interrupter set for some fixed frequency of enable pulses?  The remaining signal could be single gate-drive pulses at the start of each enable pulse, as the typical UD2.7 etc. drivers do.  Another possibility is that, unlike UD2.7 etc., this driver is self-oscillating without feedback.  (Self-oscillating drivers don't appear common on this forum, but it's what I build for myself.  With self-oscillation tuned close to the operating frequency, the feedback signal can build over several cycles to get large enough, rather than needing to get there in the initial half-cycle.  No need for a resistor on the bridge output either.)  If this driver is self-oscillating, it may not be close to your resonant frequency.  Scoping should reveal the real situation.

If feedback amplitude isn't sufficient, the solution will depend on what amplitude is required.  The UD2.7 driver circuit requires around 100mV feedback to start oscillating, into a 50-ohm load on the current transformer secondary.  If your driver is similar, but your feedback amplitude is lower, then we need to figure out why your amplitude is low.  If your driver requires much higher amplitude, then either an external preamp or modification to the driver is in order.  (In this latter case, you could see if it's possible to trace the feedback input circuit on the driver - load resistance, clamping, comparator circuit - the signal conditioning circuitry before entering the FPGA.)

Good luck!

[ritaismyconscience]

I have the same board and I got it to oscillate from feedback at around 30V on the bus. Could you maybe send pictures of your feedback setup?

The board works like this: so there's some sort of fancy data transmission going on between the interrupter and the controller, but every time the interrupter sends out a pulse, the controller detects it and sends out a single pulse on the GDTs. It then listens for feedback from the 1:1023 transformer, which allows it to oscillate further.