20KW DRSSTC / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

20KW DRSSTC

first previous 8 9 10 11 next

Move Thread

LAN_403

Marko

Thu Apr 12 2007, 11:06AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Guys, anyone noticed finn's idea from before?

hammer wrote ...
I have been working on a way to predict the zero crossing in advance, on a level where there is no PLL or other strange stuff involved.
I have gotten it to work in Orcad, this is how it is:

At the resonant frequency, let`s assume that when the current has fallen to 1/10th of the peak current, then there is 1ÂµS to the zero crossing.
I`m not using real numbers here, just some nice numbers to help draw the picture.

Now, if you take the signal from a current transformer monitoring the primary current, and rectify it, you have a pulsating DC signal that indicates the primary current.
You use this signal to do 2 things:
You sample the signal, and hold it at the peak. Divide it with a 1:10 resistive divider. Feed this into a comparator.
The other input of the comparator receives the undivided signal from the transformer.
When this signal gets smaller than the 1/10th of the peak signal, the comparator changes state, and there you have your "let`s initiate the transition" signal.

From this signal, with RC delays and schmitt triggers, you can derive the precise timing for the turn on and turn off of the bricks.

In a real circuit, there are diode drops to compensate for, source followers to add servo boost etc. but this is just the basic idea.

In orcad, this has shown, that within reasonable tuning points, the IGBT`s are not forced to switch more than 45A even if the peak current is 1200A.

A while ago I tought if a comparator with a fixed reference to generate deadtime, but this would make deadtime sine function with primary current ant it would be huge if current is too small.

finn's idea looks much better: by comparing peak and momentary primary current he could always have IGBT's switched at wanted % of primary current.

I wouldn't use any RC delays, though: I would just invert the output of comparator and use falling edge to initiate turn-off.
Since turn-on is usually much faster and we already have deadtime I would just trigger it directly from zero-crss of primary current.

This way first IGBT would start to turn off just before the current reaches zero, be off and zero and at that point another IGBT would start turning on.

This would prevent shoot trough and make us able to finely tune the coil for zero current switching.

Finn Hammer

Thu Apr 12 2007, 06:34PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Steve Conner wrote ...

Unless nobody cares,

Steve,

I, for one, care.
The stuff that leaves your hand always has that special "hard to blow up" quality. this has enabled me to run 6 coils equipped with your controllers, without failure.
Only problem is, you put so many FF`s on the board I`m unable to understand how they work.
YouÂ´re the best, Steve, we all know it.
So what if you play a little guitar once in a while, that`s just terrific!.

Of course I wish I could persuade someone to build my zero current predicter, cause I`m probably not skilled enough to do it myself.

Cheers, Finn Hammer

Justin H.

Fri Apr 13 2007, 04:45AM

Registered Member #641 Joined: Fri Apr 13 2007, 02:45AM
Location:
Posts: 3

Hi everyone. I haven't posted here in at least a couple of years. It's amazing to see how far everything has come. The last time I was here, triggered spark gaps were being developed as replacements for RSGâ€™s in conventional coils (one of the many great contributions from Terry F.). Not too long after that, things got really high power with semiconductor, then things went pulsed-modeâ€¦(OLTC) and now look around. An SMT DRSSTC was developed not too long ago (Jim from Austin, TX) and there are now more than a handful of DRSSTCâ€™s with truly impressive spark-lengths.

Iâ€™ve stopped by to congratulate my friend on his recent accomplishments. As many of you know, such milestones come with blood, sweat, tears, burns, cussing, explosions, dollars, and a never-ending train of decisions one after another after another. Itâ€™s not an easy path to take, but Aron has built his life around it. 18 foot arcs coming from a box that looks like it came out of a Stealth Fighter may be impressive. The hundreds of custom circuits that Aron has developed in the last year may be impressive. But Iâ€™ll tell you whatâ€™s most.

Itâ€™s that heâ€™s been able to pull it off, whilst keeping up with lifeâ€™s most important priorities.

Cheers to you man. The above is more of an accomplishment than anything. Youâ€™ve had some shitty times and some bad-ass times, but youâ€™re making forward progress in your projects and in your life. Keep not-worrying about the near-term setbacks. Like Iâ€™ve said before, if you keep this drive and motivation up you wonâ€™t have a choice. Some day you arenâ€™t going to have to worry about how much the latest IGBTâ€™s cost, because your soldering-iron-sponge-wetting facility alone is going to make you enough dough.

Congratulations on the 18 feet man, thatâ€™s one hell of an arc. When Iâ€™m up there next, I want to see it go straight up into a big purple streamer (And then I want to go out and climb at New River!!).

And cheers to all of you here, pushing the limits of the unknown, and often doing it single-handedly on a personal budget. Itâ€™s a great feeling, but not better thanâ€¦

Wait, weâ€™re Engineers.

- Justin

Steve Conner

Fri Apr 13 2007, 09:51AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Cool! It's great to see you back on the forum, Justin. Back in the day, Justin & Aron were the Penn & Teller of high voltage

I think what Aron has achieved is pretty awesome, and I send my congratulations too. Does he now have the world's biggest DRSSTC? I think Steve Ward managed 14ft, so it's a question of whether Chris Hooper has beaten Aron's 18ft, and I don't believe he has. Whatever, you have all stomped all over the 6'9" I got from my OLTC2

Finn: My PLL driver is complicated, but I believe every part of it has a purpose. I think it is the best possible way to drive a DRSSTC. Unfortunately I'm too busy to teach people how it works and pimp it to develop a user base. Running that kind of free technical support eats up a lot of time and energy that I would rather spend on my day job, or actually working on projects, or indeed playing a little guitar now and again :P

So, I'm happy for Steve Ward's driver to be the "industry" standard, as it saves me a lot of effort and hassles. In fact, if you include the odds of the average hobbyist being able to build it and get it to work, Steve's designs win over mine on account of being simpler. When noobs e-mail me asking for help with their *SSTCs, I tell them to go look at Steve's site. :P

As for building your zero current predictor, I submit that I already built it, just using a PLL. So instead of predicting the next current zero by linear extrapolation from the quarter-cycle before it, it predicts the next zero from an exponentially weighted average of when all the previous zeros were, combined with an initial guess. PLLs are usually analysed in the phase or frequency domain, but if you think about my circuit in the time domain, that is what it does.

If Steve Ward's driver were modified with a high-pass filter like I suggested, it would predict the next current zero using the derivative of the current waveform (assuming it to be sinusoidal)

I don't know the relative merits of these three approaches. All I know is that the PLL made a whole lot of other design aspects easier too, so once I started on it, I was hooked.

Justin H.

Fri Apr 13 2007, 07:33PM

Registered Member #641 Joined: Fri Apr 13 2007, 02:45AM
Location:
Posts: 3

Thanks Steve. I'm glad to see that you personally are still rockin & rollin with this stuff, and it's great to be in comm's with everyone again. I'll probably be posting here periodically and lurking around.

hvguy

Thu Apr 19 2007, 07:05AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Justin; thanks for the post man. Hopefully this current trend continues and I can spend more time/money developing cool HV things AND climbing. It couldn't get much better than that...

Ok, I have been researching thisâ€¦ a lot.

Steve C., I like your PLL driver and did get it to work on the bench with relative ease. My only problem with it is its relative complexity. Not so much parts count; that can be slimed down, but in understanding. Since this was, by far, not the first time I have built a PLL it came fairly easily but Iâ€™m not sure about your average noob. That being said it is technically superior to anything else out there and you did a great job developing and documenting it.

Since I decided to stick with the logic based driver I needed to optimize it to drive these bricks and my prototype parallel IGBT bridge (more on that later) with the absolute minimum switching delay. Steve W. has done a lot here but I had some ideas that needed to be tested.

First I measured the delay time of driver; this turned out to be ~150ns from CT to gate. Not bad. Next was the total turn on time of the bridge, almost exactly 1uSâ€¦ Ouch. This basically means the bridge, witch CM1000s, will hard switch more than 1KA at 70KHz and 2.4KApk (Apk*sin(F*T*360) or just look at the scope shot). The fact is IGBTs this large should be driven with the PLL. Period. There is just too much delay inherent to large IGBTs to make the logic based drivers practical; unless you donâ€™t care/have big enough parts to handle the current.

So how do we fix this problem? Well, using the PLL would work but my theory, which seems to have ever increasing validity, is that the parts will work best and last the longest (i.e. more than ~2min) when run in their SOA. This means the voltage spikes must be controlled or eliminated and current must be kept below the manufactures Icm. In order to meet these requirements I have started testing a bridge based on four parallel HGTP20N60A4s, with external freewheeling diodes, instead of working with the slower bricks. The rise, fall, and delay times on these parts are mind blowingly fast. While my current off-line gate driver (OLGD, woohoo another acronym) is fast it is not <50nS fast and is not really the best circuit for driving the only ~16nF total Cge of my parallel bridge.

At first I tried the good-ol UCC and four secondary GDT. This worked (about 200ns) but did not give the blistering fast gate waveforms I was after. At this point I though about going discrete but could not get past the idea of bouncing all those parts around at ~400V (only 600V IGBTs here). Since we all know I like GDTs I decided to try using small, very low leakage, GDTs at each gate. Each GDT would be driven by a full bridge of UCC drivers. This allows me to keep the HV well isolated from everything while keeping things as simple as possible. Hereâ€™s the schematic

This circuit works very well. Total delay at the gate is 160ns with less than 50ns rise time. Here are a few shots of the gate waveforms

Obviously you want to keep the distance between the parts as small as possible. I will probably build these drivers as surface mount boards that mount vertically right next to the IGBT stack. Here are some shots of the total 300ns delay from signal source to bridge output

So what does all this mean? A bridge of parallel parts is far superior in size, cost, and speed; provided youâ€™re using 600V parts. At the 1200V level things become a little less clear. I think the next step for driving the 16â€ will be a divided tank cap and two or three bridges, perhaps two CM1000 bridges or three CM600 bridges. Either way I think it will use a combination of my new gate driver and a PLL. More on this soon (I hope). Far as small coils (<10â€™ arc, 400V buss) go I would highly recommend the parallel bridge over bricks. Just look at the datasheets; comparing bricks to a few parallel smaller parts is like comparing a 747 to an F16.

JimG

Thu Apr 19 2007, 08:30PM

Registered Member #52 Joined: Thu Feb 09 2006, 04:22AM
Location: Austin TX
Posts: 57

Wow, Justin called me out by name.

I've been following this thread and Conner's Odin thread because I've been working on building a large DRSSTC for a few years now. I've built both the PLL and the feedback circuits and various variants of them in the past.

The nice thing about direct feedback is that the circuit is simple and there's no need to make adjustments to the circuit based on the size of the coil. If something goes wrong you don't need to know intimate knowledge about how a PLL circuit works to debug it.

What I'm intrigued with is how simple Aron's gate driver is.

hvguy

Thu Apr 26 2007, 08:14AM

Registered Member #289 Joined: Mon Mar 06 2006, 10:45AM
Location: Conroe, TX
Posts: 154

Thanks Jim. I'm a big fan of Occam's Razor but this coil is starting to push the limits of "simple". My goal now is to keep it simple "for it's size" instead.

I am currently in the process of evaluating the PLL driver for use in the 16â€ coil system and Iâ€™ve got to say it has yet to give me any problems. Steveâ€™s circuit has clearly been well thought out and tested. I have yet to see how well it makes arcs but the bench testing looks good. Although Steve, there seems to be a slight phase shift from the beginning of the burst to the end. I noticed this on your site as well; do you have any thoughts on this? Also I like the hi/lo lights; nice touch

Oh ya, there's a typo on your schematic, VCOIN is labeled as pin 8 and it should be pin 9â€¦

After tossing some ideas around far as gate drivers and over current schemes go I have settled on a preliminary design for a triple bridge of CM600s. I will post schematics once I draw them; probably in the next week.

On a different note this thing has been sucking time and money like a black hole. Iâ€™ve only killed one IGBT but all the small parts and prototypes I have gone though are starting to add up; not to mention the cost of three new full bridges and associated electronics. Then thereâ€™s the +400hrs of time to build and test it all... This coil better arc 100â€™ at 150% duty cycle when Iâ€™m done

Steve Conner

Thu Apr 26 2007, 12:32PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Hi Aron,

Wow, somebody built my driver! I'm honoured

About the phase shift thing: As far as I could tell, no matter how you set the controls, you will always have a slight phase error temporarily at the beginning of the burst. This phase error seems to be a necessary part of the operation of a dual resonant system.

What I think happens is that the ideal drive signal, if you wanted perfect zero current switching, would start out halfway between the two pole frequencies, and then bifurcate to one or the other. But if you started exactly halfway, both modes would develop equally and the PLL wouldn't know which to lock to.

So you have to start from a little above or below the ideal trajectory depending on which pole you want to end up on, and this is what causes the phase error. I always got better results driving the higher of the two pole frequencies.

A word of warning: My mental picture of these things is based on a coil that had tight coupling (so two widely separated and easily distinguishable modes) and a high operating frequency (so a lot of cycles per burst). This gave the system plenty of time to settle to a steady state with only one mode excited.

Steve Ward always built coils with looser coupling and shorter burst length. Under those conditions, there may never be a time in the burst when you can say that you are driving just one mode. I'm not too sure how this affects the theory my circuit is based on.

Also, I've not gone through the math, but I think the Steve Ward style feedback driver can produce a drive waveform that actually contains both mode frequencies at once, whereas mine is limited to one frequency at a time. I don't know what the implication of that is, though.

Steve Ward

Thu Apr 26 2007, 10:03PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

Sooo... who will be the first to implement the brains of a DRSSTC driver on a micro-controller? yeah, im probably that foolish to try it

. The PLL is cool, but its so many parts and wiring... using a single micro controller as the brains seems "better" but on the other hand, PLLs and D-flip flops dont accidentally lock up :P. Ive only begun to brainstorm such a controller... so ive got some ideas, but havent actually started programming it. Im guessing my 16MHz device will be limited to low frequency coils (maybe <100khz) in order to properly phase shift it.

I havent blown up any silicon in awhile... i need my fix!

Of course, i need to finish my uC based interrupter first (just about done!).

first previous 8 9 10 11 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob