DRSSTC log book
This will be a place for me to post about whatever is on my mind at the moment. Any information presented here is not necessarily correct, so reader beware, I definitely make mistakes as I'm learning! I've also come to realize that I really should be writing down these thoughts. I am all for sharing knowledge, so I will just simply put this online for people to see and maybe learn a thing or two from. All for the advancement of the dual resonant solid state Tesla coil!
2/11/05. I've been working on the DRSSTC for nearly a year now, so it seems like a good idea to sorta give a brief history of what I have done so far. It first started out as a normal SSTC, with an untuned primary, driving the secondary. I used an antenna to feed signal to a comparator, this generated the gate drive waveforms to control the H-bridge. I used a low impedance primary of only 3 turns (few uH) and pulsed the gate driver on and off to keep the duty cycle low. In this manner I was only able to generate maybe 2' sparks. Just slightly better than a normal SSTC, but at perhaps less power input... This I claimed to be my ISSTC (interrupted SSTC). Jimmy Hynes had been doing his DRSSTC thing a little before I started my ISSTC.
I started talking with him over the internet and he convinced me to try to make my ISSTC into a dual resonant system. And so I did, and everything started to change from that point on. I learned exactly why IGBTs were so popular for pulsed applications. MOSFETs have an ON resistance of usually 50-100mOhms at best, this means that they dissipate power as I*I*R=W. IGBTs on the other hand act more like a diode, they have a voltage drop that stays relatively small even with huge currents flowing through them, so power dissipated is given by Vdrop*I=W. Here the Vdrop is maybe 3V or so. So a MOSFET will typically be dissipating about 10 times the power as an IGBT in this sort of situation. So I sampled the best IGBT I could find at Fairchild (and then so did many other people). So the new IGBTs in combination with a tank capacitor had given my system a great boost in performance. I believe I started to get up near 3' or so soon after that. But then I started to blow up IGBTs when the arcs hit my antenna that I was using to get feedback. It was time to switch over to the current transformer! A small CT on the ground return of the secondary coil served well as a source of feedback.
Experimentation continued more and more, the pile of dead silicon was growing. Soon I was getting 4' sparks from my 19" tall coil. Eventually that 19" secondary maxed out at about 55". I then proceeded to build a few other DRSSTCs after the success of that coil. I next built the rather large DRSSTC-2, which eventually gave about 11' sparks out in the back yard. Never got a real power measurement on it, but guessed it to be around 4500W since I was running on 20A fuses. That coil being impractically large to run it as often as I'd like prompted the building of the DRSSTC-3 (which now serves as a test bed for new ideas). I then built the DRSSTC-.5 as a very small little demo coil that requires almost no setup time and can be run just about anywhere. Sort of a toy really, but it still makes fantastic sparks.
So now what am I doing? Well, I've gotten the inclination to improve upon my designs. While they seemed to all work pretty well, every so often I would blow a set of IGBTs for no good reason. The culprit? Not sure, only thing I can come up with is that the IGBTs are latching up and then going nuclear! But why are the IGBT latching up? Well, DRSSTCs should normally operate in a ZCS (zero current switching) condition. That is, the current in the primary circuit is sinusoidal when the drive voltage is at resonance with the LC. This makes it quite simple to get very close to ideal ZCS. First possible flaw here is that I am using secondary current to derive my drive signals. Who says the secondary is doing the same thing as the primary current? It isn't necessarily. Also, there is an interesting problem that comes up when the secondary arcs to a grounded object. When a ground strike occurs, the secondary appears to jump to another harmonic, usually 3X or 5X the Fres of the coil. This causes all sorts of bad switching transitions that should not occur. I believe this is why I've lost to many IGBTs in the past. So the idea now (partially inspired by Steve Conner and Terry Fritz) is to use primary current feedback to derive the drive signals. This should in theory yield almost perfect ZCS conditions!
Some important considerations when using primary current feedback: Firstly, you want to minimize the delays in your CT. That means it really needs to be loaded down well (notice that I have not been properly loading down my CTs in my previous work!!). A simple resistor works perfectly to load a CT, but not when we want to derive a drive waveform from it. In this case, we can usually use some sort of scheme involving diodes to clamp the output of the CT to a specified voltage. This yields a nice square wave from which we can work with. Most recently I have been using some 15V zeners along with some schottkey diodes (the zeners have a slooooow recovery time, so the schottkeys are there to fix that). This clamps the output of my CT to 15V or so, also note that its a balanced load in that I clamp both halves of the waveform equally (my old circuit doesn't really do that). Now I have a +/-15V square wave to feed off of, great. Second thing of importance with primary feedback is that you must use an over current detection circuit! If the secondary somehow becomes transparent to the primary, the primary will ring up without limit. There must be a circuit watching the primary current, that will disable the drivers if the current exceeds your maximum setting. When the secondary arcs to ground, the primary current will ring up excessively high as I mentioned. I will eventually post my schematics as to how I'm watching the current and disabling my driver.
Problems so far: Right now I'm finding that there is a significant phase shift in my CTs. Loading the CTs to a lower voltage seems to help the phase shift approach 0, but I'm not there yet. A few things I've learned. Don't cascade CTs! It just makes any phase shifts 2X as bad (at least). Its best to just spend the time and wind all those turns of wire onto one ferrite core instead of cheating with 2 1:30 CTs in series. I also have a problem with my JK flip flop and the way it is currently configured. In order to make the output change state (enabling the drivers) I have to feed the interrupter signal into the feedback input, so that the clock on the FF changes. This means that the feedback source impedance must be high, because the interrupter source impedance is high. It just really complicates and restricts things... in general, I don't like it and am working to change that whole setup.
Sent a letter to the TCML about what's going on with my DRSSTC-3 recently:
Yesterday I managed to
solve all of the problems that I was facing with my DRSSTC-3 and using primary
current feedback. Today I'm happy to claim longer sparks :-)
For anyone who is basing a DRSSTC design off of my designs, you might want to check out the updated schematic for my DRSSTC-3:
OCD stands for over current detection... not obsessive compulsive disorder ;-).
Anyway, I borrowed some things from our very own DRSSTC newbie, Terry Fritz ;-) I liked the idea of clamping the output of the CT with zeners and using some schottkeys to take care of the slow recovery time. I modified the flip flop circuit so that it doesn't have any issues with needing noise in the feedback input to start oscillating. You can also see the OCD in that schematic, it works beautifully. I can set it anywhere from around 200A to 600A, and I do hope to eventually be using near 600A ;-).
I found something interesting with the tuning of this coil now. Originally the primary was tuned somewhere in the middle of the 2 modes. This worked well with very high coupling as normal, but I did see some limit to my spark length (about 24" max). I really liked the idea of tuning to the lower pole, because when the coil starts making streamers nearly 3X its length, its going to detune considerably! So I started tuning the primary lower. Instantly I noticed the primary current had a much nicer linear ring up, instead of the choppy looking current ring up when tuned in the middle. I had to reduce my coupling to about .15-.2 because arcs began racing up the side of the secondary, and jumping to the primary nearby. When tuned lower, it took more input voltage to get it to produce real streamers, but once streamers start forming, the result is like an explosion of streamers! Looking at the primary current, its forming a notch at the end of the burst (current rises and then returns back to 0). If I turned the
power even higher, a second "burst" began to form after the notch. The notch occurs at about 18 cycles or so... and I'm not sure what causes it, but I think its a sign that the coils need to be tuned better. So I slapped on a turn of 12 awg at the base of my primary to get more tuning room. Now it takes even higher input voltage (about 60-70% input) but the resulting sparks are even longer! Interestingly, primary current increases with voltage input, until long streamers are formed, then the primary current doesn't increase any more at all! Right now I'm running about 420A. But even still, as I'm getting about 30" sparks, the notch is occurring again at the end of the burst... need more primary L! Its easy to see the improvement that adding more primary inductance has. I'm tuned at least 1.5 turns lower than before (the coil only had 5 turns to start with!),
obviously these streamers really do detune things a lot.
One other benefit I see to running at the lower pole is that you are reducing switching losses. You now have fewer RF cycles per burst length, and also, any delays in the gate driver become less significant as each half-cycle is now longer. I think my primary circuit is running at about 170khz, the secondary Fr is something like 220khz. Hey, but it works great so far!
I recently started tuning the DRSSTC-1 coil using the 6.5"x22" (26awg) secondary and primary feedback. Firstly, the current is damn high, I managed about 1000A so far through the 40N60s but this is a bit unnerving. Initially I was able to get about 4' sparks at a little over 50% input voltage and 750A in the primary running about 12 cycles (120uS). I then did 3 tests with varying coupling between the coils. At K=.21 it took 1000A at 1000W to get 4'. At K=.24 things looked a bit better, I recall 800+A and about 900W for 4'. At K=.29 I got the best results with 700-750Apk and about 800W for 4'. The only problem with the higher coupling is racing sparks, but making everything very round with no sharp edges has really helped there. I think I may make a new strike ring with some larger diameter copper tube (I have a bit of 5/8" tube left over from the DRSSTC-2). I would really like to see 6' from this coil (66" would be 3X the coil length, and you can usually get a bit more than that).
At the moment I am considering winding a new secondary with 30awg wire. It would have an Fr of about 83khz which could help quite a bit with switching losses since driver delays are less significant as the operating frequency lowers. The coil is running at a bit less than 100khz at the moment. It would also be a good experiment since both coils would be the same size physically.
I keep running into problems with my newer interrupter. When I first built it, it was just a circuit board, not shielded or anything, and it actually worked perfectly fine like this. Then I got the idea to put it inside a box, and that's when I started getting problems. During ground strikes, something is causing an extra long burst (I can see it on the scope). The burst looks to be maybe 2X as long as it should be, and luckily the OCD is catching these events. I still don't like when things are unstable, so now I'm thinking I should etch good PCB for the thing and likely ditch the "burst" mode where I send out several pulses with a long rest in between. This just seemed to cause random failures and also caused racing sparks sometimes, though the effects it created were really awesome I just didn't use it much. I'm sure I will figure out some sort of solution, but I'm liking the idea of a making a PCB for the thing, and probably not even worrying about an enclosure. Its funny how attempts at shielding often backfire :-p.
I was strongly suspecting that the tank impedance was just too low and perhaps it wasn't optimal for producing the best sparks from my setup. So I made a new primary, 11.5" diameter with 7 turns of .25" tube. I then disconnected half of my MMC, for a .15uF cap. When tuned to 6.5 turns (about 20% lower than resonant) I was getting 5' sparks at about 500A peak in the primary. I was pretty happy about this. It also suggests that rate of ring up is dependant on the L and C values in the tank circuit, half the C gives half the current for the same power and spark lengths. I think that at some point (when the cap gets smaller and smaller) you will not get the desired power out of the coil. I had to run the coil at about 90% input to get 5'. The pulse length was varied from 100-150uS in both cases. It might be possible to get considerably longer sparks with a longer burst length, but I fear efficiency will plummet.
Right now I'm thinking I want more current, perhaps 750A would be nice. The next thing to try will be a new secondary with 30awg wire and an Fr near 85khz or so. This will need about 7 turns on the primary and a .3uF cap. I think this new primary configuration will give me my desired primary current. The lower Fr should be even easier on the IGBTs. I will just have to see what happens when I actually wind the new coil (waiting on the wire right now).
My OCD circuit continues to work wonderfully :-).
Lots of new things since the last time I wrote. First order of business, the magnifier tests. My inclination towards the magnifier was basically to move the sparks away from the primary coil. After all was said and done, I found I really wasn't getting much benefit of running in a magnifier configuration. It seemed to simply result in a larger, more complex system. The small magnifier seemed to work well and allowed for 42" sparks vs. 36" for a 2 coil system, but the increase in physical size easily explains the spark length increase. I then tested out the large magnifier and was somewhat disappointed. Perhaps I was trying to make sparks too long for the systems size, but I was plagued with racing sparks on the secondary as well as poor efficiency compared to the 2 coil system. For now, the magnifier research is dormant, I might look at it again in the future, but its unlikely.
One newer theory is that there is a power density that limits the spark length for a given coil size. Basically, it seems that sparks generally wont exceed 3X the secondary winding length, at least it wont operate efficiently beyond that limit. My record so far is 80" from a 22" long secondary on my DRSSTC-1. 72" sparks on the 22" coil are much more common though.
Recently, Terry Fritz has been working on his Scan Tesla program which basically performs thousands of pspice simulations varying parameters until it finds an optimum setup for either 1) max secondary voltage or 2) max streamer energy. I think 2 is more important. Anyway, his program suggested that I lower the coupling on my DRSSTC-1 system to about K=.18. What I found was somewhat surprising compared with the old configuration using a K=.24. Both setups could produce 72" streamers, but with K=.18 I only required 6 cycles (rather than 9 at K=.24) and this has reduced the input power dramatically from about 1800W to 1200W (based on bang energy values... the real input power is somewhat higher due to the lousy load that charging giant filter capacitors presents). The peak current was a little higher at 700A now. On the plus side, the primary RMS current is lower (less cap heating). I should mention I upgraded the MMC to 450nF (that's 5 strings of 2, .15uf 2kv CDE caps). I can also operate now without a breakout point without flashover between primary and secondary. The random streamer breakout is quite exciting :-).
New problem: Its summer, and everyone has their air
conditioning running. This lowers the line voltage and seems to make
voltage sagging even worse when I try to run my DRSSTC. I find that using
a 120V line (by either using a 140V variac and a voltage doubler, or a 280VAC
output from the variac and a full-wave rectifier, both cases yielding 400VDC
optimally) I cant get 6 foot streamers now! I can only manage maybe 5
feet. When I plug into the 240V line (280VAC out of the variac into the
FWR) I can achieve the 6 foot sparks. What seems to be happening is that
the very low impedance load of the filter capacitors (note, they only charge at
the PEAK of the AC waveform, when the diodes are forward biased) is clipping off
the tops of the peaks. The result is that I only get maybe 350VDC (maybe
less!) rather than 400VDC. That's a difference of a few joules of bang
energy (about 7J vs. 10J!!). For now I can just run the coil from a 240V
line, but I have some things I plan on trying. First would be to reduce
the buss capacitance to 4700uF instead of 11000uF. While the smaller cap
stores less energy, it should hopefully charge to a higher voltage. I'm
not sure if this will get my sparks back up to length or not, it should be an
interesting test. If that doesn't work, then the next plan is to make a
switching power supply. This could potentially charge to 400VDC even when
the line voltage is sagging terribly. I'm looking at boost converters at
the moment. I can also design it so that it pulls current for nearly the
whole AC cycle instead of only at the peaks, this would reduce the RMS current
drawn from the line, improving my apparent efficiency.