Testing the TK Tar Baby

[TinselKoala]

Ha-ha !

I have achieved ClassE breakout, at about 27 volts input. At 33 volts and about 3 amps, there is a nice constant hissing blue corona, and I can draw whitehot 1/2 inch power arcs. The secondary is heating up, so I have to be careful... it's held together with hot glue!
The ClassE waveform seems fairly stable; I can vary the input voltage quite a bit before I have to retune the gate bias pot to maintain the nice clean drain spikes and the fair sinusoidal secondary field. If there's too much gate bias it gets very noisy and the mosfet heats fast, although the voltage output can still be high. If the bias is too low the oscillations increase in frequency and the resonance is lost.

This is as high as my "brute force" DC filtered supply will safely go. Next I'll try it with 36 volt battery pack, then 48 volts. Have to move it a bit further away from the bench first, though. My goal is nice spectacular corona, on batteries alone, for total portability. But even if I have to give it line voltage, rectified and filtered, that will still be OK... as long as the mosfet survives, that is.

I'm still trying to find the optimum price/durability ratio for the mosfet used here. The original design used 17n80, I've seen spectacular results from a slightly larger ClassE coil using IRF720, which is a TO-220 little powerhouse, I've used the IRF830a in this one and it works OK, even the IRFPG50s from Tar Baby work, sort of, but their Rdss is too high for good results. Obviously the transistor needs high drain-source voltage rating, low on-state resistance, fast response rates, and high current capacity. The IRFP450 is working well so far, I've only popped two while tuning (one with a nice explosion and power arc) and now that ClassE performance seems here to stay, they should last longer, unless the voltage swings too high as I increase the input power.

[Magluvin]

Basic FG anatomy showing where the 50 Ohm output resistor actually resides.

Ok, I see now. Thanks. So when the gen is not in control of the amp(1ohm) or just not powered up, we have 51ohms from output to ground?  Or if the gen is set for dc on pulse, or just ac with offset, and the gen sig goes to zero V, we have our 51ohm path again.

But that is not the issue. It is the fact that the path exists in the circuit where it is placed. And that changes the circuit quite a bit from the original idea, if one believes that the signal gen is place, is only injecting a signal and not affecting the circuit in any other way.

So those big dense bursts in the waveform are only when the gen is in a state of phase that causes the oscillation to happen. And the flat spots are when the gen is in the opposite phase that doesnt allow the oscillation to happen.

So that is why you guys can get the oscillation without the gen. You just supply the circuit with a permanent oscillation turn on, and just skip the off time.

So the gen is just turning the oscillation on and of for intervals, as the oscilation effect is much higher in freq than the gen is putting out.

So, no real need for the gen. ;]

I think I got it, to some degree.    Thanks 

Mags

[picowatt]

Ha-ha !

I have achieved ClassE breakout, at about 27 volts input. At 33 volts and about 3 amps, there is a nice constant hissing blue corona, and I can draw whitehot 1/2 inch power arcs. The secondary is heating up, so I have to be careful... it's held together with hot glue!
The ClassE waveform seems fairly stable; I can vary the input voltage quite a bit before I have to retune the gate bias pot to maintain the nice clean drain spikes and the fair sinusoidal secondary field. If there's too much gate bias it gets very noisy and the mosfet heats fast, although the voltage output can still be high. If the bias is too low the oscillations increase in frequency and the resonance is lost.

This is as high as my "brute force" DC filtered supply will safely go. Next I'll try it with 36 volt battery pack, then 48 volts. Have to move it a bit further away from the bench first, though. My goal is nice spectacular corona, on batteries alone, for total portability. But even if I have to give it line voltage, rectified and filtered, that will still be OK... as long as the mosfet survives, that is.

I'm still trying to find the optimum price/durability ratio for the mosfet used here. The original design used 17n80, I've seen spectacular results from a slightly larger ClassE coil using IRF720, which is a TO-220 little powerhouse, I've used the IRF830a in this one and it works OK, even the IRFPG50s from Tar Baby work, sort of, but their Rdss is too high for good results. Obviously the transistor needs high drain-source voltage rating, low on-state resistance, fast response rates, and high current capacity. The IRFP450 is working well so far, I've only popped two while tuning (one with a nice explosion and power arc) and now that ClassE performance seems here to stay, they should last longer, unless the voltage swings too high as I increase the input power.

TK,

Nice coil you have here.

Have you considered putting a gate driver in front of the MOSFET?  That circuit looks like it would switch the MOSFET fairly slow, hence more dissipation.  I would suspect that very little heat sink would be needed with your current setup if the MOSFET were being switched on and off in less than 50 nanoseconds, and there are plenty of 2-4 amp gate drivers out there that can easily switch 2nF in less than 50ns. 

When it comes to MOSFET switching speed, it is all about how much current you have available to charge/discharge the gate C.  Higher voltage and/or low RDSon types will typically have more C to charge/discharge.  With class E, even though you are aimimg to switch at maxV/minI and minV/maxI, you still want to switch as fast as possible to reduce losses in the MOSFET. 

PW 

[TinselKoala]

@Mags:
Yep, you've got it. The oscillations only happen when the FG is sending a negative voltage out its "positive" probe lead, what we are calling Gate LO. Take a look at the circuit schematic, and imagine the FG as a "negative" battery. That is, simply imagine it as a battery, applying its positive pole to the gates of Q2 and the negative pole to the gate of Q1. Now trace the current flow. Notice how the FG's "battery" is in series with the main battery. The voltage measured from the main battery positive to the FG's (or other bias source) negative is the sum of the two voltages provided.


Now... you may wonder about the ClassE coil I've been working with. Why have I been working with it, why am I talking about it here?

I've cut out the irrelevant components and noted the bias supply section in the following schematic. Notice anything interesting about the circuit? For example... why does it oscillate?

Now, as I'm showing with this ClassE sstc, when mosfet(s) are switched _properly_ for the purpose desired, and parasitic oscillations avoided and only the desired feedback oscillations are actually USED for something.... very spectacular results can be obtained. This ClassE sstc is _probably_ much more "overunity" already, than any of Ainslie's miserable little knockups IF it is measured in the same way, by undersampled spreadsheet integration and improperly filtered probes. I am melting copper wire, radiating power in the RF, lighting up high-voltage neon bulbs, burning flesh, making half-inch continuous power arcs... all with a single mosfet carrying under 50 Watts of input power.

And it's using nearly the identical circuit as the single mosfet NERD circuit, with a bit of deliberate "stray wiring" for oscillations and a lot of added capacitance for power. And of course... it's doing something with the field from the primary inductive load: it's using Faraday's Laws of INDUCTION and the phenomenon of voltage rise through standing wave resonance to produce a voltage amplification far in excess of what turns ratio alone would predict. Turns ratio is about 70, input voltage 30, transformer effect voltage 2100 volts, actual voltage reached by vrswr in excess of 6 kV, judging by corona and arc effects.

[TinselKoala]

@PW: Yes, you are right and I agree with you totally, and I've got a more sophisticated design with a mosfet half-bridge, driven by gate driver chips and triggered by RF pickup in a CMOS inverter, that will likely blow the socks off this one, with true SGTC style long lightning bolts from the top capacity.

This coil is actually being used for a different purpose, which you might be seeing now. it's never going to make those long lightning bolt sparks, but when I have it working at full power with the straight, unconditioned gate pulses, however slow they are, it will be proving a point that I've been trying to drive home for years now, I think. And it will make a good audio-modulated demonstrator as well.

Don't worry... there are more and better SSTCs in my future. I have a nine-pound spool of that #27 magnet wire and I only used 1/4 pound for this coil, and I've been constructing a coil winder in my spare time. When I go to the trouble of using mosfet drivers and I expect the thing to run for a while, making lightning bolts, I'll use a full H-bridge and a PLL circuit, probably, with driver chips, of course. I have a few Intersil H-bridge driver chips right here, in fact, now that I think of it, HIP4080AIP. Thanks for reminding me, I had forgotten those were in there. I got them as _free samples_ from Intersil about ten years ago.