Testing the TK Tar Baby

[TinselKoala]

OK, I went through the manual calculation using one cycle from the above scopeshot. I got a mean power of negative half a Watt.
If we can get this up to negative One Watt.... then I'll send it off for the "One Watt Challenge" prize offered by somebody or other.

This is of course uncorrected for the inductive reactance of the CVR at 3.6MHz. The correction will lower the power figure but it won't change the sign.

[TinselKoala]

I'm a little surprised that the polarity of the LEDs can't be reversed and have them still light up. Since the circuit works and oscillates with just a 100R in there, the diode action isn't required, I think...

Here's a funny: If I put the LED in backwards it doesn't light and of course there are no oscs. But if I then "short" across the reversed LED with another 2.2uH inductor, the LED lights up! And the oscs return but with an extra, large, positive going peak in the CVR trace.
If I have the LED in the "right way around" as shown in the diagram and glowing well,, shorting across it with the inductor kills the light and the oscs.

Well, it turns out I was wrong about that very last part. The LED dies... but the oscillations don't. They smooth a bit and look even more strongly negative. Removing the LED load entirely and just dropping in one of those 2.2 uH inductors makes the best result of all, I think: nice smooth waveform with lots of negative math.

I made a video and it's uploading now. I'll post the link when it's ready. Meanwhile here's this, showing the waveform with the load LED reversed and paralleled with the 2.2uH inductor.

[TinselKoala]

Curiously strong:

http://www.youtube.com/watch?v=--JF12FH2Zc

[TinselKoala]

@ .99:

Altoid seems to run as well or perhaps a bit better.... with the Zener completely removed ??!?

[poynt99]

Hi TK,

With the simulation, removing the LED and adding 2.2uH more inductance does increase the VBAT amplitude slightly, but other than that, there is little difference.

However, removing the zener makes quite a difference; the VBAT went from ~60V to ~200V swing! This ought to make an even better negative mean power (NMP) computation. The Fo decreases quite a bit too.

Obviously there is more going on in terms of resonances in the real build compared to the simulation, as my traces are relatively "clean", and my "VBAT" amplitude is much higher.

Nonetheless, it's nice to see it running, and you making "improvements" on the NMP computation.