Pulse-width-modulation and the Baxandall Class-D oscillator

I thought I'd have a play.....because I am a bit of a pedant when it comes to 'pretty' circuit diagrams and pictures of them.

Not pedantic enough though....

Errrr, it sort of oscillates but. I don't know.

Anyway, have fun

DNA

Why the counter? It's not for a delayed-sweep scope equivalent. If it's for a PWM time delay, couldn't you use a time delay, and manually adjust the delay to meet your needs? Crude, but...

Bill, I enjoyed a 2003 12-post thread titled, "Peter Baxandall's class-D oscillator revisited." Message-ID:

But I wasn't able to find the March 2006 thread you mentioned.

It clear to me a PWM version of the Baxandall oscillator would be useful, at least for the small crowd who needs a wide output-power range. For example, I spent a few months designing a 300kHz 500W phase-shift PWM driver for my 10kV transformer, featuring PWM and power-supply-voltage variation together, so it could precisely operate over a 1000:1 voltage range, from 10V TO 1kV. Note, the feedback loop included a square-root converter to account for the double parallel feedback terms. I doubt I could have achieved the wide voltage range without the PWM modulation. A Baxandall power oscillator might have greatly simplified my circuit, eliminating the ucc3895 phase-shift controller and hip4081A H-bridge driver, but it would have needed PWM as well as power-voltage modulation.

I would have needed to drive the Baxandall resonant circuit with an external frequency reference and tune the load, as opposed to letting the switching follow the resonance, but that's OK.

It would let me realise a "magic sinewave" approximation to the perfect drive waveform (at least for high-Q tank circuits) plus non-overlapping drives for M3 and M4. The nearest thing I can find to a time delay in LTSpice is a lossless transmission line, which could be made to work, but is worryingly unrealistic.

You seem to have managed to set up non-overlapping switching between S1 and S2 - I'm not enough of a Spice guru to understand how - which means that you need the catching diode DSTOMP to make the circuit work. The third harmonic spike is worse (37dB down as opposed to 40dB down) than it is on my "pretty" educational model of the classic Baxanadall circuit. I've not yet managed to build a real Baxandall oscillator that needed that catching diode.

Anything that did something was more by luck than design.... I threw DSTOMP in there because node VCT was swinging off to -millions of volts. Of course the thing that prevents that in 'real' life is the body-source diodes of the mosfets, being silly I left them out.

I'm not enough of anything to initially know why it behaves the way it does but I like to keep things simple so I can selectively prod parts and scratch my head over what things might be doing.....

DNA

That pretty much duplicates my classic Baxandall circuit, including the initial burst of "squegging". I had to run the simulation for 10msec and ignore the first 5msec to get a clean waveform for the DFT routine, and it gives pretty much the same thrid harmonic content that I got.

I personally dislike these self oscillating things......

DNA

It looks like your "squegging" is the presence of two resonances in the circuit, maybe.

The fundamental operating frequency might want to be set by the transformer primary magnetising inductance and the resonant capacitance. However the input inductor also resonates with it and the only source of damping is the output load... sort of...

Try....

I made LIN 10mH and set the two load resistors for critical damping assuming that the capacitors appear in parallel referred to the input inductor.

It's still a bit woofty, like it's not going to weld heads together at fifty paces.

DNA

OK.... I changed that a bit because it looked overdamped so maybe I was thinking about something wrong.

DNA

The squegging doesn't seem to be a problem in practice - it just limits LIN. Make LIN too big, and the squegging never dies away.

I'm with you on the problems of self-oscillating circuits - my pulse-width-modulated circuit started out self-oscillating, but proved to oscillate in a whole lots of modes that I didn't like and couldn't get rid of.

Unfortunately, the Baxandall circuit does require zero-voltage switching, and if you don't let the thing self-oscillate, you need some kind of control loop to move the oscillation frequency to match the resonant frequency of the tank circuit (or adjust the resonant frequency of the tank circuit to match the desired resonant frequency - as in Win's application).

I'll still say the "squegging" is the input inductor resonating with the resonating capacitor and if you properly damp it with the load it gets sorted. Of course that's not a very good solution.

I was chasing some sort of deadtime inclusion type thing but I finally figured that this is a current fed coverter so having the switches on at the same time isn't so much of a problem and that leads me on to the next one....

Now we might be cooking. I've tagged a current mode control buck converter on the front end of things. Watch out, as I've been diddling I changed the transformer turns ratio.

It seems that there are two limiting conditions, maybe. The first is you shouldn't operate with a load close to the resonant impedance of the tank (above it but not too close to it). The second is you shouldn't try to develop a reflected voltage across the load that is twice the input voltage.

Anyway..... there's something to have a play with. It looks like the input control stomps on the "squegging". You can regulate the output using PWM and... I dunno.... generally poke about and see what happens.

Cheers

DNA

Baxandall's value for the feed inductor was:

Lfeed = Rload(referred back across coll-coll)/2w.

Notice that with Rload referred back across the collector-collector then there is a relationship between Lpri (collector-collector), Q, and Lfeed.

Yarrow is supposed to have based his value for Lfeed on O.H Schade's L-C power supply calculations.

Lmin = Vcc/3.w.I and then he used 5*Lmin.

I've used Baxandall's, but not Yarrow's.

Here you go..... LET'S SHUNT SOME POWER......

How do you fancy 1KW into 50K........?

Took a bit of fiddling because I haven't figured it out yet but I'm close. Mind you it's probably a bit tra...laa...laa.

DNA

Just when things start to get interesting, my wife drags me off to Paris - we won't be back until next Monday. Sod's Law working at full strength.

Thanks Tony. That is interesting and important - I completely missed that when I read Baxandall's paper, back in 1968 (when I knew very little about electronics).

think harder.

the ideal switch makes a great comparator, R-Cs can easily add the delays you want. lots of different ways to do that. Not necessarily a good idea tou build real circuits like that, but SPICE dont care.

I have a spice model of a 3-phase regenerative rectifier using parke and clarke transforms, incorporating a behavioural model of the TMS320F2811

And I built a behavioural thermal model of my IGBTs too, which includes switching loss. took perhaps 30 minutes.

Cheers Terry

I just got my copy of Roddam yesterday. good, isnt it.

Cheers Terry

That's him. As mentioned in previous threads I was unable to obtain a copy of Baxandall's original paper so had to do with Thomas Roddam's description.

Now then, move on to page 191, where TR developes the voltage-driven series-resonant form into a half-section filter. Can you follow the final step, to figure 10.25, and add values to it?

Figure 10.25 is given below.

L1 C2 Squarewave voltage----((((((---+---||---+---+ | | | | ) \\ C1=== L2) /Rload | ) \\ | | | -----------------+--------+---+

L2 is the primary of the output transformer, with Rload referred to the primary.

On page 192, do you know what a "Boucherot Impedance Transforming Network" is?