If I had a Baxandall class D resonant oscillator, would it be possible by modulating the current input to the drive circuit to produce a rounded off triangle like waveform?
Bill Sloman's excellent work
formatting link
found that the odd harmonic distortion was caused mainly by the AC ripple current flowing through the source inductor (and hence the driving windings). This presumably causes a perturbation in the dB/dt of the flux which modifies the output waveform.
So would it be possible (in theory) by controlling this drive current more accurately to produce a rounded off triangle waveform without losing all the efficiencies and advantages of a resonant class D oscillator?
It would be messy. In theory, to create a triangular waveform, you want to add the odd harmonics of the fundamental, with each harmonic added at an amplitude that is related to the amplitude of the fundamental in proportion to the inverse of the square of the harmonic number - the third harmonic at one nineth of the fundamental, and the fifth harmonic at one 25th (4%) would seem to be as much as you'd need.
So, three centre-tapped tank circuits, tuned to be resonant at the fundamental, the third harmonic and the fifth harmonic. Then three separate feed inductors, each going from the same voltage rail to a different centre-tap, and three pairs of MOS-FET switching transistors to drive the three separate tank circuits.
Then a 4046 running at - say - thirty times the fundamental frequency, divided by six to drive the fifth harmonic tank, by ten to drive third harmonic tank and by thirty to drive the fundamental tank, with a second divide by thirty output in quadrature with the first that you can phase lock to the output from the fundamental tank.
This would give you three sychronised sine waves; put a 225 turn floating coil on the fundamental tank circuit, a 25 turn floating coil on the third harmonic tank circuit and a 9 turn floating coil on the
5th harmonic tank circuit, and connect the three coils in series ands you should be able to end up with a not-too-round triangular wave.
The feed inductors could probably have quite a lot more inductance than the inductance of the tank circuits - the original Baxandall class-D oscillator built with bipolar transistor switches "squegs" when the feed inductor is too big, but oscillators driven by MOS-FETs don't seem to have this problem.
What's interesting is that, once it's all going, the power supply can be cranked down to zero and you can make the triangle forever, for free, since the ideal circuit is lossless. The slopes are technically segments of sine waves, not linear bits, so there will be some small curvature, less as L gets bigger. Given a real inductor, simple tweaks could make the slopes straight.
Blimey, I didn't consider that solution! Thanks for this. The tanks I assume would be close to resonance but may be a little out due to tolerances and drift, which I guess will add a bit of cross-over distortion, but that might get filtered out eventually.
Anyway, I was thinking more on the lines of modulating the current through the single source inductor. Maybe by reducing the source inductor to a much lower value, and PWMing it such that the resulting dB/dt in the core generates a triangle wave. The reason I say that is because your work seems to suggest the ripple through this source inductor (due to the centre tap voltage rising and falling) actually adds harmonic distortion to the output waveform, and one of the solutions was to try to remove it by PWMing.
Also, this source inductor is quite a large component and suffers from I2R losses, so couldn't you make it look much bigger to the circuit by tracking a proportion of the centre tap voltage and PWMing a much smaller source inductor with it? The result from that I think would be to effectively put a much smaller ripple voltage across the inductor, and you could get away with a smaller one as a result.
BTW I have made a large-ish class D oscillator which worked fine even with a large source inductor relative to the drive inductance, and yes I did use MOSFETs :)
One thing of note, your experiments with PWMing produced some ringing which you suggest non-overlapping drive might help. I came to the conclusion that you *need* a small amount of overlapping, because when both transitors are off the current stops flowing instantly, which would mean the source inductor voltage would rise to try to keep its current flowing and there's nowhere for the current to go (both transitors are off). The one I built has a PLD that guarantees a few 100ns of overlap and had no issues with noise.
Thanks for the correction. That saved me some time.
I like to leave some of the details to peoples' imaginations. A few people actually do have imaginations; I most always get along with them.
Maybe a controlled current source to add a compensating curvature, for a small added power dissipation. Or kick in some smaller inductors here and there on the curve, to pick up the droop. We don't really know the requirements, which is nice, because it allows for more ideas.
Switching inductor taps is always interesting. If you switch to an intermediate tap, current jumps up... just what you need to keep the triangle slope up, in a bang-bang sort of way.
I once did a buck switcher that converted +24 to +5, for a control system on some Navy ships. Schottky rectifier dides were fairly new, and none I could get were good for 24 volts reverse. I connected the catch diode to the center-tap of the buck inductor, which reduced the peak reverse voltage to about 15, good enough. But the consequence of, essentially, tap switching was that the ripple current into the +5 filter cap went way, way up. Had to use a big wet-slug tantalum.
So, puzzle of the day:
Take a 2 henry inductor that has a center-tap. Run 1 amp through the whole thing, steady-state. Then short one end of the inductor to the CT, and remove the current source. What happens when you do that?
Nice - and a much neater solution than mine. I am fixated on the Baxandall circuit, which lead me to a much more complicated (and less good) solution.
I'm wondering if you could get away with putting a second winding on your inductor, return one end to Vcc/2, and drive the second winding with the buffered (and offset by Vcc/2) voltage appearing at the switch end of the inductor, thus making up the voltage drive lost as the capacitor charges up
That wouldn't work with a single supply rail, but tapping the inductor say one third of the way down from Vcc and buffering that voltage into three times as many turns of over-winding might be persuaded to work.
It's feedback, and could well oscillate, but you are driving a capacitor, which would kill the high frequency gain - which is where oscillations like to happen - so it might be worth a try.
Something like that should work. Or just add a modulated linear current source at the top of the h-bridge, to add a little correction current, so you don't have to buy such a big inductor.
There may be a passive network that improves triangle linearity, too, at least at one frequency. Possibly a parallel LC tank in series with the main L, to raise the effective current-source impedance at 2F, the major ripple current frequency. Yeah, that might work.
It's interesting that the triangle sides are in fact s-curves, steepest in the middle. That's because they are actually little slices of a sine wave, straddling the zero crossing. I actually drew the sawtooth waveform in my sketch with such a little s-curve, but I didn't do that consciously. This circuit has some beautiful voltage and current waveforms.
Run Fred's sim and change L to about 0.85 H. It becomes a sinewave generator, sort of Baxandall-like.
--
He says it's pretty hush-hush, but since he's specified it as
capacitive and fairly high-voltage, I'd guess a piezo ceramic, poled
PVDF film, or some sort of varactor-like device.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.