Envelope-tracking power amplifier

[earthtonesaudio]

...Or by another name, Class G or class H amplifiers.  Basically this is a normal class A or A/B amplifier output stage, but the supply voltage is modulated by the input signal, so that larger input levels result in larger supply voltages for the output devices.  Conversely, when the input signal is low, the output supply is reduced, increasing efficiency. 

Normally these designs are somewhat complex, but I think one could do this rather simply with a few power MOSFETs and a pulse-width modulated audio signal, as long as you got the details right.  First you would convert the audio to PWM, then drive both the output stage and a SMPS with the PWM signal.  The SMPS would increase the voltage to the output stage as a function of the audio duty cycle, and the output stage would simply switch whatever voltage appeared across it, whenever the PWM signal told it to switch.  Add an LC filter and speaker, and off you go.

Now about those details...
-First the PWM would have to be pretty robust.  Getting stuck high or low is not an option, and a stable frequency would be nice.
-Next you'd need to make sure the PWM circuitry could drive 3 power FETs without problems.  Buffers=good.
-Then there's the matter of the SMPS.  You'd want it to respond quickly to both attack and release, so depending on the maximum desired voltage you could possibly make the inductor and capacitor slightly smaller.  This would mean a higher operating frequency, but that's desirable for audio bandwidth anyway.
-Sticking to a single-ended power supply would greatly simplify the design, at the cost of a mandatory output capacitor.  I'm not sure about this, but I think you could use a smaller value, higher voltage capacitor before the output filter rather than after and not lose too much low end.

Any other considerations I may be missing? 


Note: I know this is not technically a stompbox, but it could be stompbox-sized, battery powered, etc.

[petemoore]

Any other considerations I may be missing?   
  Almost completely uncharted territories often have these, after making marks on the map it will be easier to find them.
  Note: I know this is not technically a stompbox, but it could be stompbox-sized, battery powered, etc.
  Dunno, sounds like an 'effect' to me.
  I thought about using the power supply as an input, typed about it a few times too.
  Sounds like a DC amplifier, which would need to be a near perfect ''moving voltage reference with no ripple'', and able to supply current...whew...let me know if you work it out or if theres a workaround...there might be a fairly simple way.
  If the ripple was hard to get rid of [in the sweeping DC reference voltage], perhaps it could be made to be of an input frequency such as the input source has...my early posts leaned toward some ripple as 'fuzz' ..seemed a possible or impossible alternative to diode clipping at the time.
  then I had a novel concept and it still haunts my lazynass.

[Cliff Schecht]

Modulating the supply voltage will also modulate the bias current and will affect gain and linearity as well. May sound good, may sound bad. Remember that not all distortion sounds good.

[earthtonesaudio]

Any other considerations I may be missing?   
  Almost completely uncharted territories often have these, after making marks on the map it will be easier to find them.

Haha, true!

Modulating the supply voltage will also modulate the bias current and will affect gain and linearity as well. May sound good, may sound bad. Remember that not all distortion sounds good.

This makes perfect sense to me if the output stage was to be operated in the linear region.  But what about for a class D/switched output stage?  Would this still be a concern?

[Andi]

Modulating the supply voltage will also modulate the bias current and will affect gain and linearity as well. May sound good, may sound bad. Remember that not all distortion sounds good.

Not if you go bipolar...

Class G is in use in commercial power amps.

[Cliff Schecht]

Modulating the supply voltage will also modulate the bias current and will affect gain and linearity as well. May sound good, may sound bad. Remember that not all distortion sounds good.

Not if you go bipolar...

Class G is in use in commercial power amps.

Well even with bipolar, you can run into non-linearities. I've never played with, or even looked at, class G or class H amplifiers though. If the supply modulation is limited enough then achieving a predictable output is easily possible. I like the idea still, please don't take my last post as a negative one, I was just pointing out that bias conditions do affect beta/transconductance (although these effects can be minimized).

[earthtonesaudio]

I've posted this a few times before, but here's a VERY ROUGH outline of what I'm talking about. 

Please disregard all the horrible smoke-producing errors (there should be an output filter, for one), and instead just look at it as a circuit that converts the audio to PWM, then sends that to a boost converter and also the output transistors. 



And then assume that the PWM circuit can easily drive all the power FETs, there are no cross-conduction issues, and perhaps the output transistors are delayed slightly behind the SMPS transistor.