Composite amplifier with Class-D

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Composite amplifiers with Class-A or Class-(A)B are a well known technique to extensively improve the total performance of a system.

I was wondering if this could be used with a Class-D output stages as well?
What are the practical (and theoretical) limits and issues?

So in other words, using an high performance (linear) opamp with a Class-D output stage in the loop.
 
feedback around a Class D stage is typically limited by upper audio frequencies by the Class D power output filter phase shift

Bruno Putzeys probably does the best anyone knows how to today in audio

Bruno Putzeys: Head of the Class (D, That is) | Sound & Vision


some put a power Class A stage in series with the Class D, let the Class D do the main V swing

and the feedforward "Current Dumping" technique could let you parallel a Class A with a Class D - but you need a "power combiner" with serious passive component matching

mentioned at the end of: https://web.archive.org/web/20060221012225/http://quad405.com/jaes.pdf
 
feedback around a Class D stage is typically limited by upper audio frequencies by the Class D power output filter phase shift

Bruno Putzeys probably does the best anyone knows how to today in audio

Bruno Putzeys: Head of the Class (D, That is) | Sound & Vision


some put a power Class A stage in series with the Class D, let the Class D do the main V swing

and the feedforward "Current Dumping" technique could let you parallel a Class A with a Class D - but you need a "power combiner" with serious passive component matching

mentioned at the end of: https://web.archive.org/web/20060221012225/http://quad405.com/jaes.pdf

Mr. Putzeys is also a superb author per his discussion of his technology. He further excels at philosophy which pemeated his article.
 
It's been done for years. (With Pre, Post or Pre+Post-Filter-Feedback)

If you would mind to read the stuff jcx posted, you would already have the answer.

The article is not really readable, contains a lot of blacked out pieces.

I also don't see how the following default composite setup compares to that article?
 

Attachments

  • Class-D composite.png
    Class-D composite.png
    32.1 KB · Views: 428
The question is answered through logic elimination of all objective reasons for advantage. Only subjective advantage remains.
"Subjective advantage" * "objective disadvantage" = null end
"Subjective advantage" * "objective advantage" = high end
"Subjective disadvantage" * "objective advantage" = null end
"Subjective disadvantage" * "objective disadvantage" = low end

btw: Thanks for posting links to serious papers.
 
Your question is flawed. If you look into class D amp design, they indeed have op-amps and other linear devices embedded in them. As with class A and AB amplifiers, crappy linear componentry is a recipe for crappy sound. As Eva has stated, there are subjective and objective elements to the answer of your question. Yes, you can do it, but how is another question. Your answer lies in your domain.
 
Composite amplifiers with Class-A or Class-(A)B are a well known technique to extensively improve the total performance of a system.

I was wondering if this could be used with a Class-D output stages as well?
What are the practical (and theoretical) limits and issues?

So in other words, using an high performance (linear) opamp with a Class-D output stage in the loop.
It may help to have an intuitive feel for how an all-linear (okay, almost-linear) composite amplifier works. I resist the claim that I have this, but I remember reading through much of the MyRef and similar threads for composite chip-amps.

Composite amps have a lot of careful compensation for the increasing phase shift with increasing frequency (this happens with each amplifier alone, and cascading two amplifiers complicates the situation). I recall the use of 1 percent components in places where 20 percent would good enough to make a stable single-amplifier circuit (well, a single-amp circuit doesn't even have those intermediate 'places', but I hope the point comes across).

The phase shift of a Class D amplifier is way beyond that of most linear amps, due to several things. There's the input sampling, which causes the output to be an inherently delayed version of the input. Then there's the output filter, often a two-pole (inductor and capacitor) circuit which adds its phase shift, and then the feedback is taken from that point. The output must be low-pass filtered for many reasons - the one pertinent here is that you can't use the raw output, a PWM'ed square wave, as feedback for a linear amplifier.

I suppose the way to do this is to plot both gain and phase shift with frequency of a Class D amp, and figure out what kind of response the other amplifier would need to have to prevent the gain/phase shift situation that would give oscillation. As I mentioned, composite amps already use high precision components to "work right" and be flat up to 20kHz. No doubt it would be that much more difficult for a Class D composite amp. Yes, you could "do it" but with the much greater accuracy needed, things such as adding shielding (thus changing the capacitance to ground) and temperature variation (causing slight component value changes) may be enough to make the circuit unstable.
 
This plot shows phase shift of state-of-the-art class D modulator (post-filter feedback), a constant 6.7us delay across the audio band (equivalent to 2.3mm), frequency response deviation is +/-0.05dB, output impedance at 20khz is less than 0.005 ohms.
 

Attachments

  • class_D_phase_shift.png
    class_D_phase_shift.png
    3.5 KB · Views: 260
And this plot shows same modulator in action:
- Magenta trace: input voltage
- Green trace: output voltage
- Red trace: PWM
- Cyan trace: current (load: 100uH+4r load simulating a tweeter)
- Blue trace: input to comparator
- Grey trace: integrator volt
- Dotted rectangle: phase shift, indicated at bottom left.

No ornaments.
 

Attachments

  • class_D_phase_shift_2.png
    class_D_phase_shift_2.png
    48.8 KB · Views: 268
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.