Difference between self oscillating and non-self oscillating D-amp.

[sildex]

Hello. I'm kind of new in D-amps. I read the D-amp tutorial from IRF.com, I had a friend in the campus which builded one.
What I can't understand is the feedback loop in a D-amp. In a classic AB class the feedback is collected from the output back to the negative in of the amp and it akt's as a error correction.
But in D-class how can you take the pwm back into the analog in? The pwm will be mixed with the analog IN signal with is a sinwave.
Someone told me that that is trick to make a triangle wave and it ackts like an integrator. But these words doesn't have too much sens to me...
I can't understand what it happends step by step. Can anyone explain please how the loop works?
And why you will need a self oscillating topology? I mean this is not a T-class to self adapt the sample freq...
Practically my questions are regarding IRS2092 chip.

[RealForce]

Can someone answer this one? thanks.

[hitsware]

The ' pwm ' feedback is filtered back to analog by the action of the
47k resistor and then the capacitors to ground at the input.

One doesn't need self oscillation, but it is one approach.
This way the triangle wave (which is needed) is inherent
in the amplifier rather than having to be injected from
a separate source.

[RealForce]

Originally Posted by hitsware

The ' pwm ' feedback is filtered back to analog by the action of the
47k resistor and then the capacitors to ground at the input.

One doesn't need self oscillation, but it is one approach.
This way the triangle wave (which is needed) is inherent
in the amplifier rather than having to be injected from
a separate source.

Ahaaaaaaa GREAT EXPLANATION. Thank you so much man. You made the things very clear.

Thanks.

[sildex]

Yes, this kind of loop is a trick. And I think it's also hard to control. But...with simulations, and calculations of the transfer function of the loop I think it's possible to do it. From what I read extra about this loop is that it was invented to reduce the nr of components and using it you could have problems(ringing and so on) at high power >1000W. So at this kind of power over 1kW I think it's better to use the classical triangle generator and the sinusoidal loop.

[AP2]

In theory, the amplifier can come right in the esoteric world is with triangular wave modulator. but ... since it is very difficult to do work well (snr, linearity, resolution high-frequency sound etc) then we prefer selfoscillator.yes is simple But has other defects

Regards

[Eva]

Self oscillating loops have several interesting properties. For example, open loop gain is automatically compensated against changes in supply voltage, while in clocked modulators it's a direct function of supply voltage. This makes self-oscillating amplifier performance quite independent of supply rail voltage and sagging and improves PSRR dramatically.

Another advantage of self oscillating loops is that switching frequency is automatically reduced as the output approaches the rails, while keeping constant carrier residual amplitude at the output. This results in the minimum amount of switching events for a given carrier residual amplitude. In other words, switching losses are always as low as possible in self oscillating amplifiers. Switching frequency usually drops 2:1 or more before clipping.

Clocked modulators force the output stage to switch always at the same frequency, resulting in many switching events that are not really required, as the output gets closer to the rails and carrier residual amplitude becomes smaller and smaller.

Additionally, post-filter self-oscillating recycles the extra gain due to filter resonance as open loop gain, thus reducing output impedance (and THD too). In other words, the filter can resonate close to 20khz or even at a lower frequency without compromising frequency response.

Once you have put together a prototype which takes advantage of all this (and more), you don't feel like going back to clocked modulators.

[AP2]

Hi Eva,
I apologize to you for not having responded. but this is another fact).
ok, what you said is true though.

(Is that self-oscillating loops is Automatically Reduced switching frequency as the output Approaches the rails).
yes, this is one reason of Intermodulation audio signal, typically the mid range (vocals, trumpet, etc.) infact this defect is proportional to output power.
(Clocked modulators force the output stage to switch always at the Same Frequency, Resulting in many switching events That are not really required, as the output gets closer to the rails and residual carrier amplitude Becomes smaller and smaller.)
This makes no sense.because,while continuos clock if audio signal is smaller,current is smaller or zero. also this is very good because it is "sampleRate" it is necessary high and fixed for high performances.
My post ref to "absolute" Esoteric amplifier with two feedeback and active filter. it is not simple but I can assure you that you can not compare to a selfoscillator with post-filter feedback. (measurements and listening)

Regards

[81bas]

Originally Posted by Eva

Self oscillating loops have several interesting properties. For example, open loop gain is automatically compensated against changes in supply voltage, while in clocked modulators it's a direct function of supply voltage. This makes self-oscillating amplifier performance quite independent of supply rail voltage and sagging and improves PSRR dramatically.

I have an working example of tri-wave generator, which compensates it's output amplitude and offset from supply rails, improving PSRR of clocked class-D amp dramatically too Overall design surely becomes more complex, than in self-oscillating amp, but...

Originally Posted by Eva

Another advantage of self oscillating loops is that switching frequency is automatically reduced as the output approaches the rails, while keeping constant carrier residual amplitude at the output. This results in the minimum amount of switching events for a given carrier residual amplitude. In other words, switching losses are always as low as possible in self oscillating amplifiers. Switching frequency usually drops 2:1 or more before clipping.

I think, dropping down the switching frequency cannot have any positive effect for audio reproduction (at 10kHz-20kHz)

Originally Posted by Eva

Once you have put together a prototype which takes advantage of all this (and more), you don't feel like going back to clocked modulators.

well, if the result is satisfying, sure you will not rebuild it

[Eva]

Your main misconception about PWM amplification is assuming that a non-constant switching frequency results in non linearity.

The absolute minimum switching frequency is 2 times the maximum frequency you want to amplify, 44Khz for 22Khz audio.

The main reason for using 10 times higher frequencies (400Khz) is being able to deal with carrier residual with a 12db/oct output filter.

In clocked modulators carrier residual amplitude is maximum near 0V output and drops to 0 when the output approaches the rails (ugly looking waveform btw). Carrier residual amplitude is modulated by audio signal.

In self-oscillating modulators carrier residual amplitude is almost constant and frequency is smoothly reduced to keep it constant. There is no penalty, it just takes advantage of a "feature" of the output filter that you can't use with a fixed clock.

This reduction in switching frequency does not involve any reduction in the amount of negative feedback, so linearity is unchanged.

Another advantage of variable switching frequency is spread spectrum EMI. Variable switching frequency is now being used in many SMPS control ICs. I nearly always use it in clocked modulators too, for example in PFC.