Are Class D Amplifiers really "digital"?

[Miska]

I would say class-D amplifier can be more "analog" or "digital", depending on how the modulator works. If the modulator is analog, it is closer to analog one, although still switching PWM device. While one using digital modulator is essentially a "power DAC" and thus much closer to "digital".

 

[Miska]

On 9/13/2020 at 6:30 PM, kumakuma said:

Yes, the PWM waveform appears digital but the width of Vo will vary in an "analog" way. The non-digital nature of the waveform will be even more apparent when Vs is more complex than a sine wave.

 

Or you could say Vo works in a DSD'ish way...

[Miska]

4 hours ago, firedog said:

This is really another pointless audiophile debate that gets us nowhere.

But I agree with Paul M.: a PWM signal can be played back directly by analog equipment and produce sound. Poor quality sound, but sound. Therefore it is analog.

 

Since DSD is essentially PWM, we can also conclude that DSD is analog.

 

[Miska]

13 minutes ago, SoundAndMotion said:

That's not true, as the video @kumakumalinked below and many web pages explain. DSD is PDM, pulse density modulation, and the way it's encoded and decoded is different from PWM, pulse width modulation. In some ways they look similar, especially with full scale simple sine waves, but they're not the same.

This OT post of mine doesn't answer the "analog-ness" though.

 

No, PWM is just a special cased version of PDM as Wikipedia also says. In practice all DSD is treated as PWM, because none of the converters transition state between subsequent samples of same value.

 

Essentially a DSD converter / PWM amplifier has no idea where the pulse train originates from, whether it originates from analog or digital modulator.

 

We've already demoed a "power DAC" amplifiers in Munich couple of times, based on my software modulators.

 

[Miska]

1 hour ago, SoundAndMotion said:

So the methods and ideas are related, but the implementation is different, i.e. the signals will be different, except at the peaks where they're the same. And several engineering/programming sites describe them as related, but different.

 

Every modulator producing DSD output gives you different bit stream. Some are better, some are worse. Actually same for analog modulators too. So there's no one way to do things, modulators vary a lot.

 

So where I draw the distinction is whether the same pulse train / D/A conversion stage works or not.

 

[Miska]

On 9/15/2020 at 9:01 PM, SoundAndMotion said:

Even the Wikipedia article you cite says:"For a 50% voltage with a resolution of 8-bits, a PWM waveform will turn on for 128 clock cycles and then off for the remaining 128 cycles. With PDM and the same clock rate the signal would alternate between on and off every other cycle. The average is 50% for both waveforms, but the PDM signal switches more often."

 

Actually none of real modulators produce such output, or then the modulator is bad. Yes you can use 01010101 as a DSD silence pattern. As much as you can also use 01101001 or many other variations. But if you look for example output of my modulators, they don't produce either one for that case, but instead a random pattern.

 

[Miska]

4 hours ago, semente said:

What about FM broadcast? Is it digital or analogue (or both)?

 

I would guess most of the transmitters are digital these days, possibly using some sort of SDR (Software Defined Radio).

 

The more complex modulations like QAM256 used for DVB and mobile networks are always digital creation.

 

By the way Technics also just recently announced a digital class-D amp using GaN FETs:

https://www.technics.com/uk/news/414/index.html

 

[Miska]

8 hours ago, barrows said:

So far, I do not agree with this sentiment.  I have been working with the Purifi class D modules here, trying to get the best out of them.  So far they are on par, or sounding better than, any amplifiers I have heard here.  The only amp which has come close (and could be considered preferable, maybe, in some areas), has been the Bricasti M-25, which is a $20K component.  And I am still making improvements to my Purifi amp build.  As far as I can tell, the Purifi modules themselves are as near "perfect" output stage as exists.  Getting the best out of them requires designing and implementing the best input stage, power supplies, layout and wiring, as their performance appears to be limited only by the implementation.

So far, the GaN amplifiers which are available, have worse measured performance than the Purifi demo set up: so I do not see a real world advantage for higher switching frequencies, at least yet, no one has produced a GaN amp which actually shows an improvement in performance.

To those who seem to think higher switching rates are going to result in better class D amp performance, where are the numbers to back up this assertion, I have not seen any.

 

For example most low-rate switching amps I've seen so far have issue that the bandwidth (flat frequency response) doesn't cover entire spectrum of hires material - that is 100 kHz. In addition they have aliasing problems in presence of high frequency content.

 

Another challenge is still how much ultrasonic noise they put out because the distance between the noise and the audio signals is so small that it is hard to make a simple efficient low impedance output filter.

 

I guess Purifi doesn't go flat to 100+ kHz bandwidth either?

 

[Miska]

On 9/27/2020 at 5:46 AM, sandyk said:

This isn't restricted to Digital amplifiers either, with many Analogue amplifiers having an Input filter well below the actual capabilities of many of them.

 

Mostly Naim comes to mind, but same applies to their digital gear. Hires frequency range is just cut off.

 

Many amps have input filter with something like 1st order RC filter with -3 dB point at 100 kHz.