delta sigma vs PWM for class D amp

[sarepairman2]

What are the benefits of each method?

[moffy]

I suspect that delta sigma would require more bandwidth than PWM.i.e. a lot higher pulse frequency probably between 10MHz and 20MHz.

[coppice]

Think of a narrow PWM pulse. Its pretty close to a Dirac sample. Think about a fat PWM pulse. Its close to a zero order hold sample. So, as the PWM pulse width varies the frequency response also varies. PWM is a pretty funky way to represent a signal.

[rs20]

I suspect that delta sigma would require more bandwidth than PWM.i.e. a lot higher pulse frequency probably between 10MHz and 20MHz.

Alternative view: Digitally produced 100kHz 8-bit PWM is generally produced by having a (for the sake of argument) 25.600MHz clock, and the PWM generator wanting to output 50% simply counts out 128 0's, and then 128 1's, etc etc.

If you could use that same 25.6MHz clock but output Delta Sigma, you'd get alternate 0 and 1's; the noise is now much higher frequency and easier to filter away now.

In short, for a given underlying clock rate, Delta Sigma is actually superior to digitally produced PWM.

[NiHaoMike]

To blur things a bit more, you can mix Delta Sigma with classical PWM, a technique often referred to as MASH.

And then there are modified Delta Sigma algorithms that try to limit the maximum switching frequency. That is often used in advanced power inverters.

[coppice]

To blur things a bit more, you can mix Delta Sigma with classical PWM, a technique often referred to as MASH.

And then there are modified Delta Sigma algorithms that try to limit the maximum switching frequency. That is often used in advanced power inverters.

I think MASH is cascading multiple SDMs, with ones error feeds into the next one's input.
There is PWM SDM, basically a multibit SDM with PWM N bit to 1 bit DAC.

MASH is a multi stage noise shaped sigma delta converter, implemented as a chain of (I think) first order modulators.

[diyaudio]

What are the benefits of each method?

Im no audio fool so I cannot speak for one. ignore my alias in that context. ( I spent a year studying this topic in 2008 when it was a hot topic..still is  )

In context of Class-D audio amplifiers. (modulator section)
 
PWM is simple, old school and just works!, requires a external continuous ref clock typical.. (also the system uses more components and requires a stable clock), from a power efficiency point of view it doesn't use the spread spectrum (fsw is contentious, right up till the output reaches the rails VCC(max), fsw remains constant), and as a result has an inherit inefficiency.
(this is hard for people to digest as PWM by design is an efficient system by context)     

Delta Sigma, clock is generated from the pole/zero (integrator order) in the system, uses less components, system uses a spread spectrum (switching events are less as the output reaches VCC(max) approximately fsw/2 ) Also, the system uses pre-feedback (sometimes both pre and post feedback), the feedback taps off the high frequency switching node (see attached image) and injects rf spurious signals into the system resulting in frequencies most comparators cannot handle (resulting in system instability), they require advance network control theory system analysis for proper compensation.
These "types" of systems are easily disturbed when design incorrectly act as an radio antenna/transmitter and interfere with near by radio receivers.   
(few can design class-d Delta Sigma modulators correctly)


There are people who favour one over the other and defend their design of modulator, this topic can easily turn into a dog fight. 

from a sound point of view, I have no comment, if that what you really asking?, no one will be able to tell the  difference in sound between Delta Sigma, PWM or UCD modulators. only audio fools with their alien abilities.   


ucd is so far the best topology from an engineering point of view, I would recommend the study of class-d papers by bruno putzeys , I always recommend people who are interested in this topic to follow him he is of few that uses science to formulate prove in the audio community.       

 

[dom0]

from a sound point of view, I have no comment, if that what you really asking?, no one will be able to tell the  difference in sound between Delta Sigma, PWM or UCD modulators. only audio fools with their alien abilities.

Not true. The distortions of older PWM and some older DSM power amplifiers are easily heard at mid-band audio. Fair enough, since they were mostly meant for high power bass amplifiers anyway.

[diyaudio]

from a sound point of view, I have no comment, if that what you really asking?, no one will be able to tell the  difference in sound between Delta Sigma, PWM or UCD modulators. only audio fools with their alien abilities.

Not true. The distortions of older PWM and some older DSM power amplifiers are easily heard at mid-band audio. Fair enough, since they were mostly meant for high power bass amplifiers anyway.

I have noticed older systems, ones that are 8 > years ago, or people who designed modulators based on older theories or even companies who were/are clueless .. have a very steep demodulator LP stages, some cutting at 25Khz/db per octave, some even employing 24db per octave stages.. and most older designs have alot of trouble with handling of the z-load (speaker)

you are correct making the point that they were only good enough for bass/mid amplifiers (mostly as PA amplifiers),  the high frequency damping was excessively high, mostly... to hide the bad designs...

we are a few days away from 2016, and alot has changed since then. I only apply what I mentioned above to properly designed modulators, not crappy designs, which of course can easily be distinguished from one another.(any day)   



 
 
       

[Kleinstein]

The digitally generated SDM amplifier essentially combines DAC and per amplifier in a single stage. As the power stage is often not that ideal in respect to linearity (e.g. R_on is not neglegible and load independent) - this way can not get the highers performance. Also the filter at the output may not be as linear als a small signal filter, especially if ferrites are used.

Digital generated PWM (without feedback) is even worse - having limited resollution, of higher clock frequenies.

Analog feedback in more classical class D amplifiers creates a kind of PWM Signal, but without quantization limit. It can compensate for nonlinearities in the output FETs and the filters.

Even if the output may not be perfect, it can be good enough so that many listeners can not hear the difference, especially as speakers are not perfect either.