DSD vs PCM resolution

[Hazard]

*Because the signal part/noise part is the mathematically useful way to work with an SDM audio bitstream, while it is not mathematically necessary to treat a PCM audio bitstream this way; and because, due to the math involved, it is possible to convert DSD to PCM with a low-pass filter (since we have been able to move the noise part into the ultrasonic range), referring to DSD as "PCM plus noise" is a shorthand mathematical truism, while indicating nothing about the actual audible noise performance, or any other type of sound quality performance, of DSD vs. PCM.

 

I have built my own DSD DAC (refer post #62 in this thread). DSD is passed throiugh a LPF and out comes - not PCM - but analog!! I have never heard that you can convert DSD to PCM with a LPF. All the theory says that if you filter DSD you get analog, and my DAC proves this. So where does your info come from?

[audiventory]

I have built my own DSD DAC (refer post #62 in this thread). DSD is passed throiugh a LPF and out comes - not PCM - but analog!! I have never heard that you can convert DSD to PCM with a LPF. All the theory says that if you filter DSD you get analog, and my DAC proves this. So where does your info come from?

 

DSD DAC is filtration DSD via analog low frequency filter.

 

DSD to PCM is filtration DSD via digital low frequency filter.

[Hiro]

DSD to PCM is filtration DSD via digital low frequency filter.

 

+ PCM decimation filter ?

[audiventory]

+ PCM decimation filter ?

 

You can keep (no decimation) full 2.8/5.6/11.2/22.5 MHz sample rate, but need filter 0 ... 20 kHz

 

But why need no decimation? For more space at hard disk?

 

Right decimation and final filtration is so slight harm, comparing full music production workflow.

 

Good final conversion like car analogy: at slight scratched glass pasted transparent film. It cause problem see through the glass at day or night?

[Hiro]

You can keep (no decimation) full 2.8/5.6/11.2/22.5 MHz sample rate, but need filter 0 ... 20 kHz

 

But why it need?

 

Right decimation and final filtration is so slight harm, comparing full music production workflow.

 

So, in the end the typical PCM process will involve downsampling, PCM decimation, digital and analog LPF, right?

[Paul R]

Expressed a little differently: If you tried to use the SDM output to express an instantaneous value it would not be mathematically useful because the quantization error for 1 bit is too large. The way to make it useful is to treat it as a stream composed of a signal part that works not as a collection of instantaneous values but as a collection of directional indicators; and an error part (noise) that is to be moved to an inaudible frequency range.

 

I am not sure you ever could treat it as an instantaneous value in any case, since the value is always either mix or min. But yes, without filters to do noise shaping, essentially putting all the noise into the ultrasonic range, it would be pretty ugly...

 

I would contend that makes DSD fundamentally different from PCM, but other people might dispute that, with some some valid justification for their opinions. Then again, I am old enough that I was taught the postulational notion in mathematics and the operational notion of science. (grin)

[audiventory]

So, in the end the typical PCM process will involve downsampling, PCM decimation, digital and analog LPF, right?

 

decimation = downsampling

 

no analog filration (digital DSD to digital PCM)

[Jud]

All the theory says that if you filter DSD you get analog, and my DAC proves this.

 

Yuri's provided the answer, but I just wanted to note (and my apologies for pointing it out regarding your comment, since we all do this):

 

- "All the theory I've read about" != all theory

 

- A proof of one possible outcome ("my DAC proves this") doesn't negate other possible outcomes.

 

I'm often surprised at what I read here and in the references linked by others, which is a great thing.

[Hiro]

decimation = downsampling

 

no analog filration (digital DSD to digital PCM)

 

Downsampling is part of it, but you have to additionally decimate the signal to PCM form for playback on a PCM DAC.

 

As for analog LPF you'll need it anyway, at least during PCM playback AFAIK.

[Hiro]

What Hiro was referring to, and what you need to remember especially when dealing with DACs is the noise at multiples of sampling rate. People tend to forget that the spectrum keeps repeating itself as mirror image around multiples of the sampling frequency.

 

Indeed.

[audiventory]

Downsampling is part of it, but you have to additionally decimate the signal to PCM form.

 

As for analog LPF you'll need it anyway, at least during PCM playback AFAIK.

 

I don't know accepted English term. In Russian sources used "децимация" (decimation).

 

"Downsampling" is decreasing sample rate. Single possible (except spline or same interpolation) way is decimation. Or decimation after upsampling (multiplying sample rate).

 

Interpolation (like spline), as rule, no need use in end-user format resolution conversion.

 

Thus, "decimation" and "downsampling" we can consider as synonyms.

 

"decimation" always downsampling.

 

Filtration used before decimation. For avoiding penitration of artefacts to final band (half final sample rate).

 

More about downsampling, decimation and filtration here How Convert Sample Rate. Divide. Downsampling

 

 

 

Analog filtration impossible for pure digital conversion (fully into software).

[Hiro]

I don't know accepted English term. In Russian sources used "децимация" (decimation).

 

"Downsampling" is decreasing sample rate. Single possible way is decimation. Or decimation after upsampling (multiplying sample rate).

 

Thus, "decimation" and "downsampling" we can consider as synonims.

 

 

Let me give you an example to explain what I mean. You can for example downsample a PCM file from 96kHz to 44.1kHz - that's simple downsampling. When you deal with DSD, however, you have to both downsample it, and convert the DSD stream to a multibit PCM form.

[Hiro]

- A proof of one possible outcome ("my DAC proves this") doesn't negate other possible outcomes.

 

A proof of one possible outcome isn't meant to prove that other approaches are not possible, only that the approach being applied is possible.

[audiventory]

Let me give you an example to explain what I mean. You can for example downsample a PCM file from 96kHz to 44.1kHz - that's simple downsampling. When you deal with DSD, however, you have to both downsample it, and convert the DSD stream to a multibit PCM form.

 

Computer inside only multibit.

[Hiro]

Computer inside only multibit.

 

Ah, I get it, ipv4 addresses are PCM

[Jud]

I've done quite a bit of beamforming DSP stuff for sonar arrays, adaptive and non-adaptive. Would be kind of fun to do the similar things in air acoustics with microphones and loudspeaker arrays. I know people who do that kind of stuff, but it has been so far limited use in music recording and home environment cases. Apart from so called "sound bars" which have not been very hifi.

 

Easy way to put something like 128 channels into good use...

 

I have a friend with several patents in this area. Here is one where the application is in chemotherapy (heating a tumor to make it more vulnerable to chemotherapy, without similarly affecting the surrounding tissue): Patent US20150055734 - Medical apparatus and methods including an array system for segmenting ... - Google Patents

[mansr]

I don't know accepted English term. In Russian sources used "децимация" (decimation).

 

"Downsampling" is decreasing sample rate. Single possible (except spline or same interpolation) way is decimation. Or decimation after upsampling (multiplying sample rate).

 

Interpolation (like spline), as rule, no need use in end-user format resolution conversion.

 

Thus, "decimation" and "downsampling" we can consider as synonyms.

 

"decimation" always downsampling.

 

Filtration used before decimation. For avoiding penitration of artefacts to final band (half final sample rate).

 

Sometimes the term "decimation" is used to denote simply removing all but one in every N samples to reduce the sample rate by 1/N. This is fine if the signal has been properly lowpass filtered first. For general M/N rate reduction, an interpolation filter must be used to reconstruct the value of the signal at times between the original sample points.

[audiventory]

Yes. Filtering should be before decimation. Filter of low frequencies is kind of interpolation.

[audiventory]

I have a friend with several patents in this area. Here is one where the application is in chemotherapy (heating a tumor to make it more vulnerable to chemotherapy, without similarly affecting the surrounding tissue): Patent US20150055734 - Medical apparatus and methods including an array system for segmenting ... - Google Patents

 

Speaker or antennas arrays is real method that work. It allow summarize power in certain point. I think, Yamaha's speakers is well known example in audio industry Sound Bars - Home Theater Systems - Audio & Visual - Products - Yamaha United States

 

But for big room need big speaker's volume anyway.

[Miska]

DSD DAC is filtration DSD via analog low frequency filter.

 

Just remember that you can model analog circuits in computers too, just like many software emulations of analog synths do.

 

For example I had a full model of the DSC1 DAC in computer before making the hardware realization.

 

 

PCM is typically binary coded (two's exponent) Nyquist sampled values without significant noise shaping.

 

SDM is typically unary coded (thermometer code) non-Nyquist sampled values and relies on heavy noise shaping.

 

 

DSC1 uses scrambled 32-bit unary coded values for conversion, while being bit-perfect with DSD data input. Values converted to analog vary between 0 and 32, thus having total of 33 possible levels.

 

While PCM ladder DACs depend on binary coded values...

[Miska]

Oh good grief. Hiro, just read below and learn. For more, see BitPerfect: Sigma-Delta Modulators - Part I. and BitPerfect: Sigma-Delta Modulators - Part II.

 

The second part has error where it says multi-bit SDM DAC would use R2R ladder for conversion. This is not the case and would leave out one very important phase of the process - DEM.

[audiventory]

I prefer always seek total root for all phenomenons

 

Software emulation of analog filter - it is digital filter.

[Miska]

I prefer always seek total root for all phenomenons

 

Software emulation of analog filter - it is digital filter.

 

Software emulation of class-A analog power amplifier is... (?)

 

You can very exactly simulate vacuum tubes, FETs and all kinds of other electronics stuff in a computer. You can make a SPICE model of pretty much anything and if you have enough CPU power you can run it in real time.

[audiventory]

If I don't mistaken, ideal class A: Y = k * X

 

That's emulation right now!

[Miska]

If I don't mistaken, ideal class A: Y = k * X

 

That's emulation right now!

 

Not really, there's nothing making it class-A. And you need to take into account different load situations such as inductive and capacitive loads (phase difference between voltage and current).