World’s First Valid Comparison of PCM versus DSD?

[Don Hills]

Wish fool thinking ? ¨-)

 

I see what you did there.

[Le Concombre Masqué]

Peace & Love, brothers & sisters!

 

Let’s make pcm and dsd help each other to grow stronger!

 

When I started to respect digital, I benchmarked my analog gear to my digital ; ie I set my cartridge tracking force close to the lowest range recommended by the manufacturer…

 

Used to harshnessless sound by the sake of EQed Pcm, I pay attention, more than ever, to Absolute Polarity when playing DSD files (btw : found out you can't set a general polarity inversion in A+ for dsd files. And as a guy who until recently converted dsd files to pcm for eqing, I have the strong suspicion that DoP fed files from A+ to TEAC 501 get polarity inverted in the process ; won't waste time investigating, and just mind AP on a record to record basis, but this might be a parameter to take into account when comparing dsd and pcm).

 

Digging dsd’s liveliness, speed, I dropped srcing and equalizing on fly and came back to Amarra (EQing offline in RX3/Alloy2).

 

And I admit that, DSD and PCM having different assets, the sweet spot for volume might be different.

 

 

So, as a consumer, whatever you pros bring, I’m happy, as long as you make the best from the best source with the output you choose.

[Allan F]

So, as a consumer, whatever you pros bring, I’m happy, as long as you make the best from the best source with the output you choose.

 

+1

[esldude]

Peace & Love, brothers & sisters!

 

 

Used to harshnessless sound by the sake of EQed Pcm, I pay attention, more than ever, to Absolute Polarity when playing DSD files (btw : found out you can't set a general polarity inversion in A+ for dsd files. And as a guy who until recently converted dsd files to pcm for eqing, I have the strong suspicion that DoP fed files from A+ to TEAC 501 get polarity inverted in the process ; won't waste time investigating, and just mind AP on a record to record basis, but this might be a parameter to take into account when comparing dsd and pcm).

 

D

 

I can tell you a simple signal if you can get it into DSD form which will allow you to check polarity with a DC voltmeter at the speakers. It is a diode processed sine wave. Check at speaker terminals. A small plus DC voltage is correct and a small negative one is reversed. Can give more details if you are interested. Handy for checking total system polarity in PCM for as well.

[Le Concombre Masqué]

I can tell you a simple signal if you can get it into DSD form which will allow you to check polarity with a DC voltmeter at the speakers. It is a diode processed sine wave. Check at speaker terminals. A small plus DC voltage is correct and a small negative one is reversed. Can give more details if you are interested. Handy for checking total system polarity in PCM for as well.

thank you ; " if you can get it into DSD form" : true, I know of no test signal as I have for LP (Cardas's) or PCM (qobuz's)

Anyway, i deal with AP on a record to record basis, by ears

[esldude]

http://www.computeraudiophile.com/f8-general-forum/digital-recording-polarity-11268/index2.html

 

Look at post #61 in this thread for attached test files.

 

One is a fixed tone, usually good enough, another is white noise. They unzip into mp3 which is fine for this test. If you can somehow record them with DSD or somehow convert them it would work. There would be no doubt of the polarity then.

 

Korg Audiogate and Weiss Saracon can do the conversion to DSD. I don't have either or I would supply the file to you. If nothing else, maybe someone with this software can convert it and make it available.

[Boris75]

 

One is a fixed tone, usually good enough, another is white noise.

 

Isn't that potentially dangerous for tweeters? I am asking the question because I would like to run this polarity test, but I am overcautious with my equipment.

[esldude]

Isn't that potentially dangerous for tweeters? I am asking the question because I would like to run this polarity test, but I am overcautious with my equipment.

 

I would advise starting such tests with a low volume as there is no need to turn it up higher than gives a good reading.

 

The single tone was 440 hz so it shouldn't be an issue. Someone wanted one with more than one frequency to measure polarity at the drivers in a multi-way system using a complex crossover. Which is why I made one with white noise. For simple polarity checks the 440 tone would be fine.

 

Now if your asking if the diode type tone has lots of harmonics dangerous to tweeters, then no it doesn't. It does create harmonics, but they are descending in level. The 440 tone I used results in -7.5 db at 880 hz and is down -50 db by 8800 hz.

 

The white noise diode tone could cause problems for tweeters. The highest single sample values are at about -6 db ocurring about one per 30 samples. The average level is at -12 db. If looked at in an FFT of 128 bands you have -24db at all frequencies. Probably not a problem for tweeters, but with a big amp turned up enough it could be. So again, just start with low levels. Even 2 or 3 volts at the speaker is plenty for this test.

[Boris75]

 

The single tone was 440 hz so it shouldn't be an issue. Someone wanted one with more than one frequency to measure polarity at the drivers in a multi-way system using a complex crossover. Which is why I made one with white noise. For simple polarity checks the 440 tone would be fine.

 

Thank you very much for your very clear response. I don't need to check the polarity of my sub, and my speakers have passive cross-overs, so a diode-processed 440Hz sine wave is enough for me.

[wgscott]

Is it possible to convert back and forth between DSD and PCM losslessly? If so, how high a sampling frequency do you need?

[Jud]

Is it possible to convert back and forth between DSD and PCM losslessly? If so, how high a sampling frequency do you need?

 

I am reasonably sure the answer is as follows, and ask that someone who knows correct me if I'm wrong:

 

- No, if what you mean by "losslessly" is "with completely identical measured characteristics."

 

- If you mean without using lossy compression formats (which I doubt), the answer is of course.

 

- This has nothing to do with sampling frequency, as just even-numbered-integer or rational-number conversions among sample rates within the PCM format (e.g., 44.1kHz to 176.4kHz), or within the DSD format (e.g., DSD 5.6mHz to DSD 2.8mHz), require use of filters and are not a matter of simple multiplication.

[Boris75]

Is it possible to convert back and forth between DSD and PCM losslessly? If so, how high a sampling frequency do you need?

 

I am not 100% certain, but I think that, with the right algorithms, upsampling DSD to 2.8224MHz 8bit PCM (so-called "DSD-wide") and then downsampling to DSD results in the same DSD file as at the beginning and that this is what Pyramix and SADiE workstations do.

[elcorso]

Is it possible to convert back and forth between DSD and PCM losslessly? If so, how high a sampling frequency do you need?

 

I think Jussi (Miska) answered this question, but don't remember the thread: He said something about the filters in PCM doesn't allow this conversion losslessly, then it will be degradation.

 

Roch

[lightminer]

A different approach - use a Sonoma vs a highly-regarded PCM recorder at the same event, no post-processing at all and see how they sound? That tells you the edge of the issue. Then we still have to figure out how to allow people to do some processing with no losses on the DSD side.

 

(Mostly related: I'd be curious to see what the Mapleshade folks think about a direct Sonoma DSD128 recording with 0 post-processing played back on a Lampizator DSD-only dac.)

[elcorso]

A different approach - use a Sonoma vs a highly-regarded PCM recorder at the same event, no post-processing at all and see how they sound? That tells you the edge of the issue. Then we still have to figure out how to allow people to do some processing with no losses on the DSD side.

 

(Mostly related: I'd be curious to see what the Mapleshade folks think about a direct Sonoma DSD128 recording with 0 post-processing played back on a Lampizator DSD-only dac.)

 

Another (different) approach,

 

How about analogue tape recording, editing and mixing and the converted to DSD (or PCM) like a lot of Cookie Marenco's recordings?

 

I would enjoy them until 2" analogue tapes continue to be available...!

 

Roch

[Miska]

Then we still have to figure out how to allow people to do some processing with no losses on the DSD side.

 

PCM processing is not any different. If you mix eight tracks of 24-bit PCM without any relative gain control you end up with 32 bits. If you multiply two 24-bit integers the result is 48-bit integer. If you adjust level in steps other than exactly 6.0206... dB you end up with way more bits. Also when you process any EQ you end up with arbitrary number of bits. If you multiply value of Pi with square root of two you end up with arbitrary number of bits (and both values have arbitrary precision to begin with). In practical system you have to limit the precision somewhere, only in theoretical formulas where you don't have to write out actual numbers you have infinite precision.

 

When you do limit the precision what matters is how you do it...

[Matias]

I think most DAWs use 64 bit internal precision by now. This should be plenty for a lot of volume, effects, mixing, etc. The real problem is recording at 44/48kHz and squeezing all dynamics...

[Miska]

I think most DAWs use 64 bit internal precision by now. This should be plenty for a lot of volume, effects, mixing, etc. The real problem is recording at 44/48kHz and squeezing all dynamics...

 

I recently changed some of the processing to 80-bit because 64-bit was running out. You also need to get that 64-bit out to something else for distribution, I haven't seen 64-bit PCM downloads yet...

 

A challenge: try to make all the DAW processing inversible (bijection) so that you can go back to the original.

[Matias]

Miska, I said wrong: the point is not exporting 64-bit files from the DAW. Today's 24 bit hi-res is the "new standard". What I meant is that internally it uses all these bits to process and mix the tracks.

[Miska]

Miska, I said wrong: the point is not exporting 64-bit files from the DAW. Today's 24 bit hi-res is the "new standard". What I meant is that internally it uses all these bits to process and mix the tracks.

 

Yes, but exporting that to 24-bit output is lossy process.

 

Processing PCM or processing DSD doesn't make practically any difference from precision loss point of view. Processing DSD64 takes a around 120x more processing resources than processing 44.1 kHz PCM, but so what, you can buy more processing power with money. And Moore's law makes sure you get roughly 2x more for same money every two years.

[Miska]

Let's look at it this way:

 

CD was published 1982 and SACD 1999. If we start from 1x processing resources of 1982.

 

1982: 1x, 1984: 2x, 1986: 4x, 1988: 8x, 1990: 16x, 1992: 32x, 1994: 64x, 1996: 128x, 1998: 256x, 2000: 512x, 2002: 1024x, 2004: 2048x, 2006: 4096x, 2008: 8192x, 2010: 16384x, 2012: 32768x, 2014: 65536x

 

Same from SACD perspective:

1999: 1x, 2001: 2x, 2003: 4x, 2005: 8x, 2007: 16x, 2009: 32x, 2011: 64x, 2013: 128x, 2015: 256x

 

So now we have roughly around 65536x more processing power available than when CD was published. Or 256x more than when SACD was published. So we can do quite a bit more sophisticated things than back then.