World’s First Valid Comparison of PCM versus DSD?

[Boris75]

Like asking Nolan Ryan and Greg Maddux to throw fastballs, and we'll pick the best pitcher based on that. (...)

 

The metaphor is difficult to grasp for the European that I am. Could you phrase it in terms of football, by which I mean soccer?

[Maldur]

I had one Pioneer DVD/SACD player that was really bad and had only 96/24 capable DACs (SACD went through horrible Mediatek PCM conversion).

 

That is interesting (in historical sense). I don't know any Pio player with Mediatek and 24/96 DAC. I have still first Pioneer DV-575 with SACD conversion (24/88) via Mediatek MT1389EE, DAC was PCM1742KE - cheapest 24/192 DAC from Ti. For DVD-A, this old man plays actually good, SACD is not so good as from real SACD player but with SACD-R capable firmware and some modified hardware (like CarlosFM did) it is good for everyday backround music. Actually, it plays now on my childrens rooms...

Did I say - Goldmund CD18 is based heavily on that player?

[Maldur]

Originally Posted by Maldur

First of all we need to know - what he smokes...?

 

In the interests of full disclosure, I'd like it to be known that I only smoke the very best.

Good to know.

Actually the reason of my such posting was in linked site, where mister T says: "And at it's best SACD still sounded WAY worse than a pure PCM derived CD played back as pure PCM".

DIYHiFi.org • View topic - Understanding DSD

---

I listened Dr. Dre headphones and without strong words, I don't wan't more...

[Miska]

That is interesting (in historical sense). I don't know any Pio player with Mediatek and 24/96 DAC. I have still first Pioneer DV-575 with SACD conversion (24/88) via Mediatek MT1389EE, DAC was PCM1742KE - cheapest 24/192 DAC from Ti.

 

I think it was the same one, this looks familiar:

 

I don't now remember the DAC chip model and don't have my photos at hand. PCM1742 doesn't have DSD support so it cannot play SACD natively, and really horribly bad digital filter:

4x/8x Oversampling Digital Filter:

Stop-Band Attenuation: -55 dB

Pass-Band Ripple: ±0.03 dB

That will spit out more ultrasonic noise with PCM than any DSD DAC built to SACD spec.

 

 

I have some inside photos and schematics of the player.

[esldude]

The metaphor is difficult to grasp for the European that I am. Could you phrase it in terms of football, by which I mean soccer?

 

Not the original poster, and don't know soccer well enough.

 

You probably know in baseball a pitcher stands at 18 meters from a batter with a stick. He will use the stick to hit a small palm sized ball. The pitcher throws the ball, and the batter tries to hit it with his wooden bat or stick.

 

Nolan Ryan was known for how fast his fastball was. He simply threw it so fast it was difficult for a human batter to react and hit. Some pitches were around 170 kph. He had a better record in some respects than any pitcher in history.

 

Greg Maddux was almost as successful. He used techniques that made the ball curve and move. Also making it hard to hit with a stick, but more due to movement than speed. However his fastest fast ball was only 140 kph. But combined with movement of even slower pitches he still was highly effective at preventing hits with a stick or bat.

 

So one guy is effective due to speed. Another guy is effective due to movement. So you set up a test to pick the best pitcher. You only test a fastball which gives an edge to Ryan. And you evaluate the result by how the catcher does catching the ball. None of which really are anywhere near the things that matter about which pitcher will be effective. In fact both were among the most effective in history. You then declare a winner with great fanfare. But you really haven't done anything other than make a spectacle of yourself.

[Maldur]

I think it was the same one, this looks familiar:

Yes, that's the same.

in my system this was a transport, I used this analog out very rarely. Limitation to 24/96 is only via optical out - as usual. SACD was for me in that time (2003-2004?) like blue dream - no one don't want to sell that.

I have schematics to, but there is nothing to do with them. Even raw DSD out is not possible, this is possible starting with MT1398EXE, found on DV-610, PD-D6/9 etc.

[Miska]

I have schematics to, but there is nothing to do with them. Even raw DSD out is not possible, this is possible starting with MT1398EXE, found on DV-610, PD-D6/9 etc.

 

"Funny" thing is that these were the family of devices used for the frequently quoted DSD vs DVD-A blind comparison tests...

 

Guess why I don't particularly fancy those results?

[PopPop]

Never mind, I didn't see that elsdude had actually posted ball speed in KPH. Nice!

[Maldur]

"Funny" thing is that these were the family of devices used for the frequently quoted DSD vs DVD-A blind comparison tests...

 

Guess why I don't particularly fancy those results?

 

Yes, I also wondering that... :)

[VandyMan]

I don't trust the judgement, period, of anybody who's never taken acid. Not when it comes right down to it.

 

Those days are behind me mostly, but I don't think there is any better teacher of the power of the mind and the variability of human perception. Plus, its fun.

[Charles Hansen]

You have read the Ayre article incorrectly. The sources are not the same.

 

Huh?

[esldude]

Huh?

 

Mr. Hansen, why are the levels so different?

[Don Hills]

Huh?

 

See post #105.

[Charles Hansen]

Mr. Hansen, why are the levels so different?

 

To be honest, I don't know. With PCM there is a maximum level. For example with Redbook CD there are 2^16 levels so basically from ±32,678. (This is a "bit" off as there is only one "0" and not a plus and minus zero, so there is one more negative LSB than the positive LSB. But you get the idea.)

 

With DSD, it is a form of pulse density modulation. A delta-sigma modulator operates via feedback, and when the level is too high the system becomes unstable and oscillates. The original full scale for 1-bit DSD was ±50% of full scale. Full scale would be plus and minus the supply rail's DC laevels. But thing keep changing with delta-sigma. For years it was said that going past first-order noise shaping was always unstable. Then you would read some AES paper about somebody that got a 2nd-order modulator to be stable. This went on for years until by the time that Sony codified DSD, they used a 7th order (!) noise shaper. But in the Scarlet Book, it allows for brief "overs" above the 0 dBFS of 50% modulation I don't recall what the limit was, but it was a few dB I believe.

 

But if you look a the spec sheet for the ESS DAC chip, it allows for modulation up to ±92.5% or something like that. So that is almost a 6 dB increase there. There is a built-in DSD output on the chip, but it is for DSD-256. Originally we were going to just leave it off because it was more trouble than it was worth.. But the main customer was imagined to be the audiophile that wanted to digitize their LP collection. That group tends to strongly believe that DSD "sounds more analog like" than PCM, so we thought we would add that feature. I assumed that it would be pretty trivial to downsample DSD-256 to DSD-64.

 

But when we contacted Arda, they said that to do it right that we really should start with the 6-bit, 256 Fs modulator output and run it through a 7th-order noise shaper. They sent as a paper on how to do it and we gave it to our FPGA programmer, who does the really hard stuff that we can't do ourselves very easily. It took the entire resources of the Xilinx FPGA, that had ten different minimum phase FIR filters. These are probably the most complex FIR filters ever doen. First of all everybody does things as half-band because that only requires half of the memory coefficients. We did everything as minimum phase which required twice the number of coefficients. Everybody else (except the "VC24" PCM filter in the original Sony SCD-1) does oversampling as a concatenation of 2x filters because it saves about 40% of the resources required (MACs and coefficiens). We did all ten of our filters as single pass, up to 16x oversampling for single data rate, and then we made a second one for preemphasized discs by convolving the deemphasis function with the oversampling function. So there is a a boar-load of processing power going on in the PCM side.

 

It turns out that not only was there not enough room to add the noise-shaping filters for DSD, we had to actually re-boot the chip when changing from PCM to DSD, and releoad he entire program. And even though that is pretty much all that is in there, the DSD algorithm took up more resources (compuiting horsepower) than all ten of the PCM filters (which are the most complex ones done in any commercial product) combined.

 

I really don't understand the math behind the noise shaping. I'm just a dumb analog hardware engineer. The ESS datasheet is light in many areas, so it may be just the way they implemented that part of the chip. I really don't know. When made the needle drop recordings, we left all of the gain setting untouched and levels that came out were different. I can't tell you exactly why. Sorry.

 

But if you have a volume control, its not to difficult to adjust for. And if you have much listening experience, you don't even need to equalize the volume to hear the differences. It's a bit more work to do the "mental offsets", but we all know that it's not like a crappy stereo sounds good if you change the volume up or down. If it were that easy, everybody would be making good sounding stereo equipment. I certainly don't need to level match to within 0.1 dB to hear if something sounds better or not. Do you?

 

Regards,

Charles Hansen

Ayre Acoustics, Inc.

[Charles Hansen]

See post #105.

That's exactly what I mean. Different environmental noise, different mains frequency (assuming a mains powered turntable motor), amount of vinyl reformation between plays. They may or may not add up to an audible difference, but it would have been easy to avoid the difference in the first place.

 

Sorry Don, I just can't take that post seriously. First of all, the turntable has its own 3-phase synchronous power supply with a pur analog oscillator that has distortion of about 0.002%. Second of all, please show me some studies, or even some posts of empirical studies showing that mains frequency variations (which in industrialized countries are well below 0.1% over just about any time scale you care to pick.

 

Next do the same with vinyl deformation and reformation studies. This is like a bad joke.

 

On the other hand finding two units that sounded exactly the same, or even two sets of cables that sounded exactly the same is a Sisyphean task. It takes months of play to break equipment in fully. Are you going to find me two units with the same amount of break-in, the same amount of matching (or lack thereof) between the transistors, the same amount of deviation from the nominal values of the 1% metal film resistors we use, et cetera.

 

All I can say is that I would never purchase any audio equipment that you designed. You can't tell the important criteria from the unimportant criteria when doing comparative listening tests....

 

Regards,

Charles Hansen

Ayre Acousics, Inc.

[Synfreak]

Charles,

 

... I certainly don't need to level match to within 0.1 dB to hear if something sounds better or not. Do you? ...

 

This will be food for some new heated discussion ;-)

 

The problem with comparisons between PCM and DSD playback "at home" is that level matching could get really tricky.

 

The best thing one could do about it - at least in a scenario like we discuss here - is to include some simple calibration files (one in DSD and one in PCM), made with the same recording chain (AtoD and software at least, not necesserally from vinyl).

 

i.e. to supply a 1kHz sine test tone, which was recorded with the same level-settings as your vinyl rips.

 

Played back at home, one can use a fairly basic setup (soundcard build into the PC and software like Audacity ...) to level-match as good as possible, based on the peak of the 1kHz sine. To avoid mismatching due to the exhaustive "DSD noise", just use 44,1kHz as samplingrate during recording the test-tones.

 

Thats like in the old analogue days, with test-tones recorded to the tape alowing to "recalibrate" playback machines.

 

Just an idea, but that might get at least that variable under some - kind of - control ... ;-)

[esldude]

 

But if you have a volume control, its not to difficult to adjust for. And if you have much listening experience, you don't even need to equalize the volume to hear the differences. It's a bit more work to do the "mental offsets", but we all know that it's not like a crappy stereo sounds good if you change the volume up or down. If it were that easy, everybody would be making good sounding stereo equipment. I certainly don't need to level match to within 0.1 dB to hear if something sounds better or not. Do you?

 

Regards,

Charles Hansen

Ayre Acoustics, Inc.

 

Actually yes, I do need .1 db matching to do valid comparisons. And if you claim you don't, then you are simply wrong. Full of it. And yes, I have much, much listening experience. Mental offsets are more illusion.

 

Sorry, but it simply is so. You can claim not to need it, but you are simply incorrect.

 

If you can't figure out why the levels are so different, so sorry, going to be blunt, but you lack enough competence in the issue. Doesn't fill me with a warm fuzzy feeling that you are the guy I should listen to on such matters.

 

If you claim you can mentally offset gross level differences you are now unmasked. No one, not one person, not you, no one can do that effectively. Learn from it.

[esldude]

Sorry Don, I just can't take that post seriously. First of all, the turntable has its own 3-phase synchronous power supply with a pur analog oscillator that has distortion of about 0.002%. Second of all, please show me some studies, or even some posts of empirical studies showing that mains frequency variations (which in industrialized countries are well below 0.1% over just about any time scale you care to pick.

 

Next do the same with vinyl deformation and reformation studies. This is like a bad joke.

 

On the other hand finding two units that sounded exactly the same, or even two sets of cables that sounded exactly the same is a Sisyphean task. It takes months of play to break equipment in fully. Are you going to find me two units with the same amount of break-in, the same amount of matching (or lack thereof) between the transistors, the same amount of deviation from the nominal values of the 1% metal film resistors we use, et cetera.

 

All I can say is that I would never purchase any audio equipment that you designed. You can't tell the important criteria from the unimportant criteria when doing comparative listening tests....

 

Regards,

Charles Hansen

Ayre Acousics, Inc.

 

Find two cables that sound the same.....you call it a Sisyphean task.....the true Sisyphean task is to find two that sound different. You are right it is a bad joke.

 

Can you defend using LP as a source material for picking amongst DSD vs PCM?

[Don Hills]

Sorry Don, I just can't take that post seriously. ...

On the other hand finding two units that sounded exactly the same, or even two sets of cables that sounded exactly the same is a Sisyphean task. ...

 

All I can say is that I would never purchase any audio equipment that you designed. You can't tell the important criteria from the unimportant criteria when doing comparative listening tests...

 

Charles,

you appear to be saying that the unit-to-unit consistency of your DACs is worse than the play-to-play consistency of an LP playback system. Is that really what you meant to say? If so, "All I can say is that I would never purchase any audio equipment that you designed."

[alfe]

Charles,This will be food for some new heated discussion ;-)

 

You mean over heated discussion ¨-)

[Jud]

Charles, here is the problem I see with the level settings:

 

Quite a while ago here at CA, we were given the chance to do blind listening tests between files of the same track (nicely recorded, from BIS) at different resolutions. Many folks, myself included, picked the version that turned out to (deliberately) be 1db louder than the others.

 

When level settings are grossly different, anyone can tell. Whether we can do a "mental offset" I'll leave for discussion by others. But it's when we do the compensation adjustment by ear that I think we can run into difficulties: Do I like A better than B, or do I have the level on A set just 1db (or less) louder than B, and *that's* why I think I like it better?

[Maldur]

I did listening session with my daughter, she is student of two musical school and we both agreed: DSD is better. Even different level of files is not an argument, Ella Fitzgerald sounds better in DSD.

My opinions are like astrotoy and Miska says. That symphony recorded with socks on mic, and I think this is done on original record session - so I don't blame Ayre's products and people's on this.

 

Because technology of both formats is different, we can't obtain exactly same volume level at all, some differences stays. We can only do universal DAC with calibrated amplification for exact matched level.

 

I did PCM conversion to DSD and we listened again. Results are same.

[Le Concombre Masqué]

Actually yes, I do need .1 db matching to do valid comparisons. And if you claim you don't, then you are simply wrong. Full of it. And yes, I have much, much listening experience. Mental offsets are more illusion.

 

Sorry, but it simply is so. You can claim not to need it, but you are simply incorrect.

 

If you can't figure out why the levels are so different, so sorry, going to be blunt, but you lack enough competence in the issue. Doesn't fill me with a warm fuzzy feeling that you are the guy I should listen to on such matters.

 

If you claim you can mentally offset gross level differences you are now unmasked. No one, not one person, not you, no one can do that effectively. Learn from it.

I thanked Charles Hansen for his test led me to reinstate dsd in its full glory in my system, i'll back him there..

 

dsd bears an unbearable lightness of being that makes it different beyond level adjustment.

 

the test showed me how badly Audirvana behaves in converting dsd to PCM and leading SRC + EQing tasks.

 

At the end of the day that means i have to forget about my EQing for dsd files ; that's way beyond .1 dB and my mids and lows are in - and + 3 dB range difference.

 

I think that i.e. for listening to Pixies I might go the pcm route and EQ for I love my 36 to strike. But I dug that for acoustic stuff, i.e. the test exemples, there's a fluidity, a fastness, a liveliness that are unmatched by PCM. I think that people who chose ESL are more appealed by dsd for it's the same nature of difference (and it shows on my huge non ESL Loudspeakers)

 

 

(btw1 Korg with 0 gain, let it clip, + EQing in RX3 is better than A+ that kills and muffles;

 

btw2 RX3 performs a much better job at upsampling and EQing (same parameters of course) than A+

 

btw3 RX3 does not handle flac (nor up samples beyond 192) ; i always claimed i preferred wav over flac in my system/processes ; what happens in A+ when you go from flac 48 to flac 384 ?? here is an added possible cause, next to endianness)

[dallasjustice]

Nolan Ryan won with his arm/legs and badass quiet Texas intimidation. Greg Madux won with his brain and overall versatility. DSD is more like Nolan Ryan and PCM is more like Greg Maddux. :-)

 

Btw, I used to watch Ryan in the Astrodome growing up in the 80s. It was exciting. But Ryan had many losses too. Maddux was more successful in the end.

 

Not the original poster, and don't know soccer well enough.

 

You probably know in baseball a pitcher stands at 18 meters from a batter with a stick. He will use the stick to hit a small palm sized ball. The pitcher throws the ball, and the batter tries to hit it with his wooden bat or stick.

 

Nolan Ryan was known for how fast his fastball was. He simply threw it so fast it was difficult for a human batter to react and hit. Some pitches were around 170 kph. He had a better record in some respects than any pitcher in history.

 

Greg Maddux was almost as successful. He used techniques that made the ball curve and move. Also making it hard to hit with a stick, but more due to movement than speed. However his fastest fast ball was only 140 kph. But combined with movement of even slower pitches he still was highly effective at preventing hits with a stick or bat.

 

So one guy is effective due to speed. Another guy is effective due to movement. So you set up a test to pick the best pitcher. You only test a fastball which gives an edge to Ryan. And you evaluate the result by how the catcher does catching the ball. None of which really are anywhere near the things that matter about which pitcher will be effective. In fact both were among the most effective in history. You then declare a winner with great fanfare. But you really haven't done anything other than make a spectacle of yourself.

[Allan F]

Nolan Ryan won with his arm/legs and badass quiet Texas intimidation. Greg Maddux won with his brain and overall versatility. DSD is more like Nolan Ryan and PCM is more like Greg Maddux. :-)

 

Actually, I believe that both Ryan and Maddux operated in the analog domain.