The Absolute Sound (magazine) take on many aspects of computer assisted music reprodu

[fas42]

So let me get this right, Amir. You're suggesting that more chaotic electrical activity in the server is better than less, because the randomness of the interference "masks problems", rather than striving for a situation with absolutely minimal interference generated at all?

Frank

 

[amirm]

Almost. Chaotic traffic creates random jitter which just serves to increase the noise floor. So having a ton of random activity is better than having a few regular ones. This effect is independent of masking. Our digital systems have very nice dynamic range and very low noise floor. Adding a few dbs of noise due to random jitter is harmless. What we want to avoid are spikes that are in the peak sensitivity range of our ears.

 

[DonH50]

Is there a paper with the masking equation somewhere? What is the shape of the skirts?

 

[amirm]

Search for "critical bands" and you will find a ton of papers. Here is a rough graph of what they look like:

In a nutshell, the shape of the masking curve is frequency and level dependent. I have not read through the Wiki but at a quick glance, it also gives useful info: http://en.wikipedia.org/wiki/Critical_bands

 

[DonH50]

Ah, Fletcher's auditory filter bands, duh! Should have picked up on that, thanks!

 

[rbbert]

I still strongly suspect that a dedicated music server has a better potential for good sound than a general purpose computer, even one optimized for music playback. I have no data t support that feeling, though, except my own experience with USB playback vs. a very definitely not optimized PC, and other's posted and written opinions of dedicated music servers like the Sooloos and Bryston.

 

[DonH50]

I think in this case you should go with your feelings. A typical PC has several things against it as a music server:

1. As a general-purpose machine, the OS is not optimized for real-time operations, leading to uncertain lag in data transfer and clocking;
2. It is a very noisy environment, with many non-random clock signals that can couple and induce audible artifacts, plus the mechanical noise of drives, fans and such;
3. A corollary is that a typical PC has many processes going at once (typically time-sliced) -- mine has 54 at the moment and I am not doing anything else -- and these can cause noise and jitter as they tie up system resources;
4. Shielding is rarely good enough for isolation, causing both analog and digital noise (random and correlated); and,
5. System stability can be an issue since some "bad" process can hog resources and cause problems with data transfer to your all-important audio (real-time) applications, and of course the thing can pause if some other app takes over, lock up, or just crash, any and all of which tends to degrade the listening experience.

I am sure there are lots more things, but that's a start... A dedicated music server can be designed to circumvent (or at least greatly reduce the impact of) all those issues.

 

[rbbert]

None of that is actually experimental data showing the audible superiority of the dedicated music server concept or implementation, though.

 

[DonH50]

Hmmm... I have seen test reports showing higher noise and spurs from your average PC sound card, but would have to dig a bit. Sweetwater Sound produced a line of music-oriented workstations at one time (not sure they still do) and a friend got one for his basement studio (a ~$200k addition to his house!) He is another designer and had a pretty substantial body of test results, theirs and his, showing how much better (quiter, lower audio noise/distortion, etc.)the dedicated machine was than his general-purpose Dell desktop. As for system stability, my personal empirical evidence suggests a dedicated server is miles ahead of my desktop...

 

[FrantzM]

Hi

I have not conducted any measurements on my personal music server. A Windows machine, modified along the advices of The Pacific NorthWest Audio Society Music Server thread in this very WBF.. So far, flawless operation but that is not my main point. Mine is that how much of that noise that we actually hear and how bad it is when certain precautions are taken. Surely there must be a threshold at which the noise don't matter, shouldn't it be ?
What I would like to understand is what kind of noise truly affects the digital stream to my DAC. After all music serving is a very light process in the case of my Duo-Core Music Server. If the PC is not dedicated, i-e a general purpose PC that doubles as Word Procesing and Net Surfing machine , I could understand but with a machine dedicated and tuned, even a modest machine...

 

[garylkoh]

We did a quick comparison about 2 weeks ago between a modified Oppo 95 playing from a USB stick and the Pacific NorthWest Audio Society Music Server driving four different DACs, and none of the members in attendance preferred the Oppo. As a point of interest, the Oppo playing from USB stick sounded quite a bit better than playing from a spinning disc of the same material (a Classic Records HDAD). I was rather busy that day when members of the society were in my office horsing around with DACs so I don't have any detailed notes to share.

 

[Ron Party]

Gary, were any of those comparisons blind?

 

[garylkoh]

They were single-blind. I didn't tell the guys which DAC was plugged into which input and let them go at it. I had previously level-matched the DACs to 0.5dB using both 1kHz sine and pink noise. After a couple of hours, some of them had drawn up nice tables and listening impressions on DAC1, DAC2, DAC3, etc. and I told them which DACs were which. However, comparison between the Oppo and the DACs were sighted.

 

[Ron Party]

Thanks, Gary.

 

[NorthStar]

Interesting Gary. :b

 

[Nicholas Bedworth]

You know, my overall take on the article was, firstly, these fellows did a huge amount of work trying to organize everything the way they did, so they deserve a round of applause for the efforts.

However, the computer audio systems they described seemed rather, well, mediocre with respect to the likely S/PDIF phase noise and jitter.

Using an onboard S/PDIF driver and a 25 ft cable seems like a disaster in the making. Why would anyone interested in critical listening use a motherboard-based S/PDIF? The jitter is possibly 1000 ps or more (Atkinson mentioned that a Mac Mini he tested has something like 1300 ps). Motherboard S/PDIF is in the "kazoo" category. And they said that USB sounded worse than that?

Now, on many mid-range DACs, sure, the USB input sounds poorly compared to the S/PDIF, but their S/PDIF configuation seems completely the wrong way to go. To me, the poor sound quality from a high-phase-noise S/PDIF would pretty much eliminate making subtle evaluations. My ears and body go into "tense" mode when the jitter is much over 300 ps. 3 ps is much, much better.

What were they thinking?

To me, computer audio is about file-based music sources. Do all the computer-to-audio signal format conversion outside the computer.

 

[fas42]

Nice comments, Nicholas. Again emphasising that a music playback system is always a sum of its parts; hoping in some miraculous way that some aspects of the setup won't have an impact of the final outcome is futile. Rather like comparing exotic sports cars, where they are all wearing very mediocre, common denominator tyres to ensure there is no bias there ...

Frank

 

[DonH50]

Frantz,

Most of the noise coupling was on the analog outputs, though jitter was higher (random and deterministic). I wish I had all the data he toook; we did see added spurs correlated to some of the internal clocks, and the noise floor was higher, but we determined that most of the damage was done from noise coupled into the analog side of the chain.

HTH - Don

 

[Nicholas Bedworth]

Their use of motherboard-based S/PDIF, in my opinion, pretty much throws everything else they say into question. That's not to say it invalidates it, but rather, we can't be sure at this point... most of what they say may well be exactly correct, but how could one tell given the interface? That's the point.

High-phase-noise S/PDIF will dominate the listening experience, DACs being equal. And I'd use a high-end DAC, such as an MSB, and would do so without an intervening pre-amp. There are way too many moving "mid-fi" pieces in their rig to be certain about their sonic observations.

If they were to reproduce their tests using high-grade USB-S/PDIF transports, it would increase the validity of their results enormously in my view. So these guys are going a great job overall, but they really need to address the "missing link" issue here. They an "A" for concept and methodology, but a "C" for equipment.

And that can be readily corrected with a couple of phone calls.

 

[Nyal Mellor]

One thing that is interesting and not generally discussed is how best to isolate the computer electrically from the rest of the system. Not just in terms of the signal ground connecting the computer to your audio equipment but also the power supply side. Computers have noisy supplies and RFI filters which pollute the live, neutral and safety ground for everything else on the same circuit. Best to isolate them using some form of power filter such as an isolation transformer or passive resistor-inductor-capacitor filter. Ideally the signal ground would be isolated galvanically from the rest of the audio equipment. Make sure your DAC has a digital transformer on its input when using a computer...