Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[JimH]

I posted this over at computeraudiophile.com today, so I thought I should also post it here.  It received a warm welcome there.

-----

I'm posting this, knowing that it will draw fire from the "bits are not bits" camp, but I think the subject is worth the trouble.

(For those who don't know me, I'm Jim Hillegass of JRiver. We publish a media player.)

If you read these forums, you may come to the conclusion that audiophile perfection requires a life devoted to the task. It isn't true. It's easy to do. Here's how:

1. Use a reliable OS (Windows 7 or 8, a new version of OSX, or Linux). Do not use Windows Server unless you're a server guru. Do not use an old OS like XP. They are now unsupported (unprotected).

2. Leave the OS alone. Don't tweak it unless you have a problem. Tweaks often cause problems.

3. Be careful not to overprotect your computer. On Windows, for example, use the Windows security software, Microsoft Security Essentials. Do use the Windows firewall. Don't add other "special" security tools.

4. Get a reliable software player, one that actually does handle the files in a bitperfect way, not one that manipulates the files without telling you what it's doing.

5. Start with easy tasks and enjoy the audio. Don't modify the default settings right away. JRiver has an Audio Setup topic on our wiki. Other players have similar instructions. Follow them.

6. Use lossless formats when you rip. WAV files or AIFF files are no better than FLAC or ALAC, and their tagging isn't as good.

A few more comments...

You will see people recommending all kinds of esoteric solutions to tuning your hardware. Most are unnecessary. A common misunderstanding is that computer processes somehow affect the timing of the PC and that can affect the timing of the DAC. This is silly. PC's are normally loafing along using a small percentage of their CPU power. And the computer has a completely separate clock from that of the DAC. They have independent timing mechanisms, and the DAC has a buffer of its own, so it draws the bits from it when it wants to.

To suggest that the timing of a DAC depends on the timing of a computer is like suggesting that the timing of flights arriving at one airport depend on the timing of their departures at another airport. They are only related if the second airport is closed for a while.

There is no such thing as improving the timing of the delivery of the stream of bits to your DAC. The DAC doesn't care how they arrive. It only cares that its buffer has bits when it needs them.

You will similarly see people say that separating the reading and processing of files from the playback of the files by using two machines is helpful. This is also silly. Both machines perform similar tasks and neither will work if it is overly simplified.

I would be the last person to say that all computers will work well. There are many many reasons they may not, but most people won't usually see such problems, and most of the problems are solvable. Don't fix what isn't broken.

I love to see people deriding others for claiming that "bits are bits". It's like saying that up and down are unrelated to gravity. A bit is a bit. It is either a 1 or a 0. If your computer didn't get this right, you would never be able to use it. You can't improve bits.

This forum is an extremely useful resource if you can find the truth, but there are a lot of less than truthful things being said here as if they were true. Unfortunately, some of the proponents of the great myths about computer audio are the most vocal and insistent about their points of view. Relentlessly repeating a claim doesn't make it true.

Most of all, enjoy the music.

[kstuart]

I agree with 99% of this, including all of the recommendations.

However, as someone who is trained professionally as both a software engineer and a recording engineer, you are missing some concepts:

* PCs are not designed to be real-time playback devices.

* PCs have electrical issues that prevent them from perfectly implementing the digital software theory. (As explained by John Swenson in the three articles posted here a few months ago.)

A variety of professional DAC designers and studio engineers (including some involved with mastering your favorite CDs) have verified the above, some in posts in Computer Audiophile.

Many agree with me that JRiver MediaCenter sounds better than the other players which you insist sound identical...

[csimon]

You are not wrong Jim.

And this is probably the truest thing:

Most of all, enjoy the music.

[dtc]

There are several issues other than being bit perfect and timing (jitter) that are often discussed when talking about computer audio quality. They include electrical noise on both the data lines and the ground and RF noise emitting from the PC. As kstuart points out, there are many reputable people in the business who believe these issues are real and can cause degradation in the sound quality. The use of different operating systems and optimizing the OS are attempts to address this noise issue. In addition, the comparison of wav versus flac files is based not on which is bit perfect, but rather on the idea that the more the CPU and disks work, the higher the potential for electrical and RF noise. The use of a high powered server and a low power player system which is electrically isolated from the server is also aimed at reducing the noise in the system.

The importance of electric and RF noise  is certainly not universally agreed on, but many DAC are taking it seriously and designing their DACs to minimize the effects. Galvanic isolation has been pretty much a standard in DAC design, although doing it on the ground can be problematic from a electrical standards perspective. There are similar efforts to reduce noise on the PC itself. Chris Connaker's CAPS projects are aimed at building a super quiet system, both sonically and electrically. J River is the sponsor of Chris's latest effort. There are other commercially available systems that take a similar approach, like Baetis, which actually uses J River software.

Bits are bits and asynchronous DACs are important issues, to be sure. But any current discussion of computer audio quality should also discuss this noise issue. It is certainly a secondary effect, but many think it is real and it should be discussed.

[AndyU]

Exasound have published some measurements here of the effects of galvanically isolating the pc from the dac. A 30db lower noise-floor with galvanic isolation. Rob Watts, the designer of the Chord Hugo DAC, which has no galvanic isolation, advocates using a USB isolator to avoid noise floor modulation, and describes the effects of so doing here. I am sure that most of the arcane tweakery advocated by some self-appointed computer audiophile "experts" is nonsense, but there's more to getting a world-class sound from your pc than getting the right bits to your DAC. In a sense the whole paradigm is silly; a pc is an horrifically noisy device electrically yet you connect it to a DAC where the designers are shooting for -160dB noise floors; USB is not a very smart way to get data to a DAC. And why should you not be able to use different security software as recommended in the o/p? Long term I am sure that network streamers - ie DACs that connect to your network via ethernet or wifi - will win out.

[JimH]

... there's more to getting a world-class sound from your pc than getting the right bits to your DAC.

I agree, but that's a hardware issue, and mainly beyond the PC.  A good DAC should take care of any problems.

[Vincent Kars]

Personally I’m not very interested in words like “audiophile”
It is reasonable non-descript and its meaning varies strongly from person to person.
From forum to forum as well

I do think technical excellence is a far more relevant criterion.
Hence all things DSP are measured and these measurements are published so we can see what it does.

BTW: nothing wrong with curing problems at the source.

[astromo]

   they're up to over 140 posts over there at CA and that's after only a day or so. Jim, did you poke the bear or a hornet's nest?

I noted a sizable proportion of responders agreeing with your premise, in the main. So, good to see it's positive.

You may want to sticky or Wiki this piece of advice.. 

[Mitchco]

One can easily measure any computer generated noise in the audio data stream by downloading freeware like the excellent Right Mark Audio Analyzer: http://audio.rightmark.org/download.shtml and user guide: http://audio.rightmark.org/downloads/RMAA%206.0%20User's%20Guide.pdf
 
Here is an old RMAA noise level test when I had a Lynx L22 sound card http://www.lynxstudio.com/product_detail.asp?i=11 (a 10 year old design) inside a basic computer:
 

 
Aside from using the built-in audio device on the mobo, one could argue that this is a worst case for computer audio noise. Meaning the sound card is physically located in the computer, both ADC and DAC are being tested in analog loopback mode, the sound card is a 10 year old design, and the computer is built using typical off the shelf components. (I am currently using a Lynx Hilo http://www.lynxstudio.com/product_detail.asp?i=59 as I needed additional channels for tri-amping my system http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/ ).
 
Turning up the volume level to maximum on my speakers or headphones yields the slightest of hiss.  Relative to listening at normal program levels, any computer generated noise is inaudible to my ears.
 
The question then becomes at what level would computer generated noise become audible when listening to music?  The first step is to test your own audibility threshold level.  One step by step example is: http://www.computeraudiophile.com/content/520-fun-digital-audio-%96-bit-perfect-audibility-testing/  The test files are included, so all one has to do is listen and determine their own audibility threshold.  In my case, relative to normal program listening level, I started not being able to hear changes at below -70 dB.  My sound card’s noise level, in the computer, is -109dBA or greater. So any computer generated noise is well below my audibility threshold while listening to music.
 
With respect to jitter, or any timing issues in general, perhaps this explanation from exaSound will help folks understand this is a non-issue in any modern DAC, using an ASIO driver, is completely decoupled timing wise from the computer:
http://www.exasound.com/Blog/tabid/74/articleType/ArticleView/articleId/131/Should-I-use-JPlay-with-exaSound-DACs.aspx
 
“ASIO is designed for and works only in Asynchronous mode. In Asynchronous mode the DAC is the master device, the computer is a slave device. The precision of playback timing is determined only by the DAC. Imperfections of computer timing are not affecting at all the sound quality. Amongst other things, JPlay claims to improve the timing precision for sound streaming.  Whatever it does, it will not affect the timing precision of the exaSound DACs in any positive or negative way. Our DACs store about half a second or so of sound data in a hardware buffer, and play this data following the beat of its own master clock.  It doesn't matter how accurate the computer is when it fills the buffer with sound data.  What matters is how the DAC chip reads the data from the buffer. The very purpose of the buffer is to decouple the timing of the two sides - computer and DAC chip.  Our master clocks have jitter levels in the range of pico-seconds and femto-seconds. This timing precision cannot be achieved in software, not even with JPlay.”
 
In conclusion, if it is a modern computer, any computer noise in the audio data stream should be a non-issue.  If in doubt, measure the noise level with RMAA and compare to your measured audibility threshold.  In addition, jitter and any timing issues using ASIO, are in the domain of the DAC, not the computer.
 
Want to learn more about Digital Audio?  I would highly recommend putting this book, which is devoid of any manufacturers or vendors special interest, under the Xmas tree:
http://www.amazon.com/Principles-Digital-Audio-Sixth-Edition/dp/0071663460
 
Enjoy the music!
 
Mitch
http://www.computeraudiophile.com/blogs/mitchco/

[Vincent Kars]

ASIO is designed for and works only in Asynchronous mode

I wonder if this is correct.
As far as I know one can use a ASIO driver with SPDIF, AES/EBU or adaptive mode USB.
Nothing async about these protocols.

[JimH]

Mitch,
I think your work is amazing.  I admire your patience.

Jim

[ssands]

There's another thing I think they are missing over there. They mention a concern about noise flipping bits (I.e., there is enough electrical noise that pushes the voltage on a bit from a 0 to 1 (as it couldn't happen the other way around)). Now, somehow they think that is happening enough to keep the soundstage from being "clear" or "more spacious" but it is not enough to create audible glitches in the sound. I'm having a hard time buying into that.

Now, certainly, we know that bits can be changed by noise, gamma rays, etc. But unless it a really noisy environment, with crappy cables/bad connectors, I don't think it is happening at anywhere near often enough for anyone to detect audibly. And, as Jim noted, that is beyond control of the software.

[dtc]

Mitch - the noise issue, as usually discussed, is not so much whether you can hear the noise generated in the computer directly, but what effect that noise, including RF, has on the downstream components, such as USB converters and DACs. Some will claim that the noise is enough to destabilize the clock in the DAC, so jitter does become a problem, even in asynchronous mode.  I am not saying I buy into that idea totally, but many in the industry do and their position comes from listening, not just from noise measurements.  This issue needs to be at least considered  when discussing both the PC and the downstream components. I am pretty sure you are well aware of these issues. They are discussed regularly on other forums. I do not think there is a consensus that noise is irrelevant in computer audio. And Chris's CAPS systems and companies like Baetis clearly believe that noise is an issue in computer audio.

I am not trying to takes sides one way or the other, but want to point out that the issue of noise, although a secondary effect, is considered important by a portion of the industry.

[JimH]

Some will claim that the noise is enough to destabilize the clock in the DAC ...

It's hard to imagine how that could happen.

[JimH]

They mention a concern about noise flipping bits (I.e., there is enough electrical noise that pushes the voltage on a bit from a 0 to 1 ...

If it's possible, that is entirely a hardware problem.  We couldn't do it even if we tried, and no software could prevent it.

[gvanbrunt]

There's another thing I think they are missing over there. They mention a concern about noise flipping bits (I.e., there is enough electrical noise that pushes the voltage on a bit from a 0 to 1 (as it couldn't happen the other way around)). Now, somehow they think that is happening enough to keep the soundstage from being "clear" or "more spacious" but it is not enough to create audible glitches in the sound. I'm having a hard time buying into that.

And you shouldn't. If "bits could be flipped" so easily that it happened enough that you would hear differences in a wave form, what would happen to software running in the computer when it's "bits got flipped"? That is about the silliest argument they have made yet.

As for noise in computers, same thing: The bits are 100% bit perfect all the way to the DAC. There is nothing changing even one bit. If noise could do that, computers wouldn't even work. Computers are a hell of a lot more sensitive to "flipped bits" than the human ear. So it's up to the DAC to reproduce the wave without injecting noise. And if noise is injected, guess what it sounds like: hiss, hum, etc. All the same things you can hear on any system that doesn't handle it properly. It doesn't make it "less spacious", "less open" or any other term they like to throw around. It is injected in the analog part of the signal as well, NOT in the digital part of the DAC. So the effects are the same as non digital.

[mwillems]

Mitch - the noise issue, as usually discussed, is not so much whether you can hear the noise generated in the computer directly, but what effect that noise, including RF, has on the downstream components, such as USB converters and DACs. Some will claim that the noise is enough to destabilize the clock in the DAC, so jitter does become a problem, even in asynchronous mode.  I am not saying I buy into that idea totally, but many in the industry do and their position comes from listening, not just from noise measurements.  This issue needs to be at least considered  when discussing both the PC and the downstream components. I am pretty sure you are well aware of these issues. They are discussed regularly on other forums. I do not think there is a consensus that noise is irrelevant in computer audio. And Chris's CAPS systems and companies like Baetis clearly believe that noise is an issue in computer audio.

I am not trying to takes sides one way or the other, but want to point out that the issue of noise, although a secondary effect, is considered important by a portion of the industry.

Mitch's measurement included the relevant downstream component (the DAC): the audio path in his measurement is software-->driver-->DAC/soundcard-->cable-->ADC/soundcard-->driver-->software.

So, to be clear, the measured noise is the result of placing an older DAC inside of a computer electrically coupled to the motherboard, converting to analog, and then looping the resulting analog signal through a second layer of conversion back to digital where the total noise registered at -109dBA.  

I've measured a dozen or so DACs with rightmark, and some are better than that, some are worse.  Some are even bad enough that they audibly interfere with music (I measured one that had noise at -40dB!).  Noise can be DAC-dependent and the method of coupling a DAC to a computer can definitely make a difference in the noise level (e.g. switching PCI slots can sometimes make a difference in the overall measured noise level).  

But here's the important part:  the noise I've measured was not meaningfully sensitive to software, the OS, or any other non-hardware/electrical coupling factor.  All hardware and electrical coupling elements being equal, the noise was (with one quasi-exception) more or less invariant.  I tried loading up PC cores, compiling things in the background, you name it.  The noise level remained the same in all but one case.  The one terrible (probably defective) soundcard that measured at -40dB was sort of an exception.  In that case the noise would audibly change pitch or tone based on PC activity, but (and this is an important but) the total volume level of the distortion only changed a few dB at most no matter what was happening in the background, and it wasn't necessarily correlated with PC load (higher activity did not necessarily mean more noise).

So IMO Mitchco's advice is very good advice: measure your DAC, and take the hearing test he linked. If the PC induced distortion from your DAC is above your hearing threshold, don't tweak your PC, buy a new DAC.  You can find stereo DACs with inaudibly low distortion profiles for less than $100.  Software that does it's job successfully (i.e. is bitperfect) is as good as it gets on the front end.  Other PC optimization/manipulation isn't going to make a dime's worth of difference to your distortion levels unless they're unlistenably loud to begin with (the PC may have a negligible electrical effect on the output, but if it doesn't show up in your distortion measurements it's not likely to be a very significant effect).  

Theoretical speculation on sources of distortion is fine, but distortion measurements are relatively easy to take (Rightmark isn't perfect, but if anytthing it tends to overreport distortion). If more people took distortion measurements in a differential manner they might realize how little of the distortion they hear is actually in any way related to PCs or DACs.  Speakers produce orders of magnitude more distortion than everything leading up to them (unless your equipment is broken).  If typical total system distortion were a 10 oz glass of water, a typical system would have nine ounces of speaker/room distortion, three quarters of an ounce of amplifier distortion, and the last quarter ounce of distortion would be the DAC, the PC, cosmic rays, and everything else put together.  

That doesn't mean it's not worth getting a decent DAC and using quality software (to make sure you're not getting more than your quarter ounce of distortion), but if you feel the burning need to tweak one can get a whole lot more done for sound quality by moving your furniture around or tinkering with EQing non-linearities in your speakers than you could ever hope to achieve by disabling windows processes.

[gvanbrunt]

 Speakers produce orders of magnitude more distortion than everything leading up to them (unless your equipment is broken).  If typical total system distortion were a 10 oz glass of water, a typical system would have nine ounces of speaker/room distortion, three quarters of an ounce of amplifier distortion, and the last quarter ounce of distortion would be the DAC, the PC, cosmic rays, and everything else put together.  

And don't forget the biggest source: your ears. There are piles of studies on just how "bad" the auditory system is at being a measuring instrument. That's not it's job. It interpolates things that are not there, filters out whole swaths of frequencies to listen to what it thinks is important, and generally fills in and connects the dots without us even knowing it. Every time I hear someone say how their "vastly superior, golden ears" hear something that isn't measured by an instrument I cringe. Look up the science and you will see just how bad we stack up as measuring devices - even the most trained ears. You may very well hear something that someone else can't - that shows your training and natural gifts. But if you believe you are hearing noise and the measurements tell you it is below human hearing, you a wrong. Period.

As for the speaker/room distortion: I get a kick out of the guys that spend 10k on a DAC and not one dime on room treatment. Like throwing a Nascar engine in a broke down junker. Then they claim how "revealing", and "resolving", etc their rig is. If people spent money proportional to where the distortion is I might even consider some of their claims. The fact is room distortion IS audible. Easily audible even to the untrained ear. So if their ears were so trained, that distortion would be driving them mad and they would be pouring money into fixing it. They don't, so they are full of it. I'm not saying all of these types, just most.

[InflatableMouse]

There's another thing I think they are missing over there. They mention a concern about noise flipping bits (I.e., there is enough electrical noise that pushes the voltage on a bit from a 0 to 1 (as it couldn't happen the other way around)). Now, somehow they think that is happening enough to keep the soundstage from being "clear" or "more spacious" but it is not enough to create audible glitches in the sound. I'm having a hard time buying into that.

As gvanbrunt said, you shouldn't.

Simply put, if in a digital video signal, if a bit flips, one pixel changes color. In an audio signal, if one bit flips, it changes a single sample, either audibly and you'll hear a crackle or inaudibly and you'll hear nothing.

For the soundstage to change, a series of precise changes are needed throughout the entire digital signal. It needs to change the waveform for both channels with precise timing changes. There's a name for that: its called DSP. DSP is complex, requires very smart people to write very complex code with a very deep understanding of acoustics. To attribute this to some fricking noise is an insult to those smart people! No amount of noise or interference, from earth or mars, will ever be able to do that.

The problem I believe with people making such claims is that they apply problems or effects from the analog domain 1:1 to the digital domain, and things simply don't work like that.

If digital interference is a problem, you'll hear shishishuwhishuswushiswhishushiswishushiswushushishishwushiswhishushiswhushushi shishu .... continuously ... no more, no less .

[JimH]

Split CAPS Servers

[drmimosa]

I posted this over at computeraudiophile.com today, so I thought I should also post it here.  It received a warm welcome there.

It sure did. Thanks for posting this, Jim.

[kstuart]

It's hard to imagine how that could happen.

You don't have to imagine anything, you just have to read Jim Swenson's three articles about how, electrically, it happens.

But if one has a fundamentalist ideology, then they won't - because they "know".

[kensn]

It is my surmise that you truly believe in your product and train of thought. I hold that in high esteem, as I do in your product and business ethics.

Ken

[JimH]

It is my surmise that you truly believe in your product and train of thought. I hold that in high esteem, as I do in your product and business ethics.

Ken

Thank you.  We try to do it right.  It doesn't always work out, but we do our best.

[ken-tajalli]

Personally I’m not very interested in words like “audiophile”
It is reasonable non-descript and its meaning varies strongly from person to person.
From forum to forum as well

I do think technical excellence is a far more relevant criterion.
Hence all things DSP are measured and these measurements are published so we can see what it does.

BTW: nothing wrong with curing problems at the source.

+++1 !
I have been a HiFi buff most of my life, but after 40 years or so, I have come to the conclusion that "Audiophile", "Bit-perfect", blah blah are not relevant to good, enjoyable sound.
For example many are swooping up 60's valve (tube) amplifiers, upgrading them and rediscovering that good sound, despite it's high distortions and noise.
Ofcourse good engineering, good hardware are a must, but keeping things "Pure" for the sake of purity is non-sense.
In the old days of Analogue (& early days of digital), it was common belief it is best to keep the audio path clean, valid, because in those days, graphic equalizers, even tone controls, used to degrade the sound in an audible way, simply there was no technology like a modern PC, advanced DAC or JRiver and its 64bit DSP!
From my profile, you can see I have spend thousands on my sound system, but what it does not say, is how I arrived at it - after many many other systems, years of experimenting!
Yet I use the DSP on JRiver extensively (which nullifies the bit-perfect notion), to correct minor recording imperfections, or my listening hardware and environment.
Good sound is what your EARS dictate - not what some guy tells you!
"Enjoy the music" no truer words spoken.