Why does the computer matter?

[ItemAudio]

Export a whole music library to an external HD. Did a single bit change? No.

Copy it back to an internal HD. Did a single bit change? No.

 

Why? Error correction.

 

Without error correction, the chaos outlined above would be a reality. So what's causing all the errors?

 

Now we're talking about something interesting . . .

 

[1audio]

There is no single resolution for multiple problems. You have simultaneous issues with noise from the power line, noise from each product, coupling from even the galvanic isolation of the ethernet cabling (the transformers have a small but significant capacitance) noise from the AES/SPDIF/USB signals radiated and conducted etc. There is no single silver bullet, just lots of time consuming careful work tweaking each element.

 

The bits are right but the timing gets degraded by everything from noise on the power supplies to data related jitter. That is a function of the line receiver or the USB/Firewire interface and its sensitivity to external stuff. everything is sensitive to external noise etc., the issue is whether the sensitivity is enough to affect the desired performance. That is very much a subjective judgement.

 

Interchassis noise coupling from power and signal connections- everytime you add another device to the chain you will get both signal connections and a power connection. The power connection has some noise leakage that will circulate through all of the other connections. If you can add a real isolation transformer (the type like an Elgar with .005 pF coupling from primary to secondary) to the computer and return the computer's ground to the power filter you may be at the best place you can for it.

 

[Desk Jockey]

Socrates7,

 

PCs are a race to the bottom in cost and pricing. An enormous amount of engineering talent is invested in wringing cost out of PCs verses increasing performance. Don't get me wrong, the platform development teams work diligently to add performance but no sooner than new technology gets to manufacturing area cost cutting begins. Cost reductions activities are across the board from materials, like motherboard materials and design, to the chip sets, memory and controllers, to functional and system testing, all aspects of the design, manufacturing and components are evaluated for cost reduction. If you doubt this for a moment just consider what 500$ can deliver in a PC today.

 

The result of all the cost reductions is that design rules are compromised. The compromises, on their own terms, are not significant, but the composite results of all compromises leaves "grey" areas of performance. Generally, the "grey" areas don't affect one another. However, when a system, in this case a PC, is pushed to the edge of its performance envelope the "grey" areas can overlap and when this occurs the PC behaves in an unpredictable manner. Given the number of possible cost reductions and the possible combinations of how those trade offs can(will) affect each other it is impossible to predict the outcome. Of course the outcome asmmeasured by the overall PC performance is not affected otherwise it would fail system testing. However, you seen expect to see subtle problems areas that you would be concerned about as an audiophile. Of the general areas you have identified, none of those areas are immune to "grey" issues in the other identified area.

 

So how does the PC matter? It matters by how closely the final product meets the design goals of the platform design team. In the commodity price driven world of PCs this is a much tougher challenge than you would imagine.

 

Cheers,

 

Greg

 

[ziggyzack]

Digital signals are not represented by square waves. They are interpreted as square waves. This is what makes them robust. It's the whole point!

 

ItemAudio, error correction schemes come in many guises. There are those which guess values when an error is detected, those which ask for data to be resent, and surely countless other approaches. The error correction should be considered part of the design. There will always be cases in digital signals where the underlying 'analogue' signal is close to the threshold of being interpreted as a 1 or a 0. Nothing particularly interesting about that, unless you think you've just discovered noise!

 

Error correction inevitably takes up some space and time, in exchange for reducing errors to a genuinely negligible level. On a CD-ROM or an audio CD, you'd never state the core capacity of the physical medium. You'd state the capacity including the error correction overhead. Error correction is part of the design.

 

So, ItemAudio, I don't understand your point. Yes, error correction exists. And it works. Robust digital transmission is a reality. What are you trying to say?

 

[audioengr]

Having worked for Intel for almost 2 decades, and in the platform division for some of that time, I can tell you that PC motherboards are poor performers by design. The first design constraint is usually that the motherboard has to be 2 layers. Once you have this constraint, the thing is noisy and beyond help by definition.

 

The reason for this is simply from "batteries and bulbs" taught to children. All digital systems have things in common with simple battery and bulb. There must be a current from the battery to the bulb and then a current back from the bulb to the battery in order for the bulb to illuminate. Where these two cases diverge is the path where the current flows. With high-speed signalling, the return current tries to flow directly under the trace that carries the source current.

 

If the motherboard of the computer is only 2 layers, then the ground-plane is sliced and diced in order to route all of the traces and the power circuits. Therefore, there is not a contiguous ground-plane under hardly ANY of the traces in the motherboard. This causes the return currents to take circuitous paths which are not only away from the optimum path under the trace, but the currents share the return path with lots of other active traces. These return paths do not have zero impedance, so as a result, there are voltage drops and crosstalk between the signals. Since there are only two layers to do the power distribution, signal routing and return paths, the power distribution also suffers.

 

All of these things add up to:

 

1) noise on the power voltages

2) ground-bounce on the return paths

3) crosstalk on the signals

 

All-in-all a very poor scenerio to achieve a low-noise, low jitter audio system.

 

My advice is: Dont even bother. You can throw money at this til the cows come home buying expensive multi-layer motherboards and expensive power supplies etc. Just get an external async Firewire or USB converter or DAC and be done with it. Then you can use any old laptop or desktop as your server and the SQ will be great.

 

Steve N.

Empirical Audio

 

[blueixus]

Steve, your expert comments about the insides of computers, does make me wonder if they really are the best things for music playback given their inherent build flaws.

 

It seems to me NONE of the processing required for music playback should be done inside the computer at all..

 

Would it not be better to have a dedicated player attached to the computer?

 

The computers task would then purely be to send files to it where they are processed and turned into SPIDF or I2S, effectively moving the software player and USB to I2S/SPIDF into one unit, and using the computer purely as a controller.

 

That one unit could be built to more conventional audio specs (decent power supplies etc etc) using an XMOS or ARM chip rather than normal computer build specs you outline.

 

 

[1audio]

Steve:

I think your comments are a little too strong. It really has not been possible to make a 2 layer motherboard that will support a BGA processor and graphics system for a long time (in PC development cycle terms), too many traces. Also meeting FCC and the EU radiation requirements forces manufacturers to make pretty decent layouts, not to mention that the high speed signals like SATA, PCI express and the memory interfaces won't work if they are not well controlled impedances.

 

However, that said, using a motherboard sound chip will be an exercise in frustration. Using a PCI card designed for high performance however can work very well. The current hot property among those with really deep pockets (the Merging Technologies Mikerinos) only runs PCI in an Intel quad core motherboard, not a quiet environment at best.

 

Also, despite the current buzz, USB and Firewire are no guaranty of good results. Execution matters far more than technology.

 

Finally the noise in the PC can propagate to the DAC through Firewire or USB. "Galvanic isolation" really only means no dc connection. There is still AC leakage through the transformers/isolators etc. that must be managed. At the frequencies involved it takes very little coupling to pass wideband noise down the chain. I suspect this is why different USB cables have different sound.

 

 

[The Computer Audiophile]

Hi Demian - I'm actually running my Pyramix machine with a PCI Mykerinos on an Intel D945GSEJT Johnstown Mini-ITX motherboard with a 1.6 GHz Intel Atom N270. Works fabulous :~)

 

[ziggyzack]

you have just invented... the digital stream player, like those offered by Linn, PS Audio, Naim, and recently screwed up by Marantz!

 

[cfmsp]

One approach to the issues you address are the uPnP-based solutions.

 

A more interesting approach was first (to my knowledge) mentioned by John Swenson (who hangs out on AA). He proposes a three piece architecture, in which a purpose built, tiny Linux 'appliance' sits between the computer and DAC to do pretty much what you propose.

 

Google John Swenson here, or preferably on AA, if you want to read up.

 

Others who want to get the computer out of their living room seem okay with putting the music playing functions (of the computer) back into the DAC (see above). Some even believe this is the way of the future! ;0

 

clay

 

[The Computer Audiophile]

We should keep in mind that there is no silver bullet or one solution that's best for everyone. All of the options discussed thus far are wonderful in the right situation.

 

At least nobody has resorted to using the term "killer" yet. For example people were always looking for the iPhone "killer" as if one device was going to put the iPhone out of business. It's the same with computer based audio. No solution is going to put any of the others out of business. In the end they all use a computer one way or the other. It's just a matter of where it resides and what processes reside in each device.

 

[1audio]

Chris:

The same platform as the Auraliti PK100, except we use a more affordable SPDIF/AES interface. The lack of a fan really helps. Thats why we chose it.

 

You must be using a different software package from Pyrimix.

 

[ziggyzack]

Clay,

 

Surely the key point of interest is that the busy operating system of a computer, and associate services, such as storage, user interface, graphics, and so on are kept in the computer, where they need to be.

 

Then you need streaming and control logic, a DAC chip, and an output stage. Whether you decide to keep the DAC chip separate from control logic via a long signal path and an external connection, or integrated and with a very short signal path, is a relatively fine decision. Either way, the DAC chip inevitably has a digital input, so it's impossible to entirely separate digital and analogue elements. The important thing is that the respected hi-fi companies I mentioned know how to do audio, and will isolate as necessary, just as they did for CD players amongst other things.

 

Two boxes or three - who really cares? Just as long as the computer proper is well separated.

 

ZZ

 

[cfmsp]

ZZ, thanks for the response.

 

you will have, hopefully, noticed the emoticon of my previous post.

I'll continue on, in a similar light hearted manner...

 

"Two boxes or three - who really cares? Just as long as the computer proper is well separated."

 

I can agree with this statement, as worded. I keep my computer well separated by a Firewire cable. ;0

 

 

"The important thing is that the respected hi-fi companies I mentioned know how to do audio, and will isolate as necessary, just as they did for CD players amongst other things."

 

If they'd been more successful (at this) with CD players (and transports), we'd not likely have the revolution we have now known as Computer-based Audiophile digital playback. ;0

 

FWIW, I don't take exception to audio companies putting bits of the processing back into the DAC necessarily. I just haven't yet bought into the "need" to get the computer way the hell away from the listening environment. To me that's simply a matter of taste (think WAF-like taste, as opposed to sonic taste), and not nearly the necessity that some would have us believe.

 

As Chris says, both (all?) approaches are valid, and after that, it's down to implementation.

 

I'll personally reconsider if/when I see/hear that a streaming approach offers significantly better sonics for the money.

 

Right now my steup of an opamp-based LIO-8 (with SMPS) fed by a Mac (via 1 meter Firewire) holds its own against an "audiophile-approved" PS Audio PWD / Bridge (at similar price point) with ethernet streaming and discrete analog output stage, etc.

 

Since I'm quite happy with the sound I get currently, I see no reason to consider switching to a different type of setup (with teething pains) and less good/familiar library management.

 

I'll take the devil I know, for now.

 

 

YMMV, it's all good!

 

clay

 

 

 

 

 

 

 

[The Computer Audiophile]

Hi Demian - I'm using Pyramix version 7 on Windows 7 currently. It's crazy what this little board and processor can achieve.

 

[The Computer Audiophile]

"I'll take the devil I know, for now"

 

How right you are Clay! I love that statement. It's fitting for everyone using every piece of gear. It's all devilish :~)

 

[ItemAudio]

Pyramix needs one reserved core to do its thing: two is (just) enough. How do you find the Pyramix interface, though, Chris?!

 

[The Computer Audiophile]

Hi ItemAudio - Pyramix actually doesn't have such high requirements as most people think, at least for playback. My tiny single core Atom processor has no problems.

 

The user interface has done something very unexpected to my listening habits. I now listen just like we all did in the old days. I load an entire album into Pyramix and listen to the whole thing because changing tracks is a pain. It's been quite fun.

 

[ItemAudio]

There's a real desire for computer transports that sound 'as good as' vinyl: as a bonus Pyramix gives you all the inconvenience of playing records!

 

I'm campaigning for a library search facility in the next version of Ovation.

 

[1audio]

Chris-

I'm really impressed with that board. I have Win 7 and J River running on that platform as well. Its much slower than the our Linux solution but it works. Using Pyramix or any editing suite as a playback system is not a real good experience. Editing on that platform also would be difficult. But its a great playback option. Are their Linux drivers for the Merging card? We could combine the card with the player system if there were.

 

 

[The Computer Audiophile]

Hi Demian - Yeah, it's a great board. As far as I know there are no Linux drivers for a Merging Mykerinos card. I've thought about using it in this way as well.

 

[audioengr]

"It seems to me NONE of the processing required for music playback should be done inside the computer at all..

 

Would it not be better to have a dedicated player attached to the computer?"

 

You have a point, however once you move the jitter-sensitive conversion electronics outside the computer, then it is just data handling. Any computer in theory should work perfectly.

 

The reality is that there is CODEC and DSP software involved in the computer, and like most software it is never perfect. This is even the case on Mac. Amarra and Pure Music prove that either the data handling and rounding/dithering is not handled properly in iTunes or the CODECs have flaws, or both.

 

So, even if you build a fortress of a good desktop computer with low noise and good power supplies, it will still have potential software issues.

 

The cheap laptop or Mac Mini does not have errors, so the fortress desktop is a waste of money IMO. In fact, most of the off-the-shelf music servers have issues with ripping software or playback software or both.

 

I prefer to have control over these software pieces: The ability to upgrade and change them.

 

Also, add-in PCI boards have issues as well, such as the clock quality. I have yet to see ANY pro audio card that uses a low-jitter clock oscillator, including RME, Lynxstudio, EMU, Apogee, M-Audio etc.. I have modded most of these or used them with my reclocker. Low jitter clocks would double the price of these products and make them uncompetitive.

 

Steve N.

Empirical Audio

 

 

 

[Wavelength]

Scott,

 

At first I did not believe that it would but really the computer and the OS have such a big influence on the output.

 

Last week we put together a MacPro Six Core, 32G, 3-1T Drives, SSD 512G for a recording studio. It's was a freaken monster. But like the iMac the external noise seems to be very high and this seems to effect everything.

 

We did some tests with my MacBook Pro and found the noise input to the AD dropped a ton. Then I pulled out some new yet to be released Firewire and USB cables and we dropped the noise on the MacPro by 3dB.

 

I have already commented on Windows vs OSX. Anyone with a MAC can bootcamp into Windows and see the difference. OSX is huge, Windows is not. Interesting I have been working with these recording engineers for years and we were looking at different new hardware devices and some of them have specs like 16 channels in/out (OSX) 8 channels in/out (Windows). Very interesting...

 

~~~~~~~~ Jitter come on... remember there is not Jitter in the computer, no jitter on the link (USB or Firewire). The only audio jitter problems happen in the DAC itself.

 

Thanks

Gordon

 

[1audio]

Most pro sound cards are focused on capture, not playback. They are designed to live in environments where a master clock is distributed to various ADC's are locked and send data back to the card. Any card that meets the basic AES/EBU spec for eye pattern response will get the data accurately. Often the data is coming from devices running at different sample rates, which is why Lynx offers a version with sample rate converters on every input.

 

Playback is a secondary part of the functionality. 24 to 55 or more channels of playback are not useful except in specialized mixing applications. Really serious listening environments on the pro world really only exist in mastering rooms, a very small market.

 

Chasing pro cards for home listening will not have a high ROI and there will be some serious limitations on how good they can get. Its not really part of the design spec. Often (RME for instance) jitter is spec'd at 1 nS and that is close to the best possible measurement with the best commercial test systems for the data stream in and out. That number doesn't translate directly into the jitter at the output of the clock recovery circuit on a receiver chip. There are many issues around the nature of the PLL and its lock time and corner frequency.

 

I believe that a design focused on playback only or primarily could sound better since it would have no need for the wide lock range or different operating frequencies needed for pro work.

 

There is more discussion of this elsewhere in these forums.

 

 

 

[Part-Time Audiophile]

Gordon -- from this, it sounds like you prefer the MacBook Pro to a Mac Pro. Is this so? Would you recommend someone to start with the laptop instead of getting one of the new towers?

 

As for the noise -- do you mean that the machine itself is noisy (eg, fan output)? Or that the machine contributed significant noise to the signal?

 

It seems you mean the former -- that the iMac and the MacPro are loud, perhaps too loud for a listening room. Is that fair?