Article: Computer Audiophile Pocket Server - C.A.P.S.

[JeffH]

Thanks for all this great information. This is very educational to us newbies and your willingness to respond to all of the comments and questions makes this one of the best places on the internet. Soon, I will build a server for myself. THANKS!

[hubsand]

Jammrock: 60Gb is fine for MP3s: you squeeze loads of albums on an iPod. <br />

<br />

But seriously, well-heeled audiophiles building a music server are surely not going to hobble its performance with a compressed format (like FLAC), not when such a leap in sound quality comes for free by using AIFF or WAV. The CAPS is built from the ground up to pass 24/192 . . . so let's talk about the space overhead: that level of quality comes at the price of roughly a meg a second, 60Mb a minute, 17 minutes per Gigabyte, maybe 12x single album CDs on a partially occupied 60gb drive -- assuming a streamlined 5Gb OS -- 24 albums per 60gb drive if we deal in 24/96. <br />

<br />

Assuming we're still talking about music lovers, we're in the realm of 200-album collections (minimum!). Even in 16-bit, that's a lot of space. We just need to forget about SSDs until they become economically viable in terabyte capacities. QED.<br />

<br />

The Zotac Ion 230 board has a single Atom chip, passive cooling (no fan) and no DC conversion: only the dual-core 330 needs a fan: the more powerful (and less suitable) 330 board also 'features' DC conversion and wireless: no, no, no. <br />

<br />

The linear PSU + Zotac Ion 230 recipe costs less than $500 (£350 in my part of the world) plus OS and case. The power supply is much better, and (plus a $100 BD-ROM) it does BluRay without a glitch.<br />

<br />

But I am curious about building the system around the Linx card: if in audition it bests the Realtek SPDIF drivers on the Zotac, then maybe it justifies the cost . . .

[The Computer Audiophile]

<i>"But I am curious about building the system around the Linx card: if in audition it bests the Realtek SPDIF drivers on the Zotac, then maybe it justifies the cost . . ."</i><br />

<br />

Hi hubsand - I connected the special cable pictured in my article directly to the S/PDIF headers on the motherboard. I was so displeased with the sound I could barely listen. The Lynx is a great card and it's very versatile for external clocking and dual wire applications.

[hubsand]

. . . as Harry Hill would say: fight! <br />

<br />

The UK price of the Lynx card is scary-expensive (double that Stateside), but we'll import the bits and audition the Lynx with an Atom ITX board. Unless anyone has built one locally and fancies a bake-off?<br />

<br />

There is a board with PCIe, a fanless single core Atom and Nvidia Ion for bluray: see here: http://www.overclock.co.uk/product/Point-of-View-NVIDIA-ION-Intel-Atom-230-Nvidia-MCP7A-ION-Gigabit-LAN-Mini-ITX-Motherboard_24658.html<br />

<br />

This would probably also run the Asus Xonar card, too . . .

[Jammrock]

Can't wait to hear your opinion on the Essence ST[X] with the updated ASIO driver! I've been eyeing that card for a long time.

[Jammrock]

I agree 60 GB is not a lot, thus my 128 GB suggestion, but again it is a moot point if you have a good home network with a media server or NAS with multiple TB's worth of storage. And that's the point made in the article. The CAPS is a great front-end, and that's what it's designed for, not necessarily a media server.<br />

<br />

Audiophiles love their separates after all and that can easily translate to the CA niche.<br />

<br />

FLAC = Free Lossless Audio Codec. Lossless being the key word...besides free. I've done bit comparisons between WAV output and FLAC output and they are identical. There is no reason to shun FLAC because it uses compression when you save about 50% on disk space and end up with identical audio quality. It does use more CPU than WAV but an Atom can handle high res FLAC with ease.<br />

<br />

I am interesting in your ION, Lynx experience. Drop me a PM if you make a write up.

[Audio_ELF]

Just a thought - but Husband menationed two or three times the Zotac Ion motherboards - but IIRC these don't have PCI slots?? There is a ASUS motherboard with Atom / ION chipset with a PCI slot (but has a fan) and another from a company called POV or Point of View (which IS fanless).<br />

<br />

Eloise

[baocrazy]

Hello Chris,<br />

<br />

Great article.<br />

I am too very interested to build a Linux Music Server.<br />

I was wondering if you talked to Alsa people about the impossibility to output the sound automatically (a la Amarra) with the original sample rate. If so, what was their answer?<br />

<br />

Baocrazy

[hubsand]

Only the new Synergy boards from Zotac have PCIe slots. None of their other boards need it for 2-channel audio: they all have SPDIF coax on board. It's a cheap socket, but with a good cable, it sounds superb. <br />

<br />

The POV board may be perfect: it has PCIe, no fan and does BluRay, but the quality isn't as good as the Zotac. Actually, most Atom boards with heatsinks don't need the fan . . . .<br />

<br />

What's twisting my melon right now is the possibility of obtaining better digital throughput using a dedicated card (like the Xonar or Lynx). Zotac's implementation is very lean: there's almost no power overhead, and that enables the system to run pure, clean and streamlined on a 90W linear regulated supply. Powering a motherboard-sized, audiophile-grade soundcard designed for D-A, then not using 90% of it, makes my Occam's Razor twitchy.<br />

<br />

At least the Asus is sensibly priced: presumably the perfect card for this application is the best shielded, but lowest spec version?<br />

<br />

I wonder whether a smarter alternative is simply to retrofit a Zotac board with an isolated WBT coax . . .

[ldolse]

@baocrazy, If the application is configured correctly then ALSA will change sample rates on the fly. The trick is to bypass the mixer, and write directly to the hardware.

[The Computer Audiophile]

Hi hubsand - Are you sure that on-board S/PDIF coax meets the requirements set out in my article? I'm willing to bet it will not handle 24/88.2 or 24/176.4 sample rates without resampling. Can you check for us?

[gvb]

"I was wondering if you talked to Alsa people about the impossibility to output the sound automatically (a la Amarra) with the original sample rate"<br />

<br />

This can be done with ease, but it is not really a feature of ALSA directly, but rather of the driver for the specific card, under ALSA. Thus, you may experience different results according to the card or chipset you are using. <br />

<br />

I can tell you that it can happen as you described with the Intel HD Audio ALSA driver, with the Realtek chipset. It also happens with the ESI Juli@. I didn't have good results with an old Creative Sound Blaster. <br />

<br />

ALSA has a mixer, alsamixer, which is the default route for all sounds in the system. The mixer resamples everything to 48KHz. So the casual user of ALSA will not get bit perfect output. In order to achieve the bit perfect output, you have to bypass the mixer and plug directly to the hardware.<br />

[hubsand]

Chris: Support for 24/192 is spotty, granted. And considerably software-dependent: preferred WASAPI on our streamlined system was fine with 24/192 but not 24/88.1, for instance. However, reaching for 24/192 was less of a priority than standardising on a rock-stable 24/96 platform . . . we're not convinced that 96+ sample rates are relevant for the foreseeable future. <br />

<br />

Getting the domestic industry to 24/96 is already a titanic (but very worthwhile) struggle. I take the view that diminishing returns, compatibility issues, availability of music, and file sizes all count heavily against 24/192 for the next 2-3 years. In fact, for playback I manually resample our 24/88.1 and 24/192 files to 24/96 in Soundbooth Pro. 'Originals' get archived, future-proof. No problem.<br />

<br />

I do admit that our recipe priorities were different: I didn't set a 24/192 requirement. But we did cast in stone that it had to be BluRay capable, and not use a switch mode power supply.

[baocrazy]

Thanks for this information.<br />

<br />

Do you have some url to share on how to realise that with those two soundcards you mentionned?<br />

<br />

[aps]

Chris<br />

<br />

The build doesn't appear to have a DVD drive? I've always thought that this minimalism makes sense (given that I can rip on another computer) but wouldn't it make installation and/or upgrades of the OS much harder?<br />

<br />

Cheers,<br />

APS<br />

[ldolse]

@baocrazy<br />

<br />

Not sure if you were looking for links from me, anyway here is the info for MPD, which is the app I would recommend on Linux:<br />

<br />

Check this thread for a detailed discussion for using plughw and hw for writing directly to the hardware, and how to check the sound subsystem to see that the sample rate is changing:<br />

http://www.computeraudiophile.com/content/Auraliti-Music-Player-and-DAC<br />

<br />

This isn't about sample rate, but how the latest builds of mpd can pad the differently bits for some DACs<br />

http://www.computeraudiophile.com/content/New-mpd-feature-cleaner-signal<br />

<br />

As long as I'm talking about mpd, I did mention in the one thread that WAV and AIFF id3 tags work flawlessly in MPD. I have since found one flaw - using the default libaudiofile sound library mpd uses for WAV/AIFF playback will actually attempt to play the tag, which comes out as a sub-second burst of static. The latest builds in GIT allow the use of libsndfile library, which doesn't have this problem, though it does have an issue with not updating the progress bar in clients, though this new issue is purely cosmetic.<br />

<br />

[Afveep]

Chris-<br />

An excellent article and obviously the result of an immense amount of work. It does beg the question though of the end result (sonics only) vs. a very simple Mac Mini solution with Pure Vinyl, for example. The $1500 target price does not appear to include about $180 for Windows 7, unless I'm mistaken, bringing the total cost of the CAPS solution to about double that of a Mac Mini and PV. <br />

<br />

I can certainly understand the immense value of this solution for those who are not interested in a Mac-based server, have a "bug" for an iTunes-based player, or have the desire, as you state, for more flexibility for an external clock or for use with a DAC that can accept only AES or SPDIF. <br />

<br />

Given that you have considerable listening experience amongst the various servers in your own system, my core question is whether you (or anyone else here) believe that this CAPS server provides a "material" sonic improvement over a mac/Amarra Mini (24/96 only) or PV (24/192), etc. off-the-shelf solution when used with a common DAC (obviously one that could also accept FW or USB). I'm not comparing this to a full Mac pro/Lynx solution, which obviously is not cost competitive.<br />

<br />

Thanks again for a great article!<br />

Tom

[hubsand]

We recently ran several auditions of our $500 dedicated Windows 7 digital transport (of similar spec to the CAPS*) against MacMini and MacBook equivalents, running PureVinyl. The MacBook was audibly less 'artefacted' than the Mini but the only way the MacBook could reach the level of the Windows machine with a moderately good coaxial interconnect was via a very high quality USB cable -- and only then when using a DAC with well implemented USB input.<br />

<br />

As a die-hard Mac user, it really pains me to say that, for the same money, a Windows 7 machine generally sounds better, and runs BluRay.<br />

<br />

Chris' version uses the (very) expensive Lynx card and drivers, which in principle offers even higher quality digital-out, and the possibility of entering the DAC via AES/EBU, which is almost always a definite plus. We'll be building this, and the Xonar/ASIO based version mentioned above, and running some comparative auditions in the next few weeks, but as things stand, Windows or Linux digital transports have some persuasive advantages, unless you're committed to end-to-end excellence in USB or FireWire.<br />

<br />

* No Lynx card, but BluRay capable and with a linear regulated PSU.

[Andrew S.]

Chris<br />

Very well researched, written and comprehensive write up. Thanks for the obvious effort you put in. Pretty thankless task putting your head above the parapets.<br />

<br />

IME Linux & MPD is perfect with USB Dacs (partic async) and not alot else - it's just too much of a PITA to configure anything requiring a driver to achieve automatic sample rate switching - RME and Lynx have both defeated me..life's too short. Having said that, along with Amarra, MPD has given me the best playback sound I have experienced from a computer digital transport.<br />

<br />

I run Win7 64 on my current Lappie - Asus UL30, SSD, M2Tech HiFace, J River 14 and Wasapi out: no problem at all with latest (beta?) 64 Hiface driver from Marco. No stutters or distortion. Very nice sound. <br />

<br />

The Hiface (esp BNC) is a nice alternative to the Lynx esp for lappie users or those on a budget:). Not as good for sure but a whole lot less expensive and works really well.... if Marco would just release linux drivers for it I would be a very happy little vegemite.<br />

<br />

Cheers<br />

A

[Encore]

Very nice setup, Chris, but i'm wondering about whether PoE is good enough. As I have written elsewhere, I'm have a strong suspicion that when a good power supply vastly improves the sound from the Sq.box, even when only the digital output is used, then the power supply of a music server is also likely to have an important influence. Something that forum members are also speculating on in the thread about the G5 vs. the Mini. <br />

<br />

As I understood your article, you also have the option to use a "real" power supply. Have you tried comparing it with PoE?

[The Computer Audiophile]

Hi Encore - PoE was only an option for the Alix PC Engines board I used on another server. The CAPS server requires power from the external supply or a more traditional internal computer power supply.

[ldolse]

It seems to me that the power supply is the one component of this system that could use an upgrade. I've been wondering if the aftermarket squeezebox power supplies would be compatible with the Intel or Alix boards. I can't find enough info to confirm 100%, but I suspect the answer is yes. I was looking at Welborne Labs:<br />

http://welbornelabs.com/squeeze.htm<br />

<br />

Never heard back from them when I emailed them inquiring about the same though.

[ahp]

Chris,<br />

<br />

I'm thinking about building my own solution based on your great article. Since I'll be using an Ayre QB-9 I won't need the sound card, making the system much cheaper to build. But I'm also thinking about using Intel D410PT motherboard since it can support 4Gbyte memory and 64bit operation. This would generate 10W heat instead of the 2.5W generated by the D945GSEJT you are using. Do you see any compatibility or heat issue with the D410PT?<br />

<br />

Thanks for the answer in advance.

[EuroChamp]

Hey ahp,<br />

<br />

you will not need the 4GB, nor will it be an advantage, when you use 64-bit. You should prefer the 32-bit version. On the other hand, the mainboard of your choice does only support 100MBit networkspeed. No problem when you play 16/44.1 files, but the 24/192 files are much bigger. You will get longer interrupts, when you load the tracks over the network.<br />

<br />

Bernhard

[iamimdoc]

For those of us who have never built a computer (but have replaced a few parts, added boards)<br />

<br />

1) Could a Compact Flash Card be used instead of a SS drive? The guy at Frys's though it could and would be cheaper<br />

<br />

2) Could the sound card be left out entirely if one were satisfied with USB DAC sound for now? <br />

<br />

3) Can you explain how one gets the operating system onto the drive (without a CD/DVD drive)? Or the music player software? Or the drivers for my NAS drives (Iocell and Maxtor drives)?<br />

<br />

Sorry if this is readily apparent to others...