|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
74.93.186.177
Recently built a fanless ITX machine and am happy with its simplicity and lack of moving parts. Have tried Windows and Foobar (somewhat difficult to set up with Asio and trying to get it to look good) and Damn Small Linux but it seemed to run slow and the attached media player software didn't seem very audiophile. I am looking for a streamlined Linux operating system that runs a media player with the lowest possible latency. The operating system will be in RAM and the wav. files will be on a USB stick. The output is to a USB dac. Seems like a simple goal but I haven't had much luck so far.
Follow Ups:
Hi there.My 2 cents:
I am running Xubuntu with Ingo Molnar realtime kernel patches!
2.6.19 is running quite stabil.I have a DDDAC USB DAC connected. It works flawless.
I get the Alsa buffer size down to 2 samples without getting XRUNS.
(You can hear it!)XMMS is a good choice. With realtime option switched on it is even better.
The player itself is not really important though. Most of them are quite similar from sound perspectice. They all stream PCM through Alsa and using pretty much the same libraries.
Be careful not to stream through dmix, the ALSA src.
With USB DACs you'll access the port directly anyhow!
No issues with dmix to be expected in this case.For myself I am using now ecasound, which is giving me best results.
I wrote a kind of shell script player, to load tracks into ram, search for files, setting volume asf! Playing from commandline is really great!
The more you get into Unix the more you'll like it!All tracks I play, are played from RAM using tmpfs!
Stay away from small distros. Support is awful! If you're not a super guru you'll spent endless hours on fixing problems, if you manage at all.
The big ones such as Xubuntu deliver all you need to be happy with Linux.Good luck and lots of perseverance!
Cheers
KLS
Very small runs great on old slow hardware with very liittle RAM.It comes with a very basic media player, but runs anything the can run in Gnome.
Xubuntu is a complete GNU/Linux based operating system with an Ubuntu base. It is lighter on system requirements and tends to be more efficient than Ubuntu with GNOME or KDE, since it uses the Xfce Desktop environment, which makes it ideal for old or low-end machines, thin-client networks, or for those who would like to get more performance out of their hardware.
Xubuntu works quite well under VMWare Server. Not for audio use, of course, but for general desktop use.
I originally had stock Ubuntu, and it was sluggish on my Core2Duo using either the VMWare emulated VGA display or the Xming server. Switching to Xubuntu was a major improvement, and using it with Xming and xdcmp is working very well.
The cool thing is that you can 'upgrade' from Ubuntu to Xubuntu simply by adding the desktop packages, which will make the Xfce session available through the display manager.
For a minimalist multi-media distro, try Geexbox.
Here's a list of media-oriented distros:http://linux-sound.org/distro.html
JackLab looks promising, but it's boot-dvd only.
/*Music is subjective. Sound is not.*/
I've had a mini-ITX fanless system running DSL to a USB DAC for a couple years now with excelent results. You mention it runs slow, can you describe that a bit more? What seems to be slow, from what event to what result?What version of DSL are you using? What are you loading it from? (CD, USB stick, hard drive etc) How much memory do you have? Are you running DSL driectly off the storage medium or loading it into memory first? This can make a HUGE difference in speed.
I've spent a lot of time running XMMS on my system and it has worked flawlessly, as long as you turn off the fancy items and EQs etc it works great for playing music.
John, since you've spent considerable time optimizing every aspect of your digital front end--server, usb cable + usb-i2s conversion (and also SB3 i2s feed), power supplies, dac and analog stage--could you comment on how critical the server piece (hardware, CPU utilization, RAM) is in this equation? Does a lean/mean music only server result in an audible difference compared to say someone's laptop or desktop that plays double duty as a home office machine (fan noise aside)? Have you tried this experiment? To what extent should one go (low cpu util, bare bones os and installed sw, ram, etc.) before incremental audible returns diminish quickly in a system chain consisting of server > usb cable > dac with usb-i2s interface?I can't seem to locate a research paper (can't recall the source) that might be 6 or so months old that published measurement data and drew conclusions about RAM, cpu utilization, lack of extraneous s/w on jitter/noise out. Ring a bell anyone?
Quick summary then details:on current system (SB3, I2S out, synchronous reclock to DAC) the computer makes no difference.
When running USB input DAC everything makes a difference.
EMI from computer in close proximity to stereo system is much worse than anything else.
Details:
I have slimserver running on 6 different machines, from old slow laptops to just put together screaming fast machines, running windows and linux, I cannot hear any difference whatsoever on the sound as long as the data stream doesn't get massively interrupted. (unplugging the cable, some program takes the whole machine down with it etc) Wifi or cable does make a difference, but not in sound "quality", my neighborhood has so many wifi networks its tough to get a reliable continuous data stream. Even the SB3's large buffer isn't always enough, I get dropouts and disconnections from slimserver, this doesn't happen when hardwired so thats the way I go.
USB DACs are a completely different story. If the USB receiver does not have a good reclocking PLL after it they are quite sensitive to pretty much anything that goes on in the computer. Some USB DACs like the Hagerman Chime have a good VCXO based secondary PLL that significantly cut down on the sensitivity to interface effects (but they do not completely eliminate it)
For ones that do not have good reclocking just about anything can cause a change in the sound. Whats running on the processor, fragmented files, power supply, how much other junk is running off the computer, the USB cable used, the temperature of the chips inside, the mechanical construction of the case etc. ANY of those things can change or modulate the noise and jitter of the USB data stream which can cause changes in the jitter spectrum of the recovered clock in the DAC. One interesting effect is aciustical feedback (commonly refered to as microphoinics), the sound from the speakers causes vibrations in the computer which causes changes in the jitter of the USB signal which causes changes in the recovered clock. Laptops are quite susceptible to this, that large open screen makes a quite decent acoustical transducer which couples nicely into the relatively light weight processor module. Its rather interesting to hook up a non-reclocked USB DAC to a laptop in the listening room and then try putting the screen at different angles and rotating the laptop at different angles.
Computers designed as "quiet" frequently have various forms of acoustical damping in the case which can help reduce this microphonic effect. Most towers and desktop cases are terrible in this regard, they are made out of thin sheet metal that rings like a bell.
Whats going on in the computer also affects the sound with a regular USB DAC. I can hear a change in sound when I'm moving the mouse! For me the biggest problem is being connected to the net, there are processes on the computer that are constantly receiving and sending packets even if no application on the computer is actively "browsing the web". On my system disconnecting the computer from the netowrk makes a significant improvement in sound when using a regular USB DAC. Don't even think about running games or spreadsheets on the same computer.
Temperature also makes a significant difference, the warmer the computer gets the better it sounds, up to the point where it gets so hot it stops working of course. I've done quite a bit of testing on this and it does correlate quite well. As the chips get hotter the sound gets better. Thus all the compulsive "cool the computer down at all costs" thinking tends to counteract good sound! Fortunately specialty items such as fanless computers tend to run pretty hot so they are good candidates for music playback.
The USB cable also makes a significant difference, I'm a known proponent of the optical USB cable, especially with a decent linear power supply on the DAC end. Length of cable makes a significant difference as does cable "brand" to a smaller degree. The USB cable specs are tightly regulated by the USB specs so there is not a large difference between them. I'm not sure what aspect of cable construction makes a difference but my guess is its either the dialectric quality or the size of the ground wire. The shorter a cable gets the better the sound, but see below. That is why I like the optical cable.
The single biggest effect on sound has nothing to do with jitter etc, its emi emitted by the computer which gets picked up by the stereo system. If the computer is sitting next to the stereo system this is a FAR greater effect than anything mentioned above. Different computers spew out different amounts of EMI so its hard to make absolute pronouncements, but I would not put any computer right next to a stereo system. When using the USB DAC my computer is 25 feet away from the system with the long optical USB cable. If the computer is only a few feet away its audible effects are quite noticeable. Laptops tend to have lower emmisions than other types but even they can be heard. The ultra quite computers are no better than anything else in this regard, I would seriously think twice about putting ANY computer in the rack with your other stereo components. Ultra low EMI computers can certainly be built, but almost everything on the market is shielded just to the point of meeting the FCC regs and no better.
I have not done extensive testing with sound cards so I don't know how these things work with sound cards. I while back I did some comparisons with several sound cards including a Lynx and a couple RME cards and I did not like them as much as my home grown USB DAC. Of course it might have been the EMI from the computer completely swamping the "native" sound quaility of the cards. When running the cards the computer was pretty close to the stereo system, I did not have any long good quality interconnects in order to place the computer on the other side of the room.
SO to sum it up: systems like the squeeze box do a very good job of decoupling you from computer issues, as long as the server can keep up with the data demand, anything will do with whatever else you want going on.A USB DAC with a good reclocker cuts down a lot on sensitivity, but finding the right one is not easy or cheap.
For a USB DAC without reclocker keep the computer as simple as possible with as few things running on it as you can possibly get. An optical USB cable lets you put the computer well away from the rest of the system.
Keep the computer away from the stereo system, period.
Well thats about it for now.
John S.
Thanks John for sharing your experience with us. Always educational. Very helpful as I'm thinking through what type of server to set up and where to place it.When you say "[I] have a very low jitter clock in the external DAC and feed it into the SB instead of its own clock. I then take the I2S signals fed to the SB DAC chip and send those to the external DAC and reclock them with the low jitter clock right before its DAC chips." The clock you feed from the dac into the SB3, is this through the same cable that transmits the i2s signal from the SB3 to the Dac for D-A conversion?
Yes, the clock from the DAC to the SB3 is in the same cable as the I2s signals.
I must admit I am a bit surprised by your concerns about EMI. Can you share with us some findings that support the concern both in terms of typical EMI levels detected and audibility.I am not sure a well built PC would leak any more EMI than any other recent DSP based electronics component, such as AV receiver, DVD/BluRay/HD-DVD player, etc. And we know that some components these days (including Tivo boxes) are essentially PCs running Linux.
Do you have any data you would care to share?
Just thought I'd add that my SB3 generates some kind of EMI on the CAT5 line such that my DSL modem goes nuts and I get ~30% ATM packet errors. Even if I have a router between them. In fact the CAT5 line doesn't even need to actually be plugged into the SB, just within a few inches to cause me grief. Logitech says I should get the unit replaced, and I will try, but I suspect it's a design issue.
I have done some tests, I hooked up a 2 inch piece of wire to the probe on my 0-40MHz spectrum analyzer and moved it around components, wires etc. I get radically different amounts and spectrum of EMI from different devices and in different directions from the same device.Laptops as a group are pretty good EMI wise (at least the ones I tested), but every "quiet PC" I've tried has leaked like a sieve. Thats four separate PCs in either fanles or very quiet fan systems that are designed to be put on racks with other components. Two were commercial systems used as is from the factory and two were home built systems using cases designed for the purpose. I wouldn't put any of those in my rack.
I also did some testing of a couple other devices such as DVD playes, the SB3, power amp etc. These varied quite a bit. The 4 DVD players I tried were not too bad, about on par with the laptops. The expensive CD transport I have was terrible, maybe thats part of why I never liked it very much. Regular analog stereo components also emit EMI but most of that is power supply related, 60, 120, 180 Hz components. Even though the power transformers are burried inside a metal box I could still eaily pick up the fields from them.
On another question of yours the cables do radiate EMI, but it was usually far less than the boxes themselves put out. I can get the same level of sound degradation from those silent PCs without any connections to them other than the power cord.
I also tried el cheapo shielding (ie wrapping in aluminum foil) and it does actually help quite a lot, but its not really very good to the devices due to air flow restrictions. (its also really hard to see a display when its wrapped in aluminum foil, I need to get that formula from Scotty for transparent aluminum!)
I think its rather intrigueing that the mass market "consumer" items seem to have much lower EMI. They probably DO meet FCC class B, but the other items don't.
The tests I did were completely uncalibrated so I have no clue as to where they stood with regards to FCC regs, they were just to get some idea of the relative amounts and spectra of stuff coming out of some of the boxes in my system.
So I tried it out on my setup. I wrapped the analog audio daughter card (containing the ADCs and DACs and analog circuits) completely in aluminium foil, using paper in between the board and the aluminium foil (to prevent any short circuits).Too early to tell whether there is an improvement or not, but the sound is nice :-)
Can you name the four PCs (or types of cases) that you tried? It will be interesting to build a repository of good and bad PC designs for future reference.Do the EMI spectrum change much across models?
My understand (but I'm not sure why) is that there is "good" and "bad" EMI and techniques like spread spectrum can convert "bad" EMI to "good" EMI without necessarily reducing overall level. Kind of like good and bad cholesterol I guess.
I'm surprised that you did not get a lot of EMI from the cables, particularly the ribbon cable you were using for the I2S.
What I read John to say is that the average PC and other non audiophile quality devices--including the stock squeezebox--spew out more EMI than other hifi 2-channel audio components sitting in one's audio rack, and this is a concern if hi fidelity is one's goal...which was the context of the question in my post.key word search "emi" under John returned 4 posts (excerpts below):
"How is the EMI form this device? My mini-ITX system puts out enough EMI that I can't put it next to the rest of the audio omponents (which is fine, I have it next to me, across the room from the audio stuff)."
"BTW I'm running I2S from my SB3 into my DAC using a ribbon cable with standard 3.3V signal levels, but I think I'm going to go with LVDS ovber CAT5, it has lower jitter and much less EMI."
"We also found that the SB3 EMI was causing part of the problem. By wrapping the whole SB3 in aluminum foil the sound significantly improved."
And from the above post: "Different computers spew out different amounts of EMI so its hard to make absolute pronouncements, but I would not put any computer right next to a stereo system. Ultra low EMI computers can certainly be built, but almost everything on the market is shielded just to the point of meeting the FCC regs and no better."
John seems to have gotten better performance out of a SB3 i2s feed relative to his usb efforts, so my conclusion is to take reasonable care in finding ways to reduce emi potential, regardless of device, via distance or reduced transmission (replacing enclosure, etc.).
Your excerpts helped a lot.*** my conclusion is to take reasonable care in finding ways to reduce emi potential, regardless of device, via distance or reduced transmission (replacing enclosure, etc.). ***
I would certainly agree with this. I would also add that it doesn't hurt to look for ways to improve EMC (electromagnetic compatibility) in all your audio equipment.
Older equipment may not have good EMC, but newer equipment should meet FCC standards which means that in theory the sort of EMI generated by an FCC compliant PC (or any other consumer electronics device that is DSP based) should NOT pose a problem. Particularly since newer PCs implement spread spectrum on the high bandwidth busses (generally memory, graphics) which also help to reduce EMI and newer power supplies have active PFC.
I suspect a factor in John's case (and possibly yours as well) is that on both his Mini-ITX PC and SB3 he's transmitting digital audio signals across components.
The problem is that any cable carrying digital signals is a very strong EMI transmitter (and receiver!). It basically acts as an EMI antenna. Ribbon cables are terrible in this respect. CAT5 is not that great either.
In my case, I have gone for the opposite approach, which is eliminate digital cables where possible. In my setup, I try where possible to do D/A conversion in the source components, and carry analog downstream to the speakers. I'm relying on the EMI shielding in each source component and EMC in downstream components to minimise any issues. I also do a lot of power supply filtering, as that can be a good noise transmitter between devices as well.
If I had to carry digital signals (say, from a PC to an external DAC) I would recommend using quad shield coax cable (plus word clock sync, to reduce jitter).
Not sure what DSL version I tried, booted it off the usb stick. Is there a particular version to use and should it load from RAM, disc, usb stick? Also are there different versions of Xmms and do you need to use something like ASIO4all to get it to output to USB dac. Thanks for your experience in this regard.
The reason your pc is running so slow is because you are running it from a memory stick. Mem sticks are very slow compared to ram or hard drive. I also expereinced very slow performance when running Linux from a dvd or mem stick. Heres why, check out the througput specsusb2.0 memorystick 10-25MB/sec
sata150 hard drive 150MB/sec
DDR-400 ram 3.2gig/sec
DDR2-800 ram 6.4gig/sec
This post is made possible by the generous support of people like you and our sponsors: