(Am I correct in believing that coaxial S/PDIF sucks less than optical? I seem to recall reading that optical cables are not to be trusted.)
I need a new sound card (because the C-Media CMI8738 I have sucks). What's a good/cheap card with coaxial S/PDIF output that is well supported by Alsa and/or Fedora Core 3 these days?
Tags: computers, firstperson, lazyweb, music
Current Music: Panacea -- Tron Rmx ♬
optical cables of the typical consumer-grade garden-variety tend to be more fragile than their coaxial brethren, if that's what you mean.
Also one last note here, I've posted 3 responses already to this post...
When you interconnect two digital devices together over coaxial SPDIF, you'll want to use a 75 ohm "digital audio" RCA cable vs a standard RCA cable. Using standard RCA audio cables won't cut it for this job (although some digital devices may work, it's not a good idea).
This is the same type of cable you'd use to connect two video sources together.
...and no, I'm not suggesting you go out and buy 'monster' cables. They make good cables, but they charge too much and their empire is built on hype. You can do just as well with a nice canare cable from Markertek (www.markertek.com) or other video dealer.
As a side note, I have DAT decks that won't clock to external clocks correctly when I use standard RCA cables for digital I/O, but when I use the 75 ohm cables, I get a perfect copy everytime.
I took it for granted that JWZ would already know this, but, yeah, what he said. With some systems you might not notice the difference, but to be safe stick to rca cables with the proper impedance.
And don't listen to the schmucks downthread trying to say that clock jitter is only a problem with datastreams from disk.
As you say, "video" cables are also 75 Ohm, and I've never had a problem using a $10 video cable for my coaxial S/PDIF connections. It is almost certain that a "digital audio cable" of equivalent quality (perhaps the same identical cable in different packaging) will cost more.
Audio cables (that don't say "digital audio") are generally of unspecified impedance and should not be used for video or digital audio.
minor physical difference - plain audio cables are generally untwisted pairs of separately insulated wire; the 75 ohm composite video or digital audio cables are coaxial internally. (my money's with you on the repackaging.)
... and then there's 120 ohm twisted pair for AES/EBU (professional digital audio)...
correction, 110 ohm. 120 is DMX...
You may also want the low-impedance coaxial cable intended for digital video (the stuff they hook up digital satellite with; this stuff looks like normal coaxial cable TV cable instead of RCA jacks on the end). You can run S/PDIF over that (stereo PCM anyway) and get way longer runs than the S/PDIF standard says you're supposed to. In one of my previous houses, I ran it under the house, about 25 feet, and soldered it to standard RCA wall jacks using a big spool of "digital video coaxial cable" from Home Depot. Audio quality seemed fine, though I wasn't using a great card or doing side-by-side signal path comparisons. There's also a couple crazy audiophiles I've seen recommend using this for DIY high-end analog interconnects.
To inject more into this "fun" debate...
One other thing to consider, that I've had to deal with in the past, it the quality of the AC power you are using, and trying to isolate it.
I've had such horrible power and 120AC wiring in my rental unit before, that if I plugged any sort of copper between the computer and the amplifier, it would result in a ground-loop. (they were using different outlets)
In such a case, using optical was a necessity, to keep the units electrically isolated. I'd hope your power situation isn't nearly that bad of a problem, but how knows. This renders most of the optical/coax debate null, though, as any jitter problems are strictly better than a horrible 60Hz buzz...
(aside: it also was a problem with a MIDI run between my v-drum module & amp, and the computer's horrible creative emu10k card. Isn't MIDI supposed to be an OPTO-ISOLATED current loop, because of this very problem?! More evidence of horrible equipment from creative labs, I guess)
Certainly. My stepfather was having a hell of a time with his home theatre setup, living a few hundred feet from a high-power fm transmitter. Most hardware he bought would start picking up the radio station as soon as he wired it up, but only when plugged into each other, none would do it solo. I figured the copper wiring was acting like an antenna and told him to get units with optical I/O, and that cleared it up.
As always, there are many factors to consider.
Try different midi cables and/or power setups between your vdrum, amp, and midi interface. I used to occasionally have problems with the data midi I/O of my nord modular causing all kinds of digital hash in it's output when editing it on the computer. Iirc it was a grounding situation between the computer and the nord, being exacerbated by an improperly shielded midi cable.
Audio quality-wise coax is better than optical. Not to mention that there are a lot of crappy cheap optical cables out there.
That's retarded. Seriously.
Please explain to me the technical reasons why a digital signal over coax is better than a digital signal over optical, considering that they both result in a bit-identical signal being fed to the DAC in all but most extreme situations -- and that in this situations, coaxial is vulnerable to electrical interference while optical is not?
Ahh, this is what everyone thinks!
They hear the word "digital" and immediately assume it's like a computer, in that the connections are a purely digital signal travelling over a medium and not subject to error.
The problem is that SPDIF (optical) and even AES/EBU (XLR connectors and copper) are subject to clock jitter. The reciever doesn't have a way to send data back to the sender to tell it something's wrong with the data and the reciever must recover the clock signal directly from the line.
This introduces jitter which is detectable as harmonic distortion in the signal.
Bob Katz has an excellent write up on this, available here:
As far as overall quality, It goes something like this:
SPDIF < SDIF-2 < AES/EBU < AES/EBU with Seperate word clock cable.
I'm familiar with jitter, but I have never, ever met anyone who can ABX a real-world 'jittered' signal from a clean one. And on a technical level, a any kind of low bandwidth phase-locked looping (which is present on pretty much every optical device) reduces any potential jittering problem to a point where they are mathmatically insignificant.
You might want to go back and read some of the clarifications that Katz posted regarding his original post on Jitter.
In playback, some digital playback mechanisms can actually remove the jitter problem because they are playing back with their own phase-locked clocks, provided there's a storage medium involved.
But, when you're doing live work (such as one A/D to an editing suite or what have you) you'll hear it.
I'll go through and re-read the Katz articles, thanks for the link. Most of my live experience is analog (nightclubs), and most of my optical experience involves storage. It's always good to have new information.
I wish a Google search for "jitter abx" got more hits. I'm increasingly coming to the conclusion that as far as these things go, if it hasn't been demonstrated in an ABX test, it doesn't exist.
Jitter happens because CDs spin inconsistantly, leading to inconsistancies in the data rate.
A crystal or PLL in a computer's sound card does not suffer from this problem.
Therefore, jitter is not a problem, in the context of this post.
And even if it were a blip on the radar, it would happen equally with TOSLINK, coax, and AES/EBU, because they all use the same clock recovery techniques.
You should try thinking sometime. It's good for the soul...
Jitter occurs during the -transmission- of the digital data which is what this discussion is about; Not in the generation of the data which is what you discuss.
What I've provided is valid given the problem space. There are varying degrees of jitter given the different transmission methods as well, which you choose to ignore.
Also, It's nice to discount what I've said because you haven't thought it out, though. Please play again, thanks.
The jitter being talked about here is induced by actual rays of light bouncing around inside a tiny little cable. "Light" is not digital, unless you want to get into a discussion about individual photons and quantum crypto. Which would be really offtopic.
But that's not jitter.
The word "jitter" refers to something inconsistant.
Reflections in an optical cable (er. at least one that isn't being moved about) are almost a definition of consistant behavior.
Reflections happen in any transmission line, be it TOSLINK, RG-11/U, 2" Heliax, Category 5 UTP, or single-mode AT&T glass. That they exist does not mean that they constitute jitter.
Reflections are not at all new problem, while jitter (in this context) is essentially an invention of the audio press.
If you folks want to be technical, you might as well get your terminology straight before ranting about it. The audiophile authors so oft quoted are just writers, and rather sensationalist writers at that. And while I do believe that they generally describe what they're hearing rather well, I find that the basis for their conclusions generally disagrees with what my oscilloscope shows me about reality.
This, in turn, causes their quasi-technical explanations to be most useful to me as a medium with which to wipe my ass.
Mpingo dots, anyone?
By reflections I was _not_ referring to the same sort of bounceback you get in a EM wave-guide cabling system. I was referring to small variations in the path length caused by the bending of the cable, etc.
Jitter: Small and rapid variations in the timing of a waveform due to noise, changes in component characteristics, supply voltages, imperfect synchronizing circuits, etc. See also DDJ, DCD, and RJ.
Light is at a much higher frequency (and therefore small wavelength) than the EM frequencies used, and is therefore going to be much more suceptible to said variations, especially when you have questionable oscillators/PLLs at either end.
Yes, perfect fibre is better. But fibre is also a lot easier to fuck up, and fucks up badly when it does.
Kind of like how over the air digital seems like a great idea with the perfect picture/sound, then you bump into your antenna and the picture disppears, instead of degrading.
As I said elsewhere, all we need is some damn error control coding and this whole thing can go the fuck away.
Bits is NOT bits. Coax/optical do not result in a bit-identical signal being sent to the DAC. Picosecond timing errors in the digital data can lead to audible degradation in sound quality. This is known as jitter and is a well understood area of digital audio reproduction. Every article I have read concludes that a good coax cable beats a good optical cable and that a bad optical cable is far worse than a bad coax cable. Therefore you are always better off with coax when given the choice.
Granted, the difference may not be audible on jwz's equipment but that is not what question he asked.
Please see above, regarding PLL.
Quote: Picosecond timing errors in the digital data can lead to audible degradation in sound quality.
1ps of jitter in a SPDIF stream of 4 Mbit/s? At a rise-time of 20ns which the signals typically have, you cannot even MEASURE it. These audiophiles never cease to amaze me by their mis-use of technical terms to sell even more overpriced junk. But ok, as long as it's audible (which includes psychological aspects, doesn't it?) it's fine with me.
So to the original question: Coax and Optical should be absolutely equivalent with the only notable difference that optical outputs are guaranteed to be electrically isolated. That way you don't have to worry at all about your computer somehow contributing to ground loops. But if that problem arises sometime you could also just add a small transformer in the "Coax" SPDIF line.
Eh? "Audio quality-wise"? This is a digital signal we're talking about, right? Are optical cables/ports typically so bad that they cause serious data loss?
It's not data loss, it's jitter and error from trying to recover the clock from the line. See above.
it's not "data loss" but timing errors. See my "jitter" comment above.
Thanks for clarifying.
So is there an answer here? Will there ever be a way to make this Just Work without having to be an audio engineer, and not have to worry about shit like jitter/clock problems? Is this an industry problem (ie: the industry isn't interested in making thing work right), or just a very hard engineering problem? (Or something else?)
Clock jitter, as is/was a popular topic in the audio press, happens because CDs are very physical things, and the speed they spin at is somewhat randomly variable. The bits therefore come off of the disc in somewhat randomly variable ways. And these bits of somewhat random timing are transmitted out of the digital output, exactly as they came off of the disc.
On the receiving end, the clock signal needs regenerated so that the DAC knows when to change the state of its parallel transistors in response to the serial data stream. And due to the nature of the specification, it takes some time for this trick to actually happen: The reciever's regenerated clock behaves somewhat like a flywheel, perfectly able to spin at various speeds, but unable to change instantaneously.
When the regenerated clock does not line up perfectly with the incoming data, things are out of sync. The bits are correct, but they're not arriving at the time that they're expected to, somewhat akin to a modem talking to a computer at the wrong baud rate.
This turns up as distortion. I'd call it harmonic distortion, but it is not always the case that it actually is a harmonic of any part of the original signal...
That all said, it's a protocol problem, if anything. But it is only manifested when using unstable playback sources like a CD, across an S/PDIF carrier of some sort, to a seperate analog-to-digital converter.
The issue can be (and is) eliminated by using a stable clock at the sending end.
So, the crystal or PLL clocks of a sound card do not suffer this problem. Dolby Digital and DTS do not suffer this problem. Some CD transports do not suffer this problem, because they buffer a bit of data to smooth things over using a PLL at the final output. (These days, "some" probably includes a large majority of DVD players in all price ranges, given that the feature is absolutely free to implement once you're already paying for all of the other parts of a DVD-playing machine.)
Even CD players themselves, no matter what vintage, when using their own internal DACs and analog outputs instead of their digital outputs, do not suffer this problem. The CD is still unstable, but the clock signal is readily accessible and does not need to be regenerated. The manifestation switches from something resembling harmonic distortion to plain old wow and flutter. And the effects of this problem are so miniscule that thus far, nobody's been able to see a measurable impact. Perfect sound forever.
Jitter, when it is a hinderance (which is quite rare nowadays, and even more rarely noticable) is medium independant, happening equally with all forms of S/PDIF and AES/EBU that do not include seperate wires for a clock signal.
It's generally a miniscule, these days, and has -never- been a showstopper. I am astounded that so much verbiage has been wasted on it here.
Why do you continue to come back to the "Jitter is a problem only brought forth by CDs" statement?
Jitter happens in the transmission of digital audio when the clock is in-band. Not just in the playback of CDs.
In mastering suites, it's reduced by the presence of a master clock (say, the Apogee Big Ben) and BNC word clock going out to all of the slaved components.
I know it's not a show stopper for joe-six-pack who wants to listen to the latest 'rush' cd, but when it's time to master and produce thousands of copies of a track for mass consumption, or a recording of a fine symphony, it becomes a tangible issue.
I just don't understand why this problem wasn't fixed in the protocol layer years ago. It's just dumb that you and I even have to have this discussion in 2006!
Thank God. This really needed to be emphasized further up the thread - there is absolutely no reason to not have built-in error control coding in the transmission standard. I first encountered 'serious' audio hardware partway through my Communications Engineering degree, and couldn't believe that none existed. Unbelievable.
You can find this card for $99 or less, it has coax s/pdif, and is capable of 24-bit/96kHz. Sound quality, it's an order of magnitude ahead of Soundblaster's best.
Seconded. These cards are also well supported by ALSA, which FC3 and beyond are using by default.
Does it have hardware mixing (that is supported by Alsa)? Since apparently my actual problem is that Alsa doesn't supprort software mixing.
Yes, it has hardware mixing. The usb version also has hardware mixing. These are the right cards unless you're really looking for that audiophile cachet. Newegg has prices for the Audiophile and the Audiophile USB.
If you decide to get a card from Creative, I recommend trying to stick with an emu10k based card. Some of the newer cards are using ca0106 chipsets and are simply awful and have delegated many things into software or are not yet implemented under the Alsa (or OSS) drivers.
Btw.. I'll be in San Francisco this week. I'm planning to stop by the lounge and take a look :-) Will I need advance tickets? (Will I even get in?)
I don't think we've got anything coming up any time soon that's likely to sell out.
Clown suits always guarantee entry.
> Clown suits always guarantee entry.
Don't tempt me :)
I also used to use a card based on the 8738. OSS, OSS/Free, and ALSA all had various issues with it. When it worked, it worked well. And when it failed, it did so in a nasty, errorless way which just plain sounded bad.
There are supposedly a variety of pro-quality cards which supposedly have good ALSA support these days. I don't trust it, because I have my doubts about any of them being very well tested. They're not popular enough to get enough eyes to be made to Always Work, and they're too expensive for that to ever change, AND what testing they do encounter is all at the studio, well removed from typical consumer use.
And it stands to reason, then, that the converse is also true: A popular and inexpensive card will, by default, have good support under ALSA.
So you might as well get a Soundblaster Live. They're cheap (I paid $30 for mine 3 years ago) and popular. They typically offer a coaxial digital output on the rear bracket. Support under ALSA is, and has been for some time, flawless (so far as I can tell).
I have used this card for both live sound and broadcast, without a bit of hesitation.
There are exactly two caveats: Some cheaper versions of the card don't have an easily-accessed digital output. Verify first before you fork over cash for it.
And everything (!) is resampled on this card to 48kHz. Some people say this is a bad thing, but it's better this way: No matter what I play with this card, whether grungy 16kHz Realaudio or AC3 DVDs, all I get from the digital output is stereo 48kHz S/PDIF.
This makes my outboard DAC very happy, eliminates a lot of the headaches that were present with the CMI 8738, and presents itself more as a system that Just Works than a collection of disparate parts. There is some measurable quality loss involved here, but remember that it is an EMU10k1: A synth chip, made by a synth company. It's entire purpose, as designed, is to resample things in unoffensive ways, and it (so far) has yet to offend my ears by doing so.
As for your other question: It depends. Typically, coax is considered better in terms of bandwidth and reflection characteristics, and it's cheaper than TOSLINK. You can run coaxial S/PDIF across tens of meters of RG-59 without detriment. But TOSLINK optical offers electrical isolation, is obviously unaffected by EMI/RFI. (Coaxial S/PDIF is supposed to be transformer-isolated at the receiving end, which does provide as much electrical isolation as TOSLINK, but this is seldom actually done.)
In general, with modern components on the recieving end, it probably makes exactly zero difference. Use whichever is more convenient, and stick with good quality cables (eg, real 75-ohm coax, or whatever TOSLINK you can find that has polished ends), and you'll never have any problems. It's just bits. Either they all get there, or some of 'em don't.
i third the recommendation for an EMU10k1-based card- I've owned my SB-Live since Creative bought out EMU- it was one of the first chips ever supported by ALSA IIRC- I can't comment on the current state of ALSA support but as an audio card nothing much can beat it.
The only issue I have with the emu10k1 ALSA support is the incredibly, incredibly non-intuitive mixer controls. The documentation for which leaves much to be desired.
At least I no longer have to put a special parameter in my configuration file to get the right controls to show up for the digital output, but I still have to control the speaker channels separately (I got sucked into the Creative digital mini-DIN connection, to avoid needing an external amp).
On the plus side, linux lets me use the digital out and my headphones simultaneously, whereas the windows drivers make this impossible.
When it comes to hardware, I am a lifetime member of the church of "buy the cheap thing that 90% of the world uses", so yeah, using some random SB sounds like a fine idea to me.
However, I can't find any SB that has S/PDIF that is not, like, $150+. What model am I looking for?
It looks like the cheap-assed $14 SoundBlaster cards can be upgraded to optical S/PDIF with the addition of an add-on "I/O card", but I'm finding that hard to Google for, so I don't even know how much that thing costs! It also sounds like the sort of thing that almost nobody probably owns.
I didn't realize that Creative Labs had fucked up their product lineup so thoroughly.
The model I have (two of) is marked SB0100, which seems to translate to "Sound Blaster Live 5.1" in marketing speak. Froogle says it's still fairly widely available, depending on which model number it is fed. It definately has a coaxial S/PDIF output.
Another option, as if you wanted one, would be the dandy little "Sound Blaster MP3+" box that I picked up at Wal-Mart. External, USB, $50. Optical S/PDIF IO, a few RCA jacks, and a headphone jack with a real level control. Works as well as any other brainless USB audio device, and was essentially as plug-and-play with ALSA as all brainless USB audio devices are supposed to be. I used it for a few days to do a few day-long recordings at an outdoor festival, and was pleased with the results and overall lack of headaches.
Of course, all that wonderful DSP resampling fun I talked about earlier: This box doesn't do any of that, instead expecting it to be done all in software like intel intended. How well ALSA does with that, these days, is anyone's guess. (It used to suck, but it's been a really long time since I've checked.)
You know what's awesome? When the first froogle hit on that says it has RCA S/PDIF, but the photograph of the card clearly does not! Way to inspire confidence in what I'm actually going to receive, dudes...
I think lots of stores are doing this, which may be part of why I was confused the other day.
As you've probably determined, it will have one each of orange, blue, pink, green, and black 1/8" jacks. The orange one, at the top, is the SP/DIF output.
So is the world of years-old computer hardware. (You'd also think that, by now, in this age of cheap bandwidth and inexpensive, amazingly high-resolution cameras, that we'd have more than postage-stamp-sized thumbnails of the stuff we're about to buy.)
just a note to say i love your blog...always so interesting. :)
"I need a...sound card...that is well supported by Alsa"
That's one of those oxymoron jokes, like "Microsoft Works", right?
You have my platonic love there.
Glass Fiber Optic
75 Ohm Coax
Plastic Fiber (toslink)
bogus audio RCA connecting coax
As far as cards, not being a unix person, I don't know what to tell ya.
Froogle is showing prices for "glass toslink" cables ranging from $8 to $80 for the same length! What does a non-crap optical cable actually cost, and how do I identify it?
By glass, I was referring to AT&T ST connector cables, not Toslink. The equipment you purchase dictates what cabled you can use and the ST gear is geeky and expensive. I didn't even realize anyone was making glass toslink (since I left employment in the industry I've spent less and less time keeping up with it...)so I don't know the details on where glass Toslink fits in the list. See this for how I understand the GFP/Plastic thing to be.
The general rule of thumb is that the toslink optical connections are inferior to the coax connections, but the more general rule of thumb is that most people don't notice the difference. Toslink has a bad rap because of the really shitty converters in some equipment. If you find something that seems to work, doesn't cost much and only has toslink, I say ignore the general rule of thumb. You're doing mp3, right? The difference won't be that big. Most of the cheaper cards will suck far worse than their link to your dac.
The best luck I've had recently is with good dacs that have a usb interface. USB is wide enough bandwidth that jitter pretty much ceases to be a problem. The Apogee mini dac seems to be usb audio compliant, sounds pretty good, and actually does 44.1 decently. You'd use a line in on your preamp, but what the hell, it's only a thousand dollars.
You'll probably insist on doing s/pdif to your receiver, though, so you'll ignore the usb dacs. Given that, you'll still want a usb audio device. The advantages here are pretty big. First of all, you don't have to open up a machine. I've seen your machines; I'd pay good money to never see the inside of another one. The other big thing is that an external box won't be exposed to the rf noise inside the pc case. This will really lower the jitter. Basically these things are easier and sound better than PCI sound cards. The problem seems to be that most of them are crap in every way. They either don't do 44.1 (resampling everyting to 48kHz, which is bad), don't work with linux, or cost too much.
The one you want is the M-Audio Audio Audiophile USB.This seems to work well with Linux (and OS X, and FreeBSD). It's consumer grade crap, but it will actually pump out a 16/44.1 pcm stream instead of butchering it to 16/48. It's pretty great as long as you avoid its analog output. The real downside is that it's both ugly and overpriced (failing your "cheap" test). It seems to work as advertised, though, so I don't begrudge them the money. You can get it from NewEgg. Disclaimers, mileage, alsa brain damage, yadda yadda.
Oh yeah, since this wouldn't be the lazyweb unless I offered a non solution... What you want for your mp3 listening joy is a Squeezebox. That has s/pdif and toslink outs. This is indeed the way to go if all you plan to do is listen to your grooviest of mp3s.
M-audio devices tend to work pretty well using ALSA drivers because afaik M-audio actually works with ALSA to get them right.
I would have said 'what I'm using', but that turns out to be a CMI8738 — can you give us a more technical description of the suckage?
I think the emu10k1 based cards suck as the 48kHz resampling they do to everything sounds a bit naff to my ears, but if your problem is that you get no sound on movies and flash that uses retarded sampling rates, then they might solve your problem, and they're cheap enough to buy one just to experiment with. I just told mplayer to do the resampling for me, but that still screws me for midi.
As for cables, every time you hear some audiophile propeller head talk about cables, just cover your ears and say "Cables don't matter! Cables don't matter!" until they go away. They do matter, but they don't matter nearly as much as every other part of the sound system if the do their job properly.
Jitter has been mentioned, but that's caused by clock drift between the sender and the receiver which really only shows up over time and after several hops. It's not a cable thing, and DACs made in the last five years use PLL's to mitigate it. AES/EBU are 'better' due to the use of balanced cables to defeat line noise, but this only really matters over the long runs you find in recording studios. The wire protocol might be a bit better than the consumer stuff, but I don't recall.
I use both optical and coax; one goes into the receiver for Dobly/DTS stuff, and the other goes into a DAC separate. I've swapped the cables around, and all I can tell is the thing that they plug into matters far more than the type of cable used.
But — optical does have the advantage of electrically decoupling the things it connects, which may or may not matter. OTOH it's annoyingly expensive for long (>0.5m) runs.
More than one program can't play audio at the same time. Other machines with the same OS but different cards do not exhibit this behavior. Also, "esdmon" doesn't work. I've encountered this class of problem before, and "get a new card" was the solution back then. Maybe there's another solution, but "get a new card" sounds like the easy way that won't require me to learn a lot of crap I don't care about...
AFAIK, it's because ALSA doesn't have a software mixer for cards that don't support it in hardware. Pretty much any modern Soundblaster card will solve the problem you're having.
ALSA's software mixer is dmix, and can allegedly be used for cards which either have no hardware mixer, or have no ALSA support for their hardware mixer.
Personally, I've never managed to get dmix to work, and just use aRts instead. Still, other people reckon that they've got it to work.
Does aRTs still introduce a ridiculous amount of buffering lag? That turned me right off aRTs, and esound. My current hope is that gstreamer will become the One True Backend...preferably outputting to that polypaudio sound server some people sound excited about.
The lag is still just about noticable, but hardly ridiculous. But then, this is on my work machine which is just used for playing Oggs and MP3s, and random application noises. I certainly don't care about it enough to fight with dmix.
dmix is supposed to "just work" in Alsa 1.0.9, which is supposed to be out real soon now. I haven't tried it, though.
I got dmix to work on my nForce2 motherboard's chipset (i810, I believe) without too much trouble a few days ago, where "work" is defined as "the sound skips every time I focus a new window or hit my browser's back button".
This would appear not to be the case.
Isn't esound supposed to handle the mixing duties in software? I mean, isn't that the reason it exists at all? If you're not able to mix different streams, and esdmon won't run, the magic 8 ball says that esd is roached, not the sound card. How and why esd got itself into that position is beyond me. That's more of an H.P. Lovecraft question than a technical issue.
Central Computer used to sell a nine dollar sound card that worked fine with linux. It sounded like shit, but it worked. They probably still sell it. I wouldn't spend much more than that until I had some degree of confidence that esound wasn't screwed.
In fact, your symptoms sound suspiciously like two copies of esd running on the same computer. This is what happens when you do that with sound hardware that doesn't do its own mixing. One copy of esd runs properly. The other one, the one that all of your software tires to use, runs but can't get the data to the device. You would expect it to die with an error. You would be wrong. The other sign pointing to this is your esdmon problem. Esdmon just dumps a copy of the data going to the audio device. If Esdmon connects to the bad esd (which it will), then it doesn't work. There is no reasonable error as such.
I'm sure that this is all explained in the copious and accurate documentation.
See, I don't entirely doubt that this could be solved in software, with enough pain. But if I can throw $40 worth of hardware at the problem to not have to think about it any more, that's a no brainer.
No, only one copy of esd running. And like I said, this machine is configured identically, software-wise, to other machines that don't have this lossage.
Ah, yes. No hardware mixing. Not a problem for me, because I don't like system sounds or websites interrupting my music with their tinny-arsed midi tunes, but it can be a nuisance.
Defintely go with the hardware fix; esd's software mixing has always been rubbish, artsd's xmms plugin is unreliable, I could never get ALSA's software mixer to work, and I'm simply not nerdcore enough to even try to make gstreamer go.
An SBLive!, as mentioned, will do the trick, but an MAudio card will probably have more brains under the hood. Their Audiophile 24/96 behaved itself under ALSA when I last auditioned one, a couple of years ago.