When you say you "convert the VBR mp3," do you mean turn it back into a .wav? Hmm. If that is what you meant, one possible explanation may be that the CPU has to do more processing on an mp3 and that somehow causes audible changes--i.e., its "easier" for the computer to play back a .wav? Strange. But, it does take some serious time to create mp3s, so maybe... |
What about something as simple as the Edirol UA-1D? Its a simple USB device with a single S/PDIF coax output. I use mine off my Win XP Pro box, going into a Theta Pro Basic IIIa. Works reliably for my set up, which is playing mp3s through the home office stereo. Hard to go wrong for $80.
I'm curious about the RME and other AES/EBU cards, however. Onhwy61, have you compared any of those to any of the simple USB devices like the Edirol or M-Audio Sonica? |
Rsbeck, the apogee is a fine DAC, from all I've heard. But, I don't think it is in the same league as the Hbrandt's $14K dCS stack. While apogee is forward thinking in incorporating USB into the box, there are very high end DACs that don't--the U24 and M-Audio pieces work well in those applications. You may want to try a Waveterminal for your application. I didn't think the stock Transit/Sonica family sounded that great compared to some other pieces, but they still sounded pretty darn good with decent source material, so I still don't understand why you had bad results. Maybe a bad unit?
As an oddball sidenote, I had my g/f close her eyes and played some short clips from the same song through iTunes. One an uncompressed .wav file ripped with EAC. The other was a derivative mp3 created with LAME using the "alt preset extreme" VBR setting--a *very* high quality mp3 that I've heard many say is overkill. Yet, she was able to clearly ID the .wav as "sounding fuller" and "better"--notwithstanding the fact that she has significant hearing loss. |
Harry, you are trying to do what I do, and the PCM conversion is necessary. I'm using an Edirol UA-1D right now, which has a USB dongle on one end and a connector for coax PCM out on the other. Nothing else, no DAC. Sounds good for background music running into my Theta (I'm using mp3s compressed with -alt preset extreme, not uncompressed AIFF/WAV/AAC). I believe the Edirol is limited to 16/44.1. Since USB is an asynch protocol, I believe there should be no USB induced jitter in such a setup. Not sure what RSBeck's set up was with the M-Audio, but I previously used a Sonica from M-Audio to do USB to PCM conversion (the version I had was not equipped with a DAC) and it sounded fine. |
Going back to basics, there is a bitstream on the CD and the idea is to feed that to a DAC to convert it to analog. PCM just (I believe) means pulse code modulation, which basically means communicating bits between two devices by an agreed upon standard for interpreting voltage pulses on the wire between them. USB is a serial interface, which means among other things (I think) that its buffered, so its a bit more hardware to implement than something translating pulses into TTL signals for digital processing. So, because coax digital is PCM, and USB is specific serial protocol, there is a conversion from a serial protocol to a different electrical format. Hopefully doesn't mean that the actual datastream of 1's and 0's is changing, but when I've got a box that is USB on one side and coax digital on the other, I call that a "converter."
Of course, I could be dead wrong. Its happened before. ; )
AIFF/WAV/MP3 is just a means for storing the bitstream on media. There is all sorts of stuff added to store things--directory records, sector sequence data, etc--even if you aren't doing any compression. But, with compression, you also need software to recover the original bitstream (or something that looks pretty close, if it was lossy compression).
I wonder, based on what you said, whether your M-Audio somehow ran the bitstream through a low end DAC and, if you used the digital out, ran it through an analog to digital converter. Doesn't seem to make much sense to me, but might explain the problems.
Sample rate conversion sounds sketchy to me in terms of rewards... Can't get more samples per second out than there are samples per second in. But, that is what oversampling is and that seems well beyond my technical competence. How does the thing *sound*? I guess the benefits of creating a 24/96 stream out of a 16/44 stream is going to depend heavily on the quality of what is doing the conversion... |
Its actually "Eric," not "Ed." I should probably stick a response in that "where does your userID come from" thread. Every time I buy or sell something on A'gon, I get an email addressed to "Ed." *sigh* |
I've used the M-Audio Sonica and switched to the Edirol b/c it sounded better to me... I've now switched to the Waveterminal U24 because it sounds better than the Edirol, and found out the reason may be b/c the Edirol outputs at 48 kHz, not native 44.1 kHz? Dunno if that is the reason or not. I have not compared the Waveterminal to Audioengr's modded Transits (the Sonica and Transit are pretty similar internally)--I gather the Transits do pass 44.1 kHz and respond well to mods. I and not really competent with a soldering iron, however, so I'll leave that to others... |
I'll try it, but my recollection is that it has no effect. Think it might even be grayed out. |
Rk_t, I think you hit the nail on the head. I'm sure someone could write software (or maybe it exists) to use the PC processor to upsample, but... seems to me that would be pretty processor intensive and subject to the vagaries of whatever other daemons or processes demand CPU time. I'm guessing you are better off matching your Elgar with a dCS Purcell if the upsampling really matters to you... |
I'd be interested in hearing from Audioengr, I'm curious how anyone can make the blanket statement that foobar's upsampling is better than hardware upsampling... Don't get me wrong, I love foobar. But, foobar is a piece of software written for consumer processors. High end companies like dCS and Theta have access to the same body of knowledge used to create the algorithms for foobar, plus the ability to use custom, specific purpose ASICs in hardware decoding that I have to believe can contribute to better upsampling...
That said, I've never run my foobar at 24/96. FBOW, I don't think my Waveterminal will pass 24/96. But, I've been quite happy with audio quality of my Waveterminal into a dCS Purcell/Delius combo... |
Interesting--glad you chimed in. Maybe I'll look at my Waveterminal again to see if it is capable of something better than 16/44 and give it a go.
I'm still a bit puzzled by the "why," however. In my admittedly simplistic view, you need software and hardware. Sure, PC software is pretty flexible, but so is a lot of firmware; I upgrade various bits of it across my network all the time. And, I think most processors these days are built to be upgradeable with software downloads.
I concur that the collective minds of sourceforge may have built a better algorithm with SRC; I just don't know. But, you still have the hardware issue...
I guess I think of a PC like a good mass market "please everyone" car like a Honda Civic. Sure, you can tack on some upgrades, open up the intake and exhaust, put on better rubber, tweak the CPU, and you might end up with a pretty quick car that might even beat something labeled as a sports car, like a BMW coupe. You can go even further, rip out the excess weight in the interior, implement engine upgrades, install nitrous, tweak the suspension, improve the aerodynamics, and maybe end up with a car that will take a real sports car, like a Carrera. But, I just have a hard time seeing one of those street racers take on something built from the ground up, no compromise, as a sports car like the Lotus Elise or a Dodge Viper.
Anyway, that was my thinking. Perhaps computers and cars aren't a good analogy. Anyway, guess if my WT24 will pass 24/96, I'll find out... |
Audioengr-
The Waveterminal does have a driver, but after poking around, I think its limited to 24/96 output tops, and I believe the 96 may, in fact, be an internal upscaler. I think, if I read right, that its 44.1/48 only.
Oh well... It still sounds damn good run through the dCS Purcell and Delius. |
I double checked, I am using the ASIO with foobar... The waveterminal does not upsample, however. |
Hmm... Neither ASIO.dll nor ASIO4ALL.dll shows up if I run a search on *.dll. Yet, the Waveterminal behaves like its ASIO, the "control panel" says "ASIO 2.0" on it prominently, and the website states: "MME & DirectSound Driver, ASIO 2.0, Direct KS for SONAR" under features. Is ASIO something they could have licensed and baked into the ESI24.dll driver? |
Sadly, I gather the Waveterminal is no longer in production. So, if you can find one, its probably an endangered species. I'd just froogle it or google it and see if there is a decent deal out there. |
