[Wodgy]

If you use the TUSB3200 or something similar you'd only have to write microcode, not a whole driver.

 

[NKT]

count me in on a good usb-->coax as well. I'm on a powerbook also, if that matters for driver consideration.

 

[Glassman]

Quote:

Originally Posted by Wodgy If you use the TUSB3200 or something similar you'd only have to write microcode, not a whole driver.

just a microcode? I don't think so.. if I want 24bit and 44.1kHz bit perfect, with the DAC as a master asking for data rather then using masterclock generated by internal PLL.. that would require writing drivers.. or at least ASIO driver, it can exist without Windows even knowing what's that USB thing connected to the computer

 

[momerath]

I'm very interested in a high quality usb->coax, no-other-considerations device and would pay the $200 glassman estimates. Just curious, though, why would it cost $200? Also, for me, it would have to work in linux, either by being usb-audio compliant or having drivers. I'd be happy to help with the drivers (windows and linux). I've not written drivers before, but I do know C and have poked around the linux kernel quite a bit.

 

[Glassman]

thanks momerath, you can oversee ASIO SDK, proceed to download here.. also find some information about DirectKS, it should be on microsoft web.. look into it and tell me if you understand

 

[nefilim]

Quote:

Originally Posted by momerath I'm very interested in a high quality usb->coax, no-other-considerations device and would pay the $200 glassman estimates. Just curious, though, why would it cost $200? Also, for me, it would have to work in linux, either by being usb-audio compliant or having drivers. I'd be happy to help with the drivers (windows and linux). I've not written drivers before, but I do know C and have poked around the linux kernel quite a bit.

I'd be interested too (working in Linux) as my Xitel Pro Link is stuck in 48khz

Dunno much about kernel development (java,persistence etc is my domain) but I'd be interested to have a look at least.

Something else I've thought about is looking at one of those Hauppage MVP devices. I see a few people has been able to enable and add a coax SPDIF output to it. Imagine that, an entire embedded linux device with 32mb ram and spdif out for $100. Probably overkill and brings a whole slew of other issues with, remote control over ethernet etc.

 

[AdamZuf]

Quote:

Originally Posted by nefilim I'd be interested too (working in Linux) as my Xitel Pro Link is stuck in 48khz

are you sure about that?

 

[nefilim]

Quote:

Originally Posted by AdamZuf are you sure about that?

yes, unless you can help

http://www4.head-fi.org/forums/showthread.php?t=74877

 

[aos]

I've finished a prototype last weekend that uses PCM2705 as USB interface and passes the digital out to a DIR1703/CS43122 combo. It is bus powered, however it is fully isolated from the rest of the DAC. Measurements via RMAA are just as good as via optical input from a PCDP or iHP-120. I'd recommend using one of those TI's PCM chips with digital out, adding an optical out to ensure that being bus powered will not affect the rest of the chain, and that's it. Coax is also possible, with a bit of circuitry and an isolation transformer. Device like that should cost <$100.

 

[Glassman]

have you tried RMAA in digital output mode? is it bit perfect? does it support 24bit? doesn't it resample?

 

[aos]

There's no such thing as digital output mode in RMAA. I don't know if it resamples, but that'd be VERY unlikely as they're usually artifacts to be seen - you'd need one hell of a filter to get rid of it and except SRC4192 or AD1896, I haven't seen one with such a stopband, certainly not in a driver. I don't know if it is bit perfect, that is not so easy to test (I need a digital recorder, but I do have iHP-120 now so maybe I'll try that). 24 bit would be too much bandwidth for USB 1.1, would it not? Anyhow I am not aware of any legal or technical way to get any music with more than 16 bits, except what you record yourself, so 24 bit is of no consequence to most anyway.

 

[Glassman]

I suppose you have some internal soundcard with digital I/O? bit perfect testing is in fact pretty easy.. take the S/PDIF output from PCM2705 and run it to your soundcard, let it sync from the input clock.. in RMAA, create the testing WAV in 16/44 and 24/44 and play them using foobar in Kernel Streaming mode, record in RMAA, then compare with the reference results, if it's bit perfect there will be no difference between them..

it seems like 44.1 is okay without resampling, that's good, now if it can handle 24bit.. oh and 24bit is really a piece of cace for even USB1, it's 2*24*44.1k ~ 2.2Mbit, USB1 bandwidth is some 12Mbit..

and why 24bit? think of software volume control.. you need the extra bits in such case..

 

[aos]

I just tested a PCDP to iHP-120, then Audacity to trim recorded wav and then EAC to compare wavs - pretty tedious task. One recording had 63 or so missing samples in the middle, the other was good. Unfortunately I do not have digital out to test the bit-perfectity, I'd have to jury-rig something; actually that'd be pretty easy. But all I really wanted to say is that there's a USB DAC thread in the DIY forum, anyone who wants cheap coax / optical out from USB can look there as a start, build one, and see if it works well enough for their purpose. Writing drivers sounds a bit too intense except as a last resort.

I don't like using foobar2000 for everything because most of the time things simply don't work (e.g. kernel streaming, setting bit deptsh for re-sampling, upsampling, whatever). I only use it for playing 24bit files (or normal 16/44) straight-up as I can select the output device without too much trouble.

Anyhow, I suspect if it wasn't bit perfect, that there would be imperfections showing up in RMAA as increase in noise or distortion or something. The only strange thing I noticed was the increased HF roloff - 1dB at 20kHz compared to 0.2-0.3 otherwise - that I don't know what to attribute to - either the USB drivers or the TI USB chip itself. On the other hand that clearly indicates filtering occuring somewhere. I'll try again.

 

[aos]

Actually there just doesn't seem to be anything that works for certain with the standard USB interface - if I use laptop with winxp then it's 16/48 that is (probably) bit perfect - but 16/44 graphs for thd/imd look horrible. On the other hand 16/44 from windows 2k looks very clean, except that there is a hf roloff which reeks of filtering - but the graphs are so clean that the filter must be good. Any other resolution/bit rate and you can see evidence of signal processing all over the measurements. It seems to me that 16/48 is meant to be USB standard and that usually works well, but anything else, including 44kHz is being processed and the quality of processing varies wildly even with different windows versions.

I guess writing your own drivers/microcode will solve these kinds of problems, but even on commercial cards like Sonica I have tons of problems and unstable signal, and >16 bit doesn't ever work.

By the way USB bandwidth may be 12Mbit/s but in any of these systems, be it ethernet, USB or firewire, I'd be on my toes if I were to exceed 10% of stated theoretical bandwidth.

 

[AdamZuf]

aos,
so in order to get a bit perfect output, can i use the one you are talking about?
(edit: probably not)
i considered to buy a cheap DVD to serve as a transport instead of my computer (won't be perfect as well, i know

), but i still prefer using my computer, and it all comes down to the price. what price are we talking about in here?