Pc -> Dac, How ?

[peufeu]

> I also think that are way too many usb dacs already
> available and that they either sound crappy or are
> very expensive, or both.

I'd like to know about the ones you heard... can you name names ?

I've heard some USB dacs too, they also sucked, so I agree with you. It's all in the implementation, though.

[Telstar]

Originally posted by peufeu
> I also think that are way too many usb dacs already
> available and that they either sound crappy or are
> very expensive, or both.

I'd like to know about the ones you heard... can you name names ?

Crap: basically all the ones that convert usb to spdif and all the ones used in windows with asio4all (pile of crap). m-audio drivers (not free) already improve things.

Good ones: wavelength from crimson up, twindac+, empirical Spoiler. And my current one, the ultracheap Valab 1543s-based one, which i'm keeping until i... get what i want

I didnt audition directly every possible model, but well, i think i have heard enough.

I've heard some USB dacs too, they also sucked, so I agree with you. It's all in the implementation, though.

You can also dig a bit into ec-design huge thread to get to know some of the noise design issues that are waiting for you

My point is that for the effort needed to make usb sound good, you could spend the same working with firewire. In the end it may be less time. Or you could license the usb implementation from Gordon or Steve, but bye bye open source then.

[Pjotr]

Originally posted by peufeu
> I also think that are way too many usb dacs already
> available and that they either sound crappy or are
> very expensive, or both.

Have a look at post #14, they sound not crappy at all, although they can be bettered.

What sounds crappy are the USB1 ones with synchronous streaming and where the DAC clock is recovered from the USB stream.

[peufeu]

> Crap: basically all the ones that convert usb to spdif



> and all the ones used in windows

I'll use Linux first, windows later. Drivers can be a problem, bit-perfectness must not be taken for granted, it must be tested... sometimes they do funny things to the audio.

> Good ones: wavelength from crimson up,

Crimson is asynchronous, which is good. Very expensive, though.

> You can also dig a bit into ec-design huge thread
> to get to know some of the noise design issues
> that are waiting for you

Yeah, that thread is scary huge.

> My point is that for the effort needed to make usb
> sound good, you could spend the same working with
> firewire.

Actually, what is the difference wrt noise ? Both are just a fat data pipe, and not isolated. And since FW is a lot harder to implement (months) versus USB (days) the choice is easy for me...

> Or you could license the usb implementation from
> Gordon or Steve, but bye bye open source then.

I don't think they'd be interested, and me neither

> What sounds crappy are the USB1 ones with
> synchronous streaming and where the DAC clock is
> recovered from the USB stream.

Yes. Definitely. The USB problem is twofold : a clock problem (just like SPDIF) which is solved by putting the clock in the DAC and not using synchronous isochronous mode, and a noise/ground problem, which needs some isolators. The clock should not go through the isolators, though.

[Pjotr]

Originally posted by peufeu
The USB problem is twofold : a clock problem (just like SPDIF) which is solved by putting the clock in the DAC and not using synchronous isochronous mode

And that is exactly what EMU did. It has 2 stand alone x-tal oscillators. One for 44.1 kHz and multiples up to 176.4 kHz and one for 48 kHz and multiples up to 192 kHz.

The rest is proper ground layout. A loop through measurement showed all spurs are far below 120 dB. Only after 50 kHz noise is rising due to the noise shaping inherent to 1 bit converters.

[audioengr]

Originally posted by peufeu
Want to join ?

The project will be opensourced as soon as it's ready, I think I'll put it on OpenCores. I would have liked an opensource USB soundcard with good quality, eq, crossover, custom oversampling etc. Since it doesn't exist, I'm building it. Those who want to help are invited

Thanks for the invitation. I'm a manufacturer, so I'm interested in a production version. I'll keep an on eye on your progress. You are definitely an asset to this community. I know your capabilities from reading past posts.

Best regards,
Steve N.

[Eric Juaneda]

Hi everyone,

Originally posted by peufeu
Fortunately I see emerging labels which start to offer FLAC @ 24-96 or 24-192 for a reasonable price... this is the future !

Where could you find FLAC downloads or other high quality audio files?

Linn Record and ...?

Eric

[tritosine]

Peufeu, from the TI patent (4192 chip dont have this feature, but you can turn it on in the 4194 in hardware mode).


"Thus, there is an unmet need for an asynchronous sample rate converter and method that provides precise phase matching, avoids use of phase locked loop circuits the input sample rate and the output sample rate, provides adequate attenuation of images, avoids the need to have its anti-aliasing filter always turned on , provides improved THD+N performance, and/or avoids the need to recalculate the interpolation filter length whenever the input or output sample rate changes.
...
If fsout always is greater than fsin, then there can be no aliasing, so there is never a need for anti-alias filtering in order to produce Audio Out. Therefore, only down-sampler 19 is needed to produce Audio Out, and this has the advantage of avoiding the group delay associated with decimator ..."


also the new Borbely dac is centered around this feature subset.

[audioengr]

What sounds crappy are the USB1 ones with synchronous streaming

You have probably only heard designs based on the PCM270X devices. I would agree about these. However, it is possible to get very close to async performance with the TAS1020 in Sync Adaptive mode. I compared my TAS1020 USB converter implementation to a total reclocker and if the USB converter uses a good USB cable (Ridge-Street or Axis Design), it was almost indistinguishable from the reclocker.

My motivation to do an async design is primarily market-driven and 24/192 capability driven, not performance-driven.

Steve N.

[Telstar]

Originally posted by tritosine
also the new Borbely dac is centered around this feature subset.

I will hear it soon in a filterless implementation and I have been told great things about it. It could likely be the base for what i have in mind.