Pc -> Dac, How ?

[Telstar]

Originally posted by DUC985
i can only agree to Russ, use COMP + CPLAY with XP follow CICs advise on how to setup the PC and buy a ESI Juli Soundcard which gives you I2S without soldering any SMD stuff and connect it to the twisted pears Sabre DAC.

You can even directly drop the Twisted Pears Dac with a 20 pin connector to the Juli.

This sounds better than most transport and DAC combos you can buy.
Just try it, it costs less than most players out there, you will be surprised.

just my two cents.

I dont doubt that sound better than most cdp. Any well optimized computer with a good dac does.

But i did study the huge thread over the asylum and there are cons:
-First cplay has an horrible interface.

edit: didnt notice the juli@ can output i2s, then most noise and jitter issues are fixed this way.

The Juli@ is kinda cheap (around 150$) but limited to 24/96 like 99% of the cheaper external soundcards. I think purfeu goal is 24/192, to be future proof.
The best buy that i have found so far (but not tried yet) is the tc konnect 8, which costs 300eur, its firewire and support 24/192. From there one can go (without bringing along any noise) any dac he prefers.

[DUC985]

agree that some parts a little outdated, but you can get most parts without any problems, at least i could.
The PC is tweaked, if you do not want to go that route, forget PC audio right away. You must have dedicated power supplies for the motherboard and all other connected devices.
I'm not connecting the Sabre through SDIF, i'm not using the analog board of the Juli, you can disconnected and use the headers to steal the I2S signal.
I gave up on SDIF connections quite some time ago.
All you need are three short shielded wires and ground.
This works without any problems with upsampling to 192khz.

The twisted pears DAC has SDIF locking problems with the Juli, i also agree to that.

[analog_sa]

Originally posted by DUC985
You must have dedicated power supplies for the motherboard and all other connected devices.

Dedicated or not ground contamination in the PC terrifies me. Don't know how viable is using optocouplers on the sound card I2S signals. Would the optocouplers require reclocking of the signals in the dac?

If async USB is possible to diy i would rather follow this route.

[tritosine]

optoisolators have their own jitter, maybe thats why accuphase used ISO150 right before multibit DAC's .

[abraxalito]

Originally posted by peufeu
For the volume control I'd like to try some LDR a la "Lightspeed" controlled by the computer (ie the volume slider), so things like ReplayGain work. We'll see if it sounds better than digital volume control...

I've been looking around for GPUs and all the new ones which support double precision have huge fans. OK, so I'll put FIR in the FPGA then.

As my wish list, I'll plump for a lean, clean approach. That would mean only implementing in the FPGA only what really has to go there, giving maximum software flexibility on the PC. Why go to the effort of isolating the DAC from the nasty noisy logic in the PC but then sticking a sweaty number-crunching FPGA right next door? By all means give us a metadata stream for each channel by which we can build an analogue volume control and/or analogue de-emphasis. But with a 38MB/s (from the Cypress DS) data pipe, there's no need to do any crossovers in the FPGA - the available bandwidth potentially gives us around 64 channels of 24/192. Ideally the FPGA and companion chips won't need to be BGA packages to give DIYers a hope of making it themselves.

R

[krautxy]

Hi DUC995,
i use sabre DAC, Juli@ and asus Xonar PCIe. i know that Juli can be used to output I2S.
Can You send me the schematicts of the output header of the Juli for I2S.
I am in frankfurt (germany ) so it will be easer to do this by telefone:
069491892 or 01793954774 or E-mail .
thanks
alfred

[peufeu]

[QUOTE]Originally posted by abraxalito
As my wish list, I'll plump for a lean, clean approach. That would mean only implementing in the FPGA only what really has to go there, giving maximum software flexibility on the PC.

Yep, for instance the device doesn't have a user interface : no LCD (those are VERY noisy), no display. User interface is better on a PC anyway (Amarok...) I'll just put a volume knob and power switch and LED.

> Why go to the effort of isolating the DAC from the
> nasty noisy logic in the PC but then sticking a sweaty
> number-crunching FPGA right next door?

It's not that bad actually. Of course, the digital +3V3 wouldn't be suitable for audio power supplies... but the decoupling is well done and it's a solid 6 layer board.

> By all means give us a metadata stream for each
> channel by which we can build an analogue volume
> control and/or analogue de-emphasis.

That's done actually, metadata stream completely separate from audio, uses a different USB endpoint. I didn't want to mix control and audio data, so since I had two pins left on the FX2 I implemented a simple serial protocol. It could also be used to control the FPGA from a uc (like Atmega) if someone wants to do it.

> But with a 38MB/s (from the Cypress DS) data pipe,

It goes to 45 without complaining... that's the FX2LP version, it's a bit faster.

> there's no need to do any crossovers in the FPGA -
> the available bandwidth potentially gives us around
> 64 channels of 24/192.

I'd like to try though, especially for oversampling, because on the PC this can get really slow. It's well suited to a FPGA : small buffers, fixed point arithmetics. I would like to experiment with the DACs in external filter mode, which needs 1.5 Msps 24 bit.

> Ideally the FPGA and companion chips won't need to
> be BGA packages to give DIYers a hope of making it
> themselves.

BGA is so much better. Anyway, this is an off the shelf board from Digilent, you can get it from them for cheaper than the parts cost in quantity 1. Much simpler.

[Wavelength]

Guys,

Just some clarification...

SYNC USB was primarily used by IN endpoints like adc's as then the PC would have to sync to the ADC's clock.

Peufeu has it right. USB 2.0 async interface supporting 24/192 is the ticket. This is exactly what I'm doing in 2009. Steve N.

Steve, one small problem. Windows and Linux only support USB 1.1. I talked to the USB engineer at Microsoft about 2 weeks ago and they are swamped and don't have any plans for Class 2.0 support.

I know a guy who wrote the driver for USB 2.0 Class but he wants $15k for the driver. Yea thanks will write my own... I've done about 20 of these in the past.

The other problem is that all devices cable of USB 2.0 have limitations in the USB 1.1 Class and cannot exceed endpoint sizes of 512 bytes. Which means it would be hard to make a dac that would support both realms as a interum to Windows/Linux Class 2.0 support.

I also talked to the Apple guys and they can support up too 24/210K right now but they are working on 32/210K USB support which would be really cool.

I still would not count Ethernet out, I just think your nuts to rely on Linux to support the platform. You would be better off doing your own thing.

Which means another driver....

So no matter what at this point a driver is required for any OS other than OSX 10.5.x

Thanks
Gordon

[peufeu]

Amen.

> I know a guy who wrote the driver for USB 2.0 Class but he wants $15k for the driver.

LOL.

> I just think your nuts to rely on Linux to support the platform.

Oh, actually I meant it is easier to write drivers for Linux than Windows... (at least for me). The time for drivers has not come yet, though, so we'll see.

[abraxalito]

>> Why go to the effort of isolating the DAC from the
>> nasty noisy logic in the PC but then sticking a sweaty
>> number-crunching FPGA right next door?

>It's not that bad actually. Of course, the digital +3V3 wouldn't be
>suitable for audio power supplies... but the decoupling is well
>done and it's a solid 6 layer board.

From reading through, I think I've made a few false assumptions about what you're planning. Definitely in my wish list is galvanic isolation - but if you were even considering the digital supply (i.e. that going to the USB interface) as available to the audio then we weren't thinking along the same lines at all. That is unless you were going to develop some kind of isolated PHY for USB (or already know of one that's available?).

A quick googling reveals something interesting in this direction from Icron but I've no idea if its genuine vapourware or not. No full datasheet is available, but they say availability Q1 2009. Even if this doesn't become reality, their standalone products might be a useful way of providing isolation on USB (any extension through ethernet must be via transformers). So perhaps its premature for me to be concerned with building isolation into the DAC itself.

My other misunderstanding was the amount of processing you were contemplating - I imagined a lot since you mentioned the potential of using CUDA. But if you're still using nothing more than the FPGA on the digilent board, that's hardly in the same ballpark as CUDA. Oversampling along the lines that's built into some DAC chips is only a few tens of MMACs so presumably is well suited to a Spartan III.

>> there's no need to do any crossovers in the FPGA -
>> the available bandwidth potentially gives us around
>> 64 channels of 24/192.

>I'd like to try though, especially for oversampling, because on the
>PC this can get really slow. It's well suited to a FPGA : small
>buffers, fixed point arithmetics. I would like to experiment with the
>DACs in external filter mode, which needs 1.5 Msps 24 bit.

Sure, yet CUDA running double precision is a far cry from fixed point on a cheap FPGA. Hence my confusion. By all means please use up the capability of the Spartan III - I'd definitely be interested in playing around with digital filters. My hardware engineering curiosity is piqued though as to how something that's successfully implemented on an NPC digital filter in late 1980s CMOS technology results in serious slowing down of a current-day PC... What software were you running?

>> Ideally the FPGA and companion chips won't need to
>> be BGA packages to give DIYers a hope of making it
>> themselves.

>BGA is so much better. Anyway, this is an off the shelf board from
>Digilent, you can get it from them for cheaper than the parts cost
>in quantity 1. Much simpler.

Here I was thinking ahead rather to something beyond your 'proof-of-concept' stage. I'd prefer the option of doing my own board design at a later stage, not relying on continued availability of the digilent board. For the prototyping stage, I can't fault your choice of that board, it looks superb.

R