Pc -> Dac, How ?

[dwk123]

I've been meaning to respond to this for a while, but with holidays and travel, haven't been able to. There are two separate pieces to this.

First, "what's the point?" For those saying "use an existing soundcard" or expressing similar ideas, I think there are two aspects that are missing:
1) multichannel. I'm not aware of any multichannel solution that uses a "good" clock architecture. No doubt some PCI/e cards probably use good design, but run into the problem of the PC power delivery system. I use an Emu 1820M which is pretty much state of the art in terms of 'affordable' multichannel rigs and yet it still uses noisy PC power for the clock etc. (although at least the DAC's can be separately powered)
2) Linux. This may be a secondary desire for peufeu, but it still seems important. There is a real dearth of good USB solutions that work under Linux. The Transit more or less works, and the 0404 USB can be cajoled into working at 16 bit, but it's a bit dicey. *IF* linux support is a goal, then IMHO rolling something from the ground up like peufeu is proposing isn't necessarily a bad approach.

Certainly, for many folks that just want a 2-channel solution under Windows, there may be easier/quicker ways to get there via a modified commercial card and/or by slaving via spdif. For multichannel it's probably not that easy.

The second aspect of this is really "should we care?". By that I mean is this actually going to be a group participation project in any meaningful sense, or is this simply going to be a showcase for peufeu's private system? I admit I have some concerns along these lines, as peufeu seems ready to charge down the path of FIR xovers and digital filters before there is really a solid playback approach in place. If this is going to be more than a one-off, the barrier to entry for interested participants/contributers has to be kept reasonably low, and it has to do something they're interested in. This means several things, I think:
- hardware resources. Boards or at least easily reproduced layouts for the isolation module, the clock/reclock module etc. Also good parts lists from common vendors to make ordering connectors, clocks, chips etc straightforward. Without this, everyone will have to do their own custom work, which means there really is very little underlying commonality.
- software. It needs to be relatively easy to configure the board to do what folks want (specifically 2-channel unfiltered playback has to be easy) If they don't want FIR/digital filters, they shouldn't have to hack verilog to get rid of it. Similarly, playback from the host shouldn't require an advanced degree to get working, and should if possible work with garden-variety distros (at least at first - I actually think a custom thumb-drive distro down the road might make sense).
- compatibility (to a degree). The TP guys have a fair number of I2S compatible modules out there, and I suspect a lot of folks that might be interested in this USB project would use the TP boards as a starting point. This of course shouldn't be a problem, but stranger things have happened.

Obviously, not all of this falls on the shoulders of one person, but without some organization of this type, this may never be more than a curiosoty for most people.

IMHO starting *really* *really* simple and getting folks on-board would be the best way to build a community of users, but unfortunately that is also likely to be the most boring for peufeu. Getting half-a-dozen folks all running a command-line user-space playback chain on common hardware would probably be enough to start generating some momentum, but that might mean defering the more advanced aspects for a while.

Since I'm a Linux guy and multichannel is embedded at the core of what I'm trying to do with my system, I'm very interested in this project. I'd be quite interested in being an early adopter if the situation is right.

[jkeny]

dwk123,
Not wishing to anticipate Peufeu's reply but:

- this isn't a showcase for Peufeu's home system - far from it - it is a truly open source/hardware system that is intended to act as a platform for open experimenting on various DAC modules/ PS modules, etc and by given a standard platform to operate from can bring the DIY community closer to being able to really evaluate different DAC, filtering schemes, etc without all the variables that so often hamper doing this in any scientific way - so I think "we should care" in seeing this fully materialise. I believe Peufeu stated why he was started this thread

- This project is already more developed than you seem to believe with a lot of research gone into selecting the most suitable FGPA board, currently the NEXYS2. This is communicating with PC via bulk USB.

Have to go!

[dwk123]

Originally posted by jkeny
dwk123,
so I think "we should care" in seeing this fully materialise. I believe Peufeu stated why he was started this thread

I think you misunderstand the point of my post. I agree that "we should care" about the basic idea of what peufeu is doing. My point is that a project will have to be approached rather differently if the goal is to build a wide user base than it would if the goal is simply to get one or two systems working doing cool stuff. So far, there is every reason to believe that peufeu will get a system up and running, but I'm not sure about all of the "other stuff" needed to make something that is usable by a larger audience. The goal of my post is to get some discussion going surrounding this type of issue.

As an example, if the only way you can use peufeu's fpga code is to roll-your-own reclocking circuit and hack the verilog to disable direct digital filtering, then 'few' other folks are likely to bother with the project. On the other hand, if someone is offering boards for the reclocker and there is a quick/easy order basket for it, then it's far more approachable. It's not reasonable for peufeu to do all this himself, but since he's providing the impetus for the project he will have to provide the primary support/organization for it.

- This project is already more developed than you seem to believe with a lot of research gone into selecting the most suitable FGPA board, currently the NEXYS2. This is communicating with PC via bulk USB.

I think I'm pretty well aware of the state of the project, as I've been following this since the Ethernet version a year or two back. I actually almost ordered a board from digilent before xmas, but figured holding off until things were a bit more clear might be wise.

[abraxalito]

dwk123:

I pretty much agree with all of what you've said - I'm also in this to make it more 'accessible' to the average punter than having to hack verilog and figure out how to attach 1mm pitch BGAs to boards...

I don't though agree that peufeu needs to provide the impetus to make this available to a wider audience - he's clearly motivated by getting something working, the day to day work of preparing parts lists and checking out availability is probably not going to appeal. I figure there are probably enough of the rest of us here to work out those aspects. We'll just have to drive peufeu to get the answers we need.

R

[tritosine]

i would like to see an open source alternative against prism dscope and audio precision

[Wavelength]

Gang,

Basically if we go with the Cypress USB 2.0 controller, it would require some kind of FPGA to have complement I2S outputs. The flow is simple and would work pretty much like this:

USB2.0<===>Cypress[DATA]->FPGA---->I2S--->DAC(s).

The MCLK at the DAC would be fed into the FPGA and that would determine the rate and so forth. The Cypress would receive Endpoint 0 setup information as to the changes in the Fs rate and output those via GPIO ports. The GPIO output would go to the FPGA and also aid in the selection of MCLK's to support various Fs rates (i.e. 44.1, 48, 88.2, 96, 176.4, 192).

The only really other alternative to the Cypress would be the new Atmel AVR32 parts which should be released soon. It does have both USB 2.0 and PHY and DMA SSC port that is capable of 24/192. I am under NDA and can't really say anything more on that till it is released.

i would like to see an open source alternative against prism dscope and audio precision

What??? Buy a lynx card and do some software buddy. There is a ton of FFT on the web, learn it understand it write it. Or buy something like the Dr. Jordan stuff.

But really you will never come close to what the Prism has. The SN ratio on that ADC is incredible.

Thanks
Gordon

[tritosine]

Originally posted by Wavelength

What??? Buy a lynx card and do some software buddy. There is a ton of FFT on the web, learn it understand it write it. Or buy something like the Dr. Jordan stuff.

But really you will never come close to what the Prism has. The SN ratio on that ADC is incredible.

Thanks
Gordon

i do have a lynx card but no adc to feed its digital ins, thanks to some hack who talked EVERYONE into buying that pcm4222 evkit with his clever marketing strategy , making it sorta out of stock
oh & what do i see in the AP1 service manual are these digitally controlled analog generators and notch filter,and since AP made this available for everyone these are sorta out in the wilderness now . Happy NY gang

[davidallancole]

Has anybody looked at this product?

http://beagleboard.org/hardware

It seems to be a mini computer that could be used for a PC and spits out I2S directly. This could then be fed to a DAC chip easily.

[Russ White]

Originally posted by davidallancole
Has anybody looked at this product?

http://beagleboard.org/hardware

It seems to be a mini computer that could be used for a PC and spits out I2S directly. This could then be fed to a DAC chip easily.

Yes, I have used it.

I have it decoding flac (from SD card) out to I2S. Even at 192/24. I have not tried to use for USB yet.

This is actually very easy to do since it runs Linux.

Cheers!
Russ

[davidallancole]

Hi Russ,

Which DACs have you ran with this beagleboard?

It seems like a clean solution because a lot of DACs I have seen have I2S inputs and generally need USB or SPDIF converter chips in front of them.