[ble0t]

Quote:

Originally Posted by Clutz Are there better USB to SPDIF chips available?

It's actually USB --> I2S...and I'm not sure who else makes them. Perhaps AKM?

 

[SnoopyRocks]

Lets not get carried away by the measurements on the other site. There's not much more that can be done about them (if they are in fact represenative) given the scope here anyway. The basic system here is fundamentally sound and reasonably addresses the problem. If done well, it should sound great. Forget about the measurements and get back on task. If the clock out of the ASRC looked like that, then we'd have problems.

 

[dsavitsk]

Quote:

Originally Posted by SnoopyRocks Lets not get carried away by the measurements on the other site.

This is right.

This test was bus powered with no noise filtering. How do we know this is representative of anything. Maybe it is, but one measurement of a questionable setup does not a result make (aren't some of you guys scientists?) Further, if you are all going to get caught up on overthinking every aspect of this project, it is never going to get done. I have used a pcm2707 into a tda1543, and it sounds good. Could it sound better? Sure, but the point of this project, as I understand it anyway, was never to build a perfect DAC with 0 jitter. It was to build a reasonably priced DAC that sounds reasonably good. The PCM2707 will do that, and it will likely do it better than any other way you plan to get digital audio out of your computer (except maybe via squeezebox, but it is already over the proposed budget.) I think perhaps the test shows that Wavelength Audio is full of it, but we knew that anyway, so it seems to not be such a startling result. Oh, and even though mine is not bus powered and filters every aspect of the usb signal, using a usb cable with a ferrite on it makes a difference as does the computer used, so a well controlled experiment is in order if you want results that matter.

Now, get back to work.

-d

 

[ble0t]

Quote:

Originally Posted by dsavitsk Now, get back to work.

You slave driver you

 

[MegaMeee]

Quote:

Originally Posted by dsavitsk Now, get back to work. -d

I agree

I pretty much expected results like these... And I pretty much don't care. As said if worst comes to worst we still end up with a good value spdif dac that you can also plug into any pc.

 

[pipp]

I just finished a usb dac. Based on pcm2706 i2s and stacked TDA1543's. Maxed out with Black Gates and tantalum resistors. Power is 2 x Regulated AC -> 9v and 2 x separate 9v to 5v supplys. It plays great !

 

[Clutz]

New PCM2706 data over at diyhifi..

http://www.diyhifi.org/forums/viewtopic.php?p=8251#8251

Much, much better looking now.

 

[Francis_Vaughan]

Pretty aggressive layout in terms of bypassing - but a bit of a "I keep telling everyone this" for me too

Got to really push the power supply de-coupling as hard as you can.

 

[00940]

Would this be sufficient ? PCM2707 receiver section, based on Bleot's schematic.

 

[SnoopyRocks]

I've decided to do parallel developement with PADS2005 instead of Eagle since learing a new software suite is not that appealing to me right now among other reasons. The schematic with the symbols I've created is attached.

number of pcb layers
IMO, this should be a 4 layer board. What does everyone else think?

board size
Do we have a target for the board dimensions? Is there a particular case in mind?

power section
The voltage regulator section is out of hand. Separate supplies (regulators) for analog and digital is reasonable. A regulator per chip is rediculous. This application is not that sensitive and any performance benefit would be mininal. Regulators don't isolate well at the the digital switching frequencies anyway. Be reasonable.

physical dimensions of the usb jack
Can someone please direct me to a mechanical drawing of the usb jack so I can make a symbol with the right footprint?

I have more comments and questions, but this is enough for now.

 

[00940]

Quote:

Originally Posted by SnoopyRocks number of pcb layers IMO, this should be a 4 layer board. What does everyone else think?

Way too expensive for small run production, especially since the price must stay around 200$.

Quote:

board size Do we have a target for the board dimensions? Is there a particular case in mind?

Eurocard sized I'd say ?

Quote:

power section The voltage regulator section is out of hand. Separate supplies (regulators) for analog and digital is reasonable. A regulator per chip is rediculous. This application is not that sensitive and any performance benefit would be mininal. Regulators don't isolate well at the the digital switching frequencies anyway. Be reasonable.

Well, the PLL sections of both the spdif and usb receiver /are/ very sensitive. And the evaluation board of the AD1896 uses two seperate regulators for the digital and analog supplies of the AD1896 (both 5V). The amount of regulators is thus not crazy in my view, we could do crazier.

But the scheme probably needs to be modified. A practical problem in the layout is that we have 3.3v and 5v digital lines. The dir1703 and pcm2707 need 3.3v, so does the pcm1794. In the middle we have the multiplexer and the ad1896 and their 5v. A big 7v line and local regulators make it easier to layout at first sight. A compromise would be to use one 3.3v regulator for both pcm2707 and dir1703, a 5v regulator for both the multiplexer and the ad1896 and one 3.3v regulator for the dac digital section.

edit : another advantage of multiple regulators is to use the reg101 which can output at most 100ma, and preferably with not too much current, to stay in its most linear operating range. If we use the same regulator for more chips, we probably have to load it more or pass to a reg102, with lesser specs.

 

[Clutz]

Quote:

Originally Posted by SnoopyRocks number of pcb layers IMO, this should be a 4 layer board. What does everyone else think?

This would be a lot more expensive, and I think we need to have a target budget for this project or else it will be very easy to go overboard. What are the benefits of going with a 4 layer board? How much more expensive is a four layerboard than a two layer board for a small run?

Maybe we should start to add up the parts costs at this point, so that we know how much of the budget we've spent so far?

Quote:

power section The voltage regulator section is out of hand. Separate supplies (regulators) for analog and digital is reasonable. A regulator per chip is rediculous. This application is not that sensitive and any performance benefit would be mininal. Regulators don't isolate well at the the digital switching frequencies anyway. Be reasonable.

I agree with 00940 here about the reg101 versus reg102, but I think that 00940's compromise seems like a good one.

 

[SnoopyRocks]

Is the goal here a $5 pcb? The pcb is one of the most important components of the system. Moreover, since the sentiment is towards abundant power bypassing, lets do it right with propper power planes. The cost is not exporbitant and IMO money is better spent here than on boutique opamps.

For a frame of reference, my last board to test an IC I designed was 3.9 x 6.9 in. (slightly larger than eurocard) with considerably more holes because of generous stiching. This was a 6 layer board. The cost was a little over $900 for 10 boards. 4 layers would have been about $550. I don't think that ~$50 for the pcb is unreasonable. With more boards, the price per boards goes down quite a bit. Is $35 for a 4 layer board unreasonable?

 

[dsavitsk]

I think there is a disconnect here. The only way you are going to get people on this or any other forum to put down that kind of money for a DAC, it had better be prototyped and it had better work way better than the other projects out there that can be built for about $100. This means that you are going to have to sell a ton of boards, you are going to have to eat the cost of prototyping, you are going to have to get it pretty much right on the first try to avoid several rounds of prototyping, and you are going to have to beat Yeo's board. Any one of these seems difficult, but doing all is probably an unreasonable thing to ask. Go with a 2 layer board, send it to batch PCB to make a handful, and be done with it. This is a fun learning project, not a best DAC ever made project. I think (and correct me if I am wrong) when Clutz says $150-200 he means that's what he wants to spend, not that he wants to put down $1000 to develop a DAC whose intellectual property rights and sunk costs he is going to give away so he can sell $35 to $55 DAC boards.

 

[mikechai]

Quote:

Originally Posted by pipp I just finished a usb dac. Based on pcm2706 i2s and stacked TDA1543's. Maxed out with Black Gates and tantalum resistors. Power is 2 x Regulated AC -> 9v and 2 x separate 9v to 5v supplys. It plays great !

What is the total cost to build? Any pictures?
Any comparison with commercial DACs or sound cards?