USB DAC Design questions

[steinchen]

The free version of Eagle is limited to half eurocard size (100x80 mm = 4x3.2"), I purchased the non-profit version and am limited to eurocard size (160x100mm = 4x6.4"). The standard version has the same limits, only the professional version supports larger boards.

 

[Clutz]

Quote:

Originally Posted by steinchen The free version of Eagle is limited to half eurocard size (100x80 mm = 4x3.2"), I purchased the non-profit version and am limited to eurocard size (160x100mm = 4x6.4"). The standard version has the same limits, only the professional version supports larger boards.

Hrmm, well then I think we're probably going to have to get the standard version - cause I somehow doubt we can get everything on a half eurocard PCB. I'm probably willing to do it, but I've never done layout before at all before. Hell, I've never drawn a schematic in it - I've only ever drawn schematics in Adobe Illustrator.

 

[steinchen]

the standard version has the same limits like my non-profit version, no need to purchase such an expensive version

nevertheless, so far the board size limit is 160x100mm, with this limit we'd have to modularize ...

 

[Clutz]

Quote:

Originally Posted by steinchen the standard version has the same limits like my non-profit version, no need to purchase such an expensive version nevertheless, so far the board size limit is 160x100mm, with this limit we'd have to modularize ...

Well, in that case- would it make sense to move the PSU on a separate PCB? We have basically three components: power supply, dac, i/v section.. and I would think it makes the more sense to keep the DAC and i/v section close to each other, rather than the DAC and power suppy, no?

Cheers,
Clutz

 

[00940]

In my view, we should be able to put everything quite easily on a eurocard. In my first layout attempts, I was able to fit everything but the I/V stage on half of that.

 

[MegaMeee]

I am also interested in a design like this... I was going to build a usb dddac (which also used the pcm2707 direct i2s), but i'm not convinced by those nonos dacs.
Personally the original budget $150 - $200 seems a little low to me. I think we are looking at ways to minimise cost too much considering how many people here seem to build "maxed" versions of everything.


Quote:

Originally Posted by 00940 In my view, we should be able to put everything quite easily on a eurocard. In my first layout attempts, I was able to fit everything but the I/V stage on half of that.

so what did you put on your first layout attempts?

Personally I don't care how many boards it is on (as long as it dosent lower the performance), but If everything will fit on a eurocard then why bother with more than one?

edit: btw ignore my earlier question. This: http://focus.ti.com/lit/an/sloa099/sloa099.pdf answers it.

 

[Francis_Vaughan]

Chiming in again.

Need to set a few parameters.

Size and configuration. I have been thinking about a DAC much like this for a while. My personal desire is a DAC that can go in the same case as my PPA. This makes for some interesting constraints. Since I built mine with the battery board there isn't a full width space anywhere. So the board would need to be long and slightly slimmer than the case to clear the power capacitors. It struck me that using the crossfeed mounting holes would be a good plan.

Power is an issue. Much like the 2702 DAC, a range of power feed options may be contemplated, but to use in a PPA - or similar - one needs to provide for a virtual earth - this probably means a BUF634 or similar on board. It also means possibly high regulator dissipation if used with PPA realisations with high supply voltages. It does have the nice effect that it can be used with a STEPS PS - something many people already have.

If laid out nicely it could use the same Molex connectors for power and audio as the PPA board, and could be configured so that it provides power loop through, enabling very neat wiring in the box. The audio could come out at the same end as the volume control in a PPA, and connect to the secondary audio input connectors right next to the volume pot.

All this is about how to make it nice for a PPA or perhaps MMM. Clearly nothing should prevent its use in more general settings. But I bet that a huge number of PPA builders would be very interested if the pairing was thought through.

The design by MWP points out something I had overlooked - the 2707 does provide a system clock output - although the data sheet is rather mute about its specification, it is clearly fast enough to run the 1793 OK. However the whole system is then totally dependant upon the clock integrity of the 2707 - something which is probably not really up to the task at hand. Hence the desire to add some jitter attenuation. An ASRC seems the right answer in this context. Sadly we have little information about the spectral content of the jitter from the different components, so it isn't clear which is the optimal balance. Personally I would even like to add a higher quality clock input for the 2707, on the basis that giving the ASRC as clean as possible a clock on input can only help. The data sheet for the 2707 cautions against an external clock, but only becuse the suspend function will not be able to stop the clock. If you don't intend using this, it is of no matter.

Crystal oscillators with logic outputs and very low phase noise are available from a number of sources, often very cheaply. The Tent clock is pin compatible, so those with boutique tastes could try one to see if it is any better. Those from Valpey Fisher look very nice.

A layout that allows for a base I/V but also allows for that part of the board to be not populated so as to support other (off board, or daught boarded) I/V stages seems very much the right idea. Personally I am rather interested in using a coule of nice Lundhal transformers. Cost as much as the rest of the DAC however. Also discrete solutions like the Hawksford or even the Pass D1 are interesting.

 

[ble0t]

When you mention the clock output of the PCM2707, are you refering to the FUNC2 pin (I2S system clock output)? I would be heistant to use it as a master clock elsewhere in the system without having anymore specs on it. Perhaps someone can get in touch with TI about it?

Also, as far as using a higher quality clock for the PCM2707 itself, I've read a couple places (and found some info here) that increasing the quality will not really do much. I would lean toward providing a very clean clock to the ASRC/DAC instead. Perhaps implementing something like this would be cheaper and work nicely for our purposes. We could possibly even look at using something like this.

 

[Francis_Vaughan]

Yes, I was referring to the Func2 pin. I agree - I would be very hesitant about using it. But it does explain why it is possible to interface a 2707 to a 1793 without additional componentry. That is all.

I'm not convinced about not cleaning up the 2707 clock. There is more to the clock than the simple Spact system. I'm not convinced that the system clock does not influence the sample clock in the 2707. Improving the system clock may improve the system clcok - it may not totally cure it's ills, but it may hekp. The rationale is that the ASRC can only attenuate - not cure. Better jitter on the way in can only help - and the spectral characteristics of the jitter are not known, which makes it hard to really know.

The quality of the system clock for both the DAC and ASRC are paramount. Jim Haggerman's clock is likely pretty good - he is a good analog designer. The Maxim clock is less so. Weasel words - "best in class" - and it quotes 21ps. We can find clocks with 3ps. Just not as flexible. But we probably don't need the flexibility.

 

[MegaMeee]

Quote:

Originally Posted by Francis_Vaughan Chiming in again. ... ... Crystal oscillators with logic outputs and very low phase noise are available from a number of sources, often very cheaply. The Tent clock is pin compatible, so those with boutique tastes could try one to see if it is any better. Those from Valpey Fisher look very nice. A layout that allows for a base I/V but also allows for that part of the board to be not populated so as to support other (off board, or daught boarded) I/V stages seems very much the right idea. Personally I am rather interested in using a coule of nice Lundhal transformers. Cost as much as the rest of the DAC however. Also discrete solutions like the Hawksford or even the Pass D1 are interesting.

about the 2707 clock I did notice this on the dddac site: "Oh, not to forget, by comparing the Tent Clock and the normal Crystal, The Tent clock brings more defined focus..... When the gain from SPDIF to USB is 100% on some kind of imaginary scale, than this gain would be on roughly 3/4 of that scale......"
whatever

I agree with the idea of a base (opamp based?) i/v stage with pins or whatever so we could use our own... although I can see myself trying about 10 different things and then going back to what I started with.

About the ppa thing would making it "ppa compatible" restrict the psu options? Oh and wouldnt it make things overly complex, because i'm assuming you would want rca outputs as well you would have a crowded back panel (not to mention a crowded inside

)

 

[ble0t]

My thought for the power supply would be to provide three options:

1) AC powered (rectifier section, filtering, etc.)
2) DC powered (bypasses AC rectification)
3) Powered from external amp (use BUF634 for ground)

This shouldn't be too difficult and/or space consuming to implement.

 

[00940]

Bleot : it would be easy only if we use a single supply for everything and a virtual ground (to provide negative voltage for the I/V stage, you'll always need the virtual ground with a single supply).

It makes it easier to integrate with other components, requires a smaller enclosure as stand alone (you can use a wallwart for example), and so on. However, it also makes impossible to separate digital and analog supplies and complicates the grounding scheme a lot.

Aiming for quality rather than ease of use, I'd say the best would be to use two boxes, eurocard sized. One for the transformers (I'd shoot for two) and basic rectification, one for the dac and the regulation.

 

[Clutz]

Quote:

Originally Posted by Francis_Vaughan Chiming in again.

I always look forward to your input.

Quote:

Size and configuration. I have been thinking about a DAC much like this for a while. My personal desire is a DAC that can go in the same case as my PPA. This makes for some interesting constraints. Since I built mine with the battery board there isn't a full width space anywhere. So the board would need to be long and slightly slimmer than the case to clear the power capacitors. It struck me that using the crossfeed mounting holes would be a good plan.

I think it's a really good idea, but it would fall on my list of "wants", not "needs", but that if it's something that can be accomodated without much sacrifice elsewhere, then it's a particularly good idea. That said, the way I will build it, will be to follow 00940's lead - one box for the DAC, one box for it's power supply. Actually, it makes me think that perhaps it'd be a good time to build a good STEPS power supply.

Quote:

The data sheet for the 2707 cautions against an external clock, but only becuse the suspend function will not be able to stop the clock. If you don't intend using this, it is of no matter.

Personally, I think the lack of the suspend function is of little consequence. It is something I would very happily give up for better performaince.

Crystal oscillators with logic outputs and very low phase noise are available from a number of sources, often very cheaply. The Tent clock is pin compatible, so those with boutique tastes could try one to see if it is any better. Those from Valpey Fisher look very nice. [/QUOTE[

This is what I would argue for, it'll allow for people to build it with a reasonable quality clock, or build it with a very high quality clock.

Clutz

 

[ble0t]

My assumption concerning the grounding scheme is one plane that is as solid as possible using ferrites on all supply lines. As long as the 'flow' of the board is done properly, this should work well. As far as the power goes, I would definitely shoot for dedicated, local regulators for all the chips. Perhaps removing the option of AC and just requiring DC would be a good compromise here. This would still allow for the use of and external PS (i.e. STEPS based, multiple TREAD based or perhaps our own design) that would be additionally regulated close to the chips.

Also, we have agreed that having Iout pads for the I/V stage is a good idea, but are we going to have a 'standard' I/V stage on the board? If so, which one?

 

[00940]

Quote:

Originally Posted by ble0t Perhaps removing the option of AC and just requiring DC would be a good compromise here.

Really, if we go through the troubles of working with only one single supply, providing an AC input is trivial. The big problem is to choose in between one supply line or four.