USB DAC Design questions

[ble0t]

Sorry...I should have made myself more clear when I was talking about DC powered. I wasn't assuming the use of a single STEPS or something of that manner. Obviously, you could use one in addition to other things, but I believe we are going to need 4 or 5 different DC supply voltages that should be further regulated onboard.

As far as the I/V stage goes, there are a couple options I've found thus far (although they might require some modification)...

http://users.verat.net/~rogic2/1541/...screte_I-V.pdf

http://users.verat.net/~rogic2/1541/pdf/max436.pdf

 

[00940]

Iirc, those I/V from pedja are made for the TDA1541 and won't really work with other dacs.

I/V options I gathered :

- opa1632
- opa134 and the like (single, dual or even quad package)
- Pass D1 (copyright issues, requires +/-30VDC)
- transformers (crazy expensive)
- 2sk389 variant of the D1 ( http://www.diyaudio.com/forums/showt...threadid=34191 )
- balanced version of Jocko's easy I/V (http://www.diyaudio.com/forums/showt...&threadid=9910 )

Glassman and Porksoda reworked the Hawksford I/V into something really interesting : http://www6.head-fi.org/forums/showt...=123317&page=2 Worth trying ?


Bleot : who will have 4 different dc voltage with no common ground (to avoid a ground star where we don't want it to be) on hand ? Really, I think we have only two options in the design of this thing. Or we decide to design the PS according to our DAC, without any considerations of compatibility. Or we decide that the ability of running it from other voltage sources is important (and I can see why it could be) and then we design the DAC according to the PS.

The first option will get the people stuck since they will need to do it like we designed it. The second option let the people quite free to use whatever they want (let's say anything providing in between 24 and 30VDC) to power the dac.

 

[ble0t]

As usual 00940, you're probably right

EDIT: What I was thinking about was something along the lines of this where they have a single higher voltage supply for all the digital supplies (regulate down from it) and a seperate analog PS.

Personally, I'd like to use a discrete I/V stage, so either the modified Hawksford that Glassman/Porksoda came up with or Jocko's would be nice. I guess then we could argue between feedback vs. non-feedback

 

[Clutz]

Quote:

Originally Posted by 00940 - opa1632 - opa134 and the like (single, dual or even quad package)

I think we should ignore the opa134 in favour of the opa1632 for reasons you've previously mentioned..


Quote:

- transformers (crazy expensive)

Define "crazy expensive"?

Increasing costs by 50%? 100%? 200%?

Quote:

- Pass D1 (copyright issues, requires +/-30VDC) - 2sk389 variant of the D1 ( http://www.diyaudio.com/forums/showt...threadid=34191 ) - balanced version of Jocko's easy I/V (http://www.diyaudio.com/forums/showt...&threadid=9910 ) Glassman and Porksoda reworked the Hawksford I/V into something really interesting : http://www6.head-fi.org/forums/showt...=123317&page=2 Worth trying ?

Do we know enough about each of the options to know the relative performance / cost of each? This is probably obvious, but I like stating the obvious, I don't think we should necessarily not consider an option just because it's expensive until we consider the relative merits of it versus a cheaper option. If it increases the project cost by 50%, but is providing that last 1 to 5%, then I think we have to ask ourselves if this is the best way to spend that money; however if one option is better than another, and it costs 10% more, and gives us 5% better performance, then I think it's worth considering. (I'm just making up numbers here.. I really have no idea what I'm talking about).

 

[dsavitsk]

Quote:

Originally Posted by Clutz Define "crazy expensive"? Increasing costs by 50%? 100%? 200%?

Probably $150-$200 to do it right. http://www.sowter.co.uk/acatalog/SOW...ORMERS_12.html

Quote:

- Pass D1 (copyright issues, requires +/-30VDC)

My occasional lurks in the Pass forum suggest that if you ask really nicely and make assurances that there is no commercial intent, creating PCBs for the circuits is not uncommon. And, +/-30V is really not that onerous, so this option is probably worth pursuing.

There is also a patent issue with AudioNote w/r/t/ the transformers, but again, a friendly request and an assurance of no commercial intent may be enough.

 

[Clutz]

Quote:

Originally Posted by dsavitsk Probably $150-$200 to do it right. http://www.sowter.co.uk/acatalog/SOW...ORMERS_12.html

okay, that meets my definition of crazy

but if we left output pads, then someone could still do this on their own if they wanted to, right?


Quote:

My occasional lurks in the Pass forum suggest that if you ask really nicely and make assurances that there is no commercial intent, creating PCBs for the circuits is not uncommon. And, +/-30V is really not that onerous, so this option is probably worth pursuing. There is also a patent issue with AudioNote w/r/t/ the transformers, but again, a friendly request and an assurance of no commercial intent may be enough.

I don't think that +/-30V is too bad- it just means a more complicated power supply set up... But I'm still interested in a discussion of the benefits of each of these options. How much better are the better options?

 

[dsavitsk]

Quote:

Originally Posted by Clutz I don't think that +/-30V is too bad- it just means a more complicated power supply set up...

Not really. You are still in LM317 territory here, so even a pair of treads could do this. Obviously, a higher voltage transformer (and a safety warning) would be necessary, bt otherwise not a big deal.

 

[00940]

my grief with the +/-30v of the D1 compared to something like +/-12v for the output stage has three components :

- we are forced to use our PS, no way to power the thing from a steps or the like. The power supply is an issue not solved yet.

- we cannot use easily the analog supply for the analog section of the dac (needs 5V) or for an opamp for the balanced to unbalanced converter.

- 50V caps take a lot more place than 16V caps.

 

[Clutz]

Quote:

Originally Posted by 00940 my grief with the +/-30v of the D1 compared to something like +/-12v for the output stage has three components : - we are forced to use our PS, no way to power the thing from a steps or the like. The power supply is an issue not solved yet. - we cannot use easily the analog supply for the analog section of the dac (needs 5V) or for an opamp for the balanced to unbalanced converter. - 50V caps take a lot more place than 16V caps.

That would mean we'd need to have or PSU in a seperate box, and three transformers, which would be getting rather expensive.

 

[Francis_Vaughan]

I think a step back is needed. This isn't intended to be the greatest DAC design ever. Furthmore I am generally unconvinced by the mad dash for utterly seperate power supplies for various parts. The issue is isolation, and that is addressable without going for seperate transformers.

I would go for a single power supply. DC input, with local regulation where needed. The issue then becomes what we are prepared to provide for the I/V and output buffer. The default answer there is usually +-15v. But +-12v is easily as good.

So this would lead us in the direction of a design that takes either +-15 and 0v, or +30 (with the addition of an optional rail splitter) but that will be totally happy running off +-12 (or 24v with splitter). This would be a very generic spec. It should work in most applications, and has the nice advantage that it is compatible with the PPA/MMM + STEPS configuration that has proven so popular here.

If there is demand and space, one could add additional pads and jumpers so that the insane

can break out the powerfeeds and use multiple additional supplies. Personally I very much doubt it will help.

Using the Pass D1 I/V is always going to be a pain. Apart from the large rails it is physically quite large, and gets hot. Is is always going to be a duaghter card. Indeed there are a few designs floating about. This very much a case of someone simply needing a couple of pads in the right place on the DAC board and then being on their own. Other discrete designs are more interesting. It may be the the board can be laid out so as to make small daughter cards easy. Maybe with pin headers in carefully selected places there can be a base daughter card layout - just defines signal and power pin locations and form factor.

 

[Clutz]

Francis, I think you've made a number of good points - particularly reminding us that this isn't supposed to be the greatest DAC ever. Regarding the power supply, I think perhaps we could take note of Alf's DAC. I believe he found that they over engineered their power regulation - and that it actually measured better when one of the regulators was bypassed.

I think that providing pads or jumpers so that a better I/V stage can be used by those who want to should be considered a design necessity. I'm curious as to which I/V stage you would favour Francis? Without really knowing much of anything about anything, I find the OPA1632 to be quite an interesting option. it's relatively low cost and easy to implement. Are the discrete designs substantially better? My preference, in no particular order would be for the OPA1632, using Jocko Homo's balanced I/V, or Glassman & Pork Soda's discrete I/V stage.

 

[00940]

Quote:

Originally Posted by Francis_Vaughan I would go for a single power supply. DC input, with local regulation where needed. The issue then becomes what we are prepared to provide for the I/V and output buffer. The default answer there is usually +-15v. But +-12v is easily as good. So this would lead us in the direction of a design that takes either +-15 and 0v, or +30 (with the addition of an optional rail splitter) but that will be totally happy running off +-12 (or 24v with splitter). This would be a very generic spec. It should work in most applications, and has the nice advantage that it is compatible with the PPA/MMM + STEPS configuration that has proven so popular here.

We will have to run the I/V at +/-10VDC. If we allow the supply voltage to be 24V, we need these drops for final regulation and possible variations in the supply line. It's also too low for discrete designs (they require from +/-15 to +/-20V). We're pretty much stuck with opamps then. We will need a strong rail splitter : almost everything will run from the positive supply.

Quote:

If there is demand and space, one could add additional pads and jumpers so that the insane can break out the powerfeeds and use multiple additional supplies. Personally I very much doubt it will help.

It would be rather useless if the grounding has been optimized for a single supply.

 

[Francis_Vaughan]

I had rather deliberately ignored the additional regulator drop for the output stage. But it is an important issue. Depending upon board space there are other options - but we can at least set some parameters.

Quote:

It would be rather useless if the grounding has been optimized for a single supply.

Mostly true - but it doesn't seem to stop people from doing this. Worse, the designs that split the grounds up too and seem utterly unable to grasp the desperately poor ground returns created. There are designs where significant RF energy is only coupled back via the chassis. It is impossible to make any sort of quality system like that.

Until one is quite sure there is no RF energy in the signal one needs to maintian ground integrity - and that typically means pretty much up to the output connections - just maybe before the output buffer - but only if the entire reconstruction filter is before the buffer.

 

[DaKi][er]

While you can get away with it in headphone amps, single supply virtual grounds should be avoided, they add far more problems than they fix

Really, there is 2 ways you could design-
Either design it to be portable, where you would want it to run off USB power or as this was set out to be, a standalone dac to run off mains with its own transformer, psu and all that with no halfway could be portable could be fixed standalone, it should be either one of the other as designing for both just makes compromises. Have pads for the AC from the transformer/s and put everything else onboard. I’m a very strong believer that the psu should be built in as part of the design and leave very little choice

I don’t have any personal interest in building one of these as I already did one for myself and I am not much of a fan for designing projects like these in groups as there are always far too many people wanting to put in their own input and in the end always throws it off course. I like to design by myself, for myself and everyone else is an afterthought.

What you guy's need to do is work out exactly where you want to end up and start listing the requirements, with everyone having different ideas that will be fun

 

[ble0t]

Given everyone's responses, perhaps lets take a step (maybe a couple

) back then and define what we're looking for piece by piece so we can come to an agreement. Since the PS has been the topic of most controversy, perhaps we can come to an agreement on that? I would suggest (after reading the previous posts and doing some research myself) going with a single DC supply that uses local regulators near the chips.