USB DAC Design questions

[Clutz]

Hello all,

I've decided for my next project I'm going to try and design and build a USB DAC. While there are already lots of interesting options available out there, designing my own really appeals to me.

I'd like to keep the parts budget between $150 to $200 (i.e. in the same range of a PPA, which will be used with this DAC). My goal is to have a USB DAC that is higher end than the PCM270x USB DACs but less complicated than DaKi][ers DAC.

Since I plan on using this primarily as a desktop DAC to use with my PC, I don't have any particular desire to include a SPDIF input. This will allow me to spend the budget elsewhere on the DAC on things I care more about, and it will simplify the signal path (i.e. USB -> PCM2704 to convert to I2S -> DAC chip). That said, I would appreciate any input on the cost/benefit of including an SPDIF in. My personal opinion is that given my price restrictions, I'd simply be giving up too much to include it.

My second question is, what are the benefits of going with two one channel DACs versus a single two channel DAC? The most obvious one that comes to mind is reduced stereo cross talk, but I'm not sure how much of an issue that is (currently the DAC chips I'm contemplating are the Cirrus Logic CS4351, the Texas Instruments PCM1738, PCM1792, PCM1798. All of the texas instraments DACs seem to have differential outputs, while the CS4351 has a single ended output - but has a 2Vrms output.

I haven't picked exactly which DAC yet, but so far I think I am leaning towards a design using a PCM2705/6 -> PCM1794. I think I would use either the OPA627 or possibly the AD843,although these will be socketed, so it doesn't REALLy matter. Another possibility would be the 6171, but if I decided to do that,or wanted that possibility, I'd have to design that into the schematic from the start.

I'm also interested if people think it would be worthwhile using an external digital filter from the DAC or if that money would be better spent elsewhere.

I'm not yet sure how I'll do the power supply, I'll need a 3v3 and 5v psu for the DAC itself, but I'm not sure about the power supply for the op-amps.

Anyway, I'd appreciate your input. I hope to learn a lot from this.

Cheers,
Clutz

 

[00940]

For I2S you need the pcm2707, not the pcm2704.

The pcm1798 seems a good candidate (less expensive that the pcm1794 but still very good).

Since you have a balanced output, I would use the fully differential OPA1632 as I/V stage, not classical opamps. You will need an usual opamp afterwards as filter and unbalanced converter.

I would try to get the best power supply possible rather than an external digital filter. Two small transformers with dual secondaries would not be very expensive. One transformer for the digital +5v and +3.3v and one for the +/-10V for the opamps.

For the opamps, a bare minimum would be +/-5V. Check the LT1964 and LT1763 for the opamps, the reg101 for the rest.

Pcb are rather expensive. Factor the cost in.

 

[Clutz]

Quote:

Originally Posted by 00940 For I2S you need the pcm2707, not the pcm2704.

Oops.

Thanks!

Quote:

The pcm1798 seems a good candidate (less expensive that the pcm1794 but still very good).

It does seem like a good candidate- but it's only about $7 less, and that into the cost of the rest of the project, I think would disappear.... though, perhaps the extra specs of the 1798 would be missed on such a design?

Quote:

Since you have a balanced output, I would use the fully differential OPA1632 as I/V stage, not classical opamps. You will need an usual opamp afterwards as filter and unbalanced converter.

What are the benefits of using a fully differential amp for the I/V stage instead of classical op-amps? The benefit I can see from using classical op-amps, is I can start out with some relatively cheap 134s and then upgrade to better op-amps later? (I'm not against using a filly differential opamp, I'm just curious).

Quote:

I would try to get the best power supply possible rather than an external digital filter. Two small transformers with dual secondaries would not be very expensive. One transformer for the digital +5v and +3.3v and one for the +/-10V for the opamps.

That's what I was thinking of doing. One transformer with dual 5 volts out for one, and the other with something like +-12 to 15?


[QUPTE]Pcb are rather expensive. Factor the cost in.[/QUOTE]

Alright. I was thinking about doing the boards with ExpressPCB. Do you think that this will result in a reasonable compromise between DaKiller's and the recent PCM270x DACs with my budget?

Cheers,
Clutz

 

[00940]

Quote:

Originally Posted by Clutz What are the benefits of using a fully differential amp for the I/V stage instead of classical op-amps? The benefit I can see from using classical op-amps, is I can start out with some relatively cheap 134s and then upgrade to better op-amps later? (I'm not against using a filly differential opamp, I'm just curious).

Cost of two opa134 at digikey : 4.5$
Cost of one opa1632 at digikey : 3.94$

The specs of the opa1632 are in line with the opa627 or better (especially settling time which is WAY better). It is what is used in the Bel Canto DAC2 btw (ths4130, same stuff as opa1632).

Using one chip instead of two saves also on decoupling caps, size of the pcb and so on. Since the structure is balanced, crosstalk is not a concern.

some additional stuff : avoid the pcm1792. It must be controlled by a microcontroller, while the pcm1794/98 (they are pin compatible btw) are easy to control in hardware mode.

 

[Clutz]

Quote:

Originally Posted by 00940 Cost of two opa134 at digikey : 4.5$ Cost of one opa1632 at digikey : 3.94$ The specs of the opa1632 are in line with the opa627 or better (especially settling time which is WAY better). It is what is used in the Bel Canto DAC2 btw (ths4130, same stuff as opa1632). Using one chip instead of two saves also on decoupling caps, size of the pcb and so on. Since the structure is balanced, crosstalk is not a concern. some additional stuff : avoid the pcm1792. It must be controlled by a microcontroller, while the pcm1794/98 (they are pin compatible btw) are easy to control in hardware mode.

Well, consider me convinced.

I'll use two OPA1632s and either a pair of 627s or AD843s.

So I guess now that I have that worked out, I should start drawing the schematic in Eagle.

 

[00940]

old stuff that should help : http://www6.head-fi.org/forums/showt...light=perreaux

 

[Clutz]

Hey 00940,

In the link you provided, it seems as though you were attempting to do something very similar. Why are you taking the 16bit 44.1kHz at 256fs from the pcm2707 and putting it into the AD1896 to convert it to 24bit 96kHz at 256fs to pass that through to the DAC? Why convert the 44.1kHz 16bit (CD audio) to 96kHz at 24bits...? If we can only get 16 bits from the I2S format from the USB Receiver? If anything, wouldn't we then want to resample up to 176.4?

Thanks
Clutz

 

[Garbz]

Whatever you do keep the output stage differential. It also gives you the option of using it as a balanced source later. If you don't stick with the differential OPAMPS, then use 4 OPAMPS for I/V and filter, one for each channel.

 

[Clutz]

Yes, I'll definitely use the differential op amps suggested.

Now my next task is to figure out how to use Eagle.

 

[MegaMeee]

silly newbie question here but what is a differential output and how can it be used to make a single ended or balanced source?

 

[00940]

Quote:

Originally Posted by Clutz Hey 00940, In the link you provided, it seems as though you were attempting to do something very similar. Why are you taking the 16bit 44.1kHz at 256fs from the pcm2707 and putting it into the AD1896 to convert it to 24bit 96kHz at 256fs to pass that through to the DAC? Why convert the 44.1kHz 16bit (CD audio) to 96kHz at 24bits...? If we can only get 16 bits from the I2S format from the USB Receiver? If anything, wouldn't we then want to resample up to 176.4? Thanks Clutz

For the 24bit part, I cannot honnestly remember. I suppose I got caught by the marketing dreams of upsampling ?

For the 96khz, the reason is that the THD figure are better than at higher frequencies, the consumption reduced, the heat reduced and then there is Wodgy's post 5. Since I used an ASRC, the exact output frequency didn't really matter, it was convenient to get a clock for 96khz.

If you want to keep it cheap and simple, scrap the ad1896. What you could do is to make provisions to cut the I2S lines at one point to add a small card with an ad1896 and a good clock later on. Through hole resistors on the I2S lines would be perfect for that.

 

[Francis_Vaughan]

You really have little choice but to use an ASRC. It is either that or a PLL. The ASRC gets you the ability to suck more than one sample rate - and since this is a USB device we can pretty much assume that people will need such flexibility.

The 1794 can only let the system clock get 6 cycles away from the LR clock - if it drifts further the chip resynchronises. You have no way of keeping the system clock this close without either a local clock (and thus you need an ASRC to manage the vagaries of the USB delivered sample rate) or you need a PLL - and thus you adapt to the vagaries of the USB sample rate.

00940's design is spot on. It is exactly what I would look at doing. You could also look at the CS8421 ASRC. Building a PLL is difficult at best.

I would be flexible about the I/V conversion. There are many options.

Also, I will keep emphasising, you need to make this design as a whole. Great care and attention to signal and ground return integrity throughout will be the make or break of it. Making the ASRC an option, expecially a duaghter card will wreck this. The integrity of the clock as it reaches the ASRC is exactly as critical as the clock that reaches the DAC. Any contamination on either will ruin the result. Careful layout is key.

 

[ble0t]

Hrmmpff...I totally forgot about 00940's design from last spring. It seems Clutz is looking for something similar. Do we as a group want to take another stab at this? I for one am quite interested

 

[guzzler]

I'd definately be interested in trying to get it off the ground again.

Another DAC worth considering is the 1738, as used in Cyrus CD players. Fantastic sound, and fairly cheap as well (think it's ~£7)

There is the other option of using the 1793 which has differential voltage outputs. Aos used that in his Ally and Piccollo DACs. It's not got such a good reputation as the 1794/8, but just thought I'd mention it.

 

[Clutz]

I should be on my way to work, so I'm going to try to be quick.

Quote:

Originally Posted by Francis_Vaughan You really have little choice but to use an ASRC. It is either that or a PLL. The ASRC gets you the ability to suck more than one sample rate - and since this is a USB device we can pretty much assume that people will need such flexibility.

People will need this flexibility because people will probably want to be able to play different audio formats through their computer (i.e. other than CD Audio)? (At the risk of offending anyone, I'm just asking to understand, I'm not trying to find a passive-aggressive way to argue).

Quote:

The 1794 can only let the system clock get 6 cycles away from the LR clock - if it drifts further the chip resynchronises. You have no way of keeping the system clock this close without either a local clock (and thus you need an ASRC to manage the vagaries of the USB delivered sample rate) or you need a PLL - and thus you adapt to the vagaries of the USB sample rate. 00940's design is spot on. It is exactly what I would look at doing. You could also look at the CS8421 ASRC. Building a PLL is difficult at best.

When I looked at 00940's design I felt it was what I was looking at trying to do.. It basically looked like exactly what I wanted to do, except it had the ASRC chip.. but I didn't know anything about it.. know that I would want it.. so.. That's kind of good in my mind (reinforces that it's a reasonable thing to attempt doing), but also means it's sort of done.

Quote:

Also, I will keep emphasising, you need to make this design as a whole. Great care and attention to signal and ground return integrity throughout will be the make or break of it. Making the ASRC an option, expecially a duaghter card will wreck this. The integrity of the clock as it reaches the ASRC is exactly as critical as the clock that reaches the DAC. Any contamination on either will ruin the result. Careful layout is key.

My two goals for this were to learn more about DIY electronics, and try to come up with a good robust design by getting lots of people involved in it. I'm still a bit confused about ground plane (and I know we're a ways off that yet), and how we go about having a single ground plane and not have the analogue return polute the digital ground... and yet from the Alf DAC discussion, I recall that two separate ground planes isn't really desirable.


Quote:

Originally Posted by ble0t Hrmmpff...I totally forgot about 00940's design from last spring. It seems Clutz is looking for something similar. Do we as a group want to take another stab at this? I for one am quite interested

Me too. I didnm't know that 00940 already had something so far along, but I was hoping to try and get the group to take a stab at something like this. It'll give me yet another reason not to work on my PhD.

Quote:

Originally Posted by guzzler Another DAC worth considering is the 1738, as used in Cyrus CD players. Fantastic sound, and fairly cheap as well (think it's ~£7)

It looks pretty good too. I think then at the very least, we've agreed that differential output is the way to go?