[Wodgy]

The best analog all-in-one choice is the PCM2702. The PCM290x is markedly inferior to the PCM2702 for analog output, but the PCM2702 doesn't have a digital output.

You can (of course) do better by using the S/PDIF output of the PCM290x and a dedicated DAC chip. The PCM3200 is too difficult for DIYers to implement and offers no advantages over the PCM290x if you're just looking for a digital out.

 

[Jupiter]

What about the PCM2704/5/6/7?

Do you have a link for the PCM3200?

 

[Wodgy]

Oops, my mistake. I meant the TUSB3200.

PCM2702 has better analog specs than the newer PCM2704/5/etc. Check the datasheets. All the members of the PCM270x series have much better analog specs than the cheaper PCM290x series.

 

[aos]

PCM2704DB looks like it has SP/DIF out pin, so one could presumably just link it up to any DAC. Higher pin count versions have digital out already decoded but they seem to be missing left/right latch signal without which it may not be possible to connect them to a DAC chip.

 

[Jupiter]

Wodgy and aos,

Thanks for the help.

I'll give the PCM2702 and maybe some of the PCM270x a try.

 

[Jupiter]

I've been looking at the schematic for the PCM2702 evaluation board. Why an external low-pass filter when the PCM2702 has an internal low-pass filter?

 

[Wodgy]

The low-pass filter is "needed" for two reasons. First, the DAC's digital filter only has about -80dB of noise attenuation above the audio band. That's actually pretty good, and there is an argument for this being good enough, but some engineers would want to do better. (In comparison, the 290x series has a fairly crummy digital filter, pushing ultrasonic noise down only to about -60dB, which makes an external analog filter more of a necessity for hi-fi audio.) Second, this is an 8-times oversampling design, meaning serious switching noise will be evident in the output signal at 176.4 kHz and beyond. Most people want to get rid of this, to reduce distortion in amplification components later on, as well as to protect tweeters, though there is a school of thought saying that it doesn't matter.

If you only care about reason #2, you can make the analog post filter a very gentle one, far away from the audio band. If you care about reason #1, then the analog post filter will need to be closer to the audio band. This is one argument in favor of using real DAC chips, rather than these integrated USB devices, but personally -80dB of ultrasonic filtration is good enough to me, and I'd move the analog filter out to at least 90kHz or so.

 

[aos]

Actually only top of the line DAC chips have better than about 80dB of stop band attenuation in the digital filter. You can usually combine them with a standalone digital filter if you do wish to keep using the same DAC chip and yet get a better attenuation.

 

[Jupiter]

Thanks again

Then I'll just skip the low-pass filter for now. Maybe I'll add it later.

 

[Jupiter]

Quote:

Originally posted by aos PCM2704DB looks like it has SP/DIF out pin, so one could presumably just link it up to any DAC. Higher pin count versions have digital out already decoded but they seem to be missing left/right latch signal without which it may not be possible to connect them to a DAC chip.

I had a look at the PCM2704/5/6/7 datasheet. The PCM2706/7 has I2S output in addition to S/PDIF. (Page 18 of the datasheet.)

 

[00940]

just to revive a bit this post, an attempt to a very simple USB-DAC/headamp.

Basically, it's just a pcm2702 with BG at the analog output right into a buffered cmoy, with discrete diamond buffers. It should run from 3 9V batteries. PS of the pcm2712 is regulated by reg102 and the headamp section is powered by +/-9V, spitted by a tl2426 buffered by a buf634.

I just realized that there are no pads to fit a pot. There should be some room just before the output caps for that.

 

[aos]

Interesting layout, you seem to have done a nice clean job of having digital paths short and placing capacitors as close to chip as possible. I am thinking of including such an USB chip with digital out into one of my DACs as an alternate input.

 

[Glassman]

all those burr brown USB chips run of 4MHz masterclock I believe, they are generating clock signal at multiples of Fs with their internal PLLs, but I have some doubts regarding their PLL quality.. however it would be interesting to include KWAK-CLOCK circuit running at say 16MHz, divide it by four using some 74HCxx logic and feed it to PCM2702.. jitter is the one and only thing which I am scared about regarding USB chips..

 

[00940]

Schematic, straight from the datasheet :

Edit : I'm gonna change the layout of the analog section to make it evolve towards this :

The amp section is from Sijosae, as well as the discrete rail splitter. (see butterfly on headwize).

 

[aos]

Some of these USB chips use Burr Brown's SpAct technology, same as with DIR1703E receiver, that claims to drastically reduce jitter. I.e. about 70ps for DIR1703E compared to about 200ps of the likes of CS8412 and AK4117. I have heard from professional engineers that measured output of modern multi-segment DAC (which is less sensitive to jitter) when using DIR1703E and they say that jitter is gone. Now, I don't know how much of those improvements went into these particular USB chips.

One drawback with receiver chip is that is is very sensitive to quality of input signal - ironic, isn't it? My digital cable box with coax out doesn't work at all, while portable CDP from Sony with G protection on has frequent very loud popping sounds as it loses lock for short periods of time (sony has lots of noise on its output when it spins up and down its motor in G prot mode). However, in the USB, quality of signal is probably more tightly controlled by the USB standard and it hopefully always works.