MINI PCM2704 HI-FI USB DAC SOUND CARD, No Brainer?

 
I needed a usb dac for headphone use from my laptop for some traveling. I did not want to spend to much, as I will not be using it much, so I decided to try one of these from ebay.
 
 
http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=260672993495&ssPageName=STRK:MEWNX:IT#ht_1883wt_905
 
 
 
The PCM2704 is a 8x oversampling delta sigma chip that is used in products such as the Musiland MD10, and the Trends Audio UD-10.1.
 
 
Cold out of the envelope it sounded much better than I expected, nice rich body, but still a bit subdued, not much air but pleasant nonetheless. For a throw away 13 quid, (20 us) its a no brainer.
 
But I thought it might be worth souping it up a bit with some simple mods using spare parts from the shoebox, see if it could be improved.
 

 
 
Just some cap swaps nothing to heavy. MKP for output instead of those electrolytes, and a couple of decent reservoir caps for the power supply and dac chip. Boxed it up with a 3.5mm stereo socket and RCA outputs and left it to cook overnight.
 

 
Well its going to take a week for those MKP to burn but already the sound has more authority and detail whilst retaining that richness. Good extension and balanced tonality, bass is clean as well. It can pass now for a very decent main system dac, sounding way above its price point.
 
I will see how much more it opens up after a week, but it does not sound at all like a dac that cost $20 US and a few parts for the price of a beer.  
 

Umm. This might be a stupid question, but aren't caps supposed to be of certain rating instead of just, well, "bigger" like you used?

Good question tuoppi, actually the output coupling caps, stock, are 470uf (talk about overkill), when in fact 2,2uf is enough for this purpose. The other 2 caps I am using are larger than the original, but the new caps have lower ESR (esp the Oscon SEPC) so can provide more energy more quickly than the stock. I think the SEPC had the biggest impact, it gave more authority and realism.
 
Today I am bypassing them both (the electros) with low value polystyrene as I think the sound is a little too polite (could be the delta sigma OS thing). Otherwise the soundstage is opening up more and its less dry. But it already far surpasses the (use once and dispose of) dac I thought I was buying. Great little thing to experiment with, esp for anyone tempted to use a soldering iron for the first time. The other thing I might try in a few days is to swap the polypropylene coupling caps for Mylar (polyethylene). Mylar is inferior but some believe more natural sounding.

I bought one of these for about $18 including postage, but I think there is a vital circuit element missing from the PCB! The PCM2704 data sheet shows a couple of components directly coupled to each output pin (a 16R in series with a 22n cap to ground) prior to the output caps, but these are omitted from the board. This arrangement strongly resembles the Zobel network often required to keep power amplifiers stable when driving reactive loads.
 
I found that as soon as I connected my headphones (via approx. 8" wires soldered to the underside of the PCB connector to a 1/4" stereo jack), there were terrible hisses and sizzles that changed slightly if I touched the headphone ground onto the USB shell (theoretically the same ground). The unit seemed to work as a line level driver OK, however. I happened to have a 47n cap to hand and found that bridging this between the +ve end of one of the output caps (i.e. one of the IC's output pins) and headphone ground killed the sizzles in that channel. Unfortunately I didn't have the 16R/22n components to hand and so have not tried the definitive test yet.
 
Has anyone else noticed this problem?
 

Hmm seems you may have a problem with that unit, I tried to repeat your experience but its as clean as a whistle.
 
My current, and best so far, caps replacements, are changing the two electrolytic for Wima 10uf, and the coupling caps for Mundorf MKP 2.2uf.  Next up ERO Mkt 1813 3.3uf as coupling.

Quote:

wood said:

Hmm seems you may have a problem with that unit, I tried to repeat your experience but its as clean as a whistle.
 
My current, and best so far, caps replacements, are changing the two electrolytic for Wima 10uf, and the coupling caps for Mundorf MKP 2.2uf.  Next up ERO Mkt 1813 3.3uf as coupling.

Click to expand...

This item costs US$22 and you want to replace the coupling caps with Mundorf which costs more than the DAC, interesting.

Mundorf MKP 2.2uf are less than half that price. People usually talk about pricey supreme and silver in oil etc, good for pricey gear, but the basic MKP are also pretty good, and cheap.
 
This is a line level soundcard, so to use it for headphone output you need to add 2 X 120 ohm resistors in the signal path to the output to the left and right channels. Adding higher values will reduce background noise but will reduce volume output.

Hi Wood
 
I beg to differ that you should need to add series resistance with headphones. The data sheet shows phones coupled directly via the output caps, but also shows the 'Zobel network' (if that's what it is) in parallel when driving phones or an amplifier. It also shows a 3k3 to ground prior to the output caps when driving phones too. These are all missing from the boards we bought from China. The huge output caps are for driving headphones directly, I think.
 
Maybe my headphones are particularly difficult to drive. I've ordered the bits to do the modification and I'll report back.
 
Here's the PCM2704 data sheet:
http://focus.ti.com/lit/ds/symlink/pcm2707.pdf

Have now fitted the 22n/16R on each output as suggested by the data sheet, and the instability is cured. The circuit happily drives my headphones now to a reasonable volume. Of course, an external amplifier and analogue volume control is preferable to controlling the volume in software, in order to preserve the 16 bit resolution...
 

Nice piece of detective work there Neil Desperandum . Although I don't seem to have a problem with sounds of frying bacon, I will also ground it with 22n/16R.
 
You are right about the coupling caps having large values for headphone use. When I used my headphones (actually Sony ear-buds as don't have my main headphones right now) I had to use the higher value to compensate for considerable bass loss.
However, (perhaps its the ear-buds), I still had to further attenuate the output with resistors before the bass emerged at normal levels.
Interesting that you obviously don't have that problem, and as you correctly point out resistors are not included in the spec for headphone out. More investigation here me thinks.
 
I have settled for ERO MKT 3.3uf for line level use and find it a very acceptable sound even in my main system. For headphone use its definitely clearer than my laptop soundcard. A handy, cheap, pocket thingy.

Hello Wood
 
I'm also listening to the PCM2704 card with a hefty (self-built) amp and good speakers (sorry if that's not quite in the spirit of these forums!) and it sounds rather tremendous - I can't imagine how it could sound any better! I'm going to build one of these cards into the amp so it's just got a USB connection to the outside world (analogue is so passe), and then sit back and congratulate myself on how little money I've spent every time I listen to it.
 
 
 
 

Hi Neil, I just did your mod with the 22n/16r..... WOW!
 
I thought it was a good little dac for the price, but after that mod, it really lowered the noise floor and injected a lot of life. I'm just laughing when I listen and remind myself how much it cost.
 
I think I might grab a couple more, box them, and give them as Xmas gifts.
 
Nice one!
 
 
 

 
Quote:

neil desperandum said:

I bought one of these for about $18 including postage, but I think there is a vital circuit element missing from the PCB! The PCM2704 data sheet shows a couple of components directly coupled to each output pin (a 16R in series with a 22n cap to ground) prior to the output caps, but these are omitted from the board. This arrangement strongly resembles the Zobel network often required to keep power amplifiers stable when driving reactive loads.
 
I found that as soon as I connected my headphones (via approx. 8" wires soldered to the underside of the PCB connector to a 1/4" stereo jack), there were terrible hisses and sizzles that changed slightly if I touched the headphone ground onto the USB shell (theoretically the same ground). The unit seemed to work as a line level driver OK, however. I happened to have a 47n cap to hand and found that bridging this between the +ve end of one of the output caps (i.e. one of the IC's output pins) and headphone ground killed the sizzles in that channel. Unfortunately I didn't have the 16R/22n components to hand and so have not tried the definitive test yet.
 
Has anyone else noticed this problem?
 

Click to expand...

Yes i just received mine and soldered headphone port to output like you and there is terrible background hiss when i connect my grado sr60.  
so from the datasheet is it c11, c12, r5, r6 i need to add to eliminate hiss??
 
also does the LED on your board turn on? mine isnt on

The LED flashes briefly when first plugged in then goes of. I added the 22n/16r between the input on the coupling caps and the common ground pin on the output, and a couple of 120 ohm resistors between the output caps and the positive and negative line out to headphones to attenuate the signal, will be trying it with grado sr60 in a couple of days.

Quote:

wood said:

The LED flashes briefly when first plugged in then goes of. I added the 22n/16r between the input on the coupling caps and the common ground pin on the output, and a couple of 120 ohm resistors between the output caps and the positive and negative line out to headphones to attenuate the signal, will be trying it with grado sr60 in a couple of days.

Click to expand...

ah ok i see the LED flash now
I will need to do the mod asap, its unusable with all this hissing and squealing I've got. apart from that I can already hear it is going to sound better than i thought it would