usb dac
Nov 17, 2004 at 12:41 AM Post #391 of 519
Here's what could be very well be my next USB dac. It won't see the light before june or july though.

pcm2707%203.jpg


PCM2707 => I2S => PCM1794 => Passlabs D1 I/V stage (minimalist, without the buffer)

Regulation onboard for the pcm2707 and pcm1794. Outboard regulation for the gain stage. The output caps are outboard too. Balanced to unbalanced will be done by a dynalo that will double as preamp. Bypassing of the IC by 100uF oscon + ceramic (solderside).

Size is half an eurocard.

a.jpg
 
Nov 17, 2004 at 2:00 AM Post #392 of 519
so...Would this be a better D/A conversion than the one in my ADCOM GCD-700? Dual Burr/Brown 20bit D/A converters? Thanks

Not the above DAC but the familiar board you all have built.
 
Nov 17, 2004 at 3:08 PM Post #394 of 519
00949 - please please please dont split your groundplane. The signals crossing the bridge need a return path and thus will create more noise on the board. You have two options, either use a continuous plane and keep analogue and digital signals away as far as you can physically move them or, crate a bridge between the planes (and only one to avoid loops) directly below those signals.

If return current cant flow beneath signal wires, it will only go elsewhere on your board and create noise
frown.gif
 
Nov 17, 2004 at 6:19 PM Post #395 of 519
The grounding question is what I'm working on right now. I've read about anything and it's contrary on that topic. The only thing I'm quite sure of is that the DAC must be completly on the analog plane.

Option 1 : one single groundplane
Option 2 : two, joined by a ferrite bead somewhere
Option 3 : two, joined by a trace under those resistors carrying the I2S signal
Option 4 : two, joined at the power supply
Option 5 : two, with the I2S passing through some isolating device such as the ISO150 (costly and big)
Option 6 : two, joined by a pair of schottky's, back to back
Option 7 : two, joined by a resistor

Honnestly, for this, I'm still researching what's best.
confused.gif


The power supply will be made out of two transformers. One 2X5VAC for seperated supplies for the 2707 and 1794. One 2X30VAC for the I/V stage.
 
Nov 19, 2004 at 2:43 PM Post #396 of 519
Option 1 : one single groundplane
In my opinion, the best option. Just keep the analogue and digital traces as far apart as possible


Option 2 : two, joined by a ferrite bead somewhere
Bad plan for digital signals. The I2S will lose its sharp edges and you will likely see jitter as the sample point moves. Your circuit should be designed to handle up to the 5th harmonic of a signal with minimal losses to get a good edge to sample. This rules out a ferrite connection.

Option 3 : two, joined by a trace under those resistors carrying the I2S signal
OK option, but power supplies should be isolated for each board, or the power connections should also be made at or around the same point. The reason for this is that both ground and power can carry a high frequency signal. It doesnt care which plane it jumps to as it is ac

Option 4 : two, joined at the power supply
Bad idea, the return path will be a big ground loop that will run round the gap and transmit noise onto the entire board via the power supply. Also, very poor emc and esd protection as the high frequency created on one plane may just jump across the planes using the I2S wires and fry the chips

Option 5 : two, with the I2S passing through some isolating device such as the ISO150 (costly and big)
Good option, I suggest two boards are used with completely isolated power supplies or a single board with three islands and one psu for each. The ISO150 sits as in the middle island and only allows the I2S to hop.

Option 6 : two, joined by a pair of schottky's, back to back
Bad idea, there is no low voltage DC return path. Noise will be created at a level equal to the voltage drop across the diode on each board. You will most certainly hear this on the output

Option 7 : two, joined by a resistor
Similar to option 6, we're looking to lower impedance, not create it


What you have to remember is this - any signal will take the path of least impedance. Note that two ground planes near to each other form a capacitor, a length of track is a resistor (and inductor) and a via is a bigger inductor and little resistor. A signal will take the easiest path, wherever it has to jump. This will depend on frequency.

The best board is a 4 layer board with central ground and power planes. This allows return current to flow beneath any signal thus giving the smallest possible loop area. By minimising loop area, we minimise coupling between signal lines. By keeping a distance, we prevent crosstalk at the signal decreases proportional to a power of distance from the track.

I suggest the following

1 - Use a 4 layer board (or 5 layer for two supplies, alternatively split only the power plane (2nd layer from base, make the ground the third from base and run all important signals on the top layer thus next to a complete plane)
2 - Keep power and ground on central planes
3 - Decouple the supplies next to each chip
4 - Keep Analogue and Digital as far away as possible
5 - Ensure separation between signal lines
6 - Match the length and impedance of the I2S lines for the best signal quality
7 - Place in a metal box, but only provide a single starpoint to the box for all grounds to avoid loops forming
 
Nov 19, 2004 at 4:56 PM Post #397 of 519
wow, I didn't expect such a nicely detailled reply
eek.gif
Thanks a lot
smily_headphones1.gif


- 4-5 layers boards : not an option for me, I need to keep everything on two layers because I'm using eagle freeware and because it'd be awfully expensive to produce in very low series. The project is intended to be purely a one-shot project. Perhaps two or three other boards done but that's it.
- decoupling as close as possible : I'm trying to. For now, I can hardly do closer. What would you use for decoupling ? I'm planning 47-100uF electrolytic coupled to a 100nf ceramic.
- I think I'll go with the common groundplane. Iso 150 are too much of a problem to get working
 
Dec 3, 2004 at 12:12 AM Post #398 of 519
Does anyone have a part # (preferably Mouser or Digikey) for a pad/cushion for the crystal used in this DAC? I haven't been able to find one on either site. Or is there a more DIY method?

EDIT: Typical...right as I post it, I find the Mylar spacer in the catalog. For anyone else who is curious, the part # at Mouser is 520-700-9001.

I guess I have a side question...what discernable difference would I see with/without something?
 
Dec 3, 2004 at 12:29 AM Post #399 of 519
Quote:

Originally Posted by ble0t
Does anyone have a part # (preferably Mouser or Digikey) for a pad/cushion for the crystal used in this DAC? I haven't been able to find one on either site. Or is there a more DIY method?

EDIT: Typical...right as I post it, I find the Mylar spacer in the catalog. For anyone else who is curious, the part # at Mouser is 520-700-9001.

I guess I have a side question...what discernable difference would I see with/without something?



wait - is one even necessary?
 
Dec 9, 2004 at 5:31 AM Post #400 of 519
Well, I got my DAC assembled, but I'm having an issue and I was hoping someone had some ideas...

If I use two AA batteries (i.e. 3.15V) I get 1.75 volts out of the MAX1722 and if I use the USB supply (~5V) I get 3.75V out of the MAX1722 . Obviously something is a miss at or before the converter. I did some measuring around the board and the voltage drop over R10 (10R resistor) is close to 2V when using batteries, which obviously isn't a good thing. I assume this would be the sign of a faulty or shorted MAX1722? It doesn't appear that any traces are bridged, but I wanted to get some opinions before I started desoldering it. Thanks in advance.
 
Dec 9, 2004 at 2:23 PM Post #401 of 519
Quote:

Originally Posted by ble0t
Well, I got my DAC assembled, but I'm having an issue and I was hoping someone had some ideas...

If I use two AA batteries (i.e. 3.15V) I get 1.75 volts out of the MAX1722 and if I use the USB supply (~5V) I get 3.75V out of the MAX1722 . Obviously something is a miss at or before the converter. I did some measuring around the board and the voltage drop over R10 (10R resistor) is close to 2V when using batteries, which obviously isn't a good thing. I assume this would be the sign of a faulty or shorted MAX1722? It doesn't appear that any traces are bridged, but I wanted to get some opinions before I started desoldering it. Thanks in advance.



I had the same problem on the second one I build. It turned out that two pins of the dac were soldered together. Another option may be to replace the max1722, which fixed my first usb dac. Good luck!
 
Dec 17, 2004 at 11:16 PM Post #402 of 519
Sorry for the thread necromancy, but I'm looking for some assistance in building one (actually two) of these. I am having a very difficult time finding a source for the 2.5mm pitch WIMA MKS-02 0.1uF capacitors and the max7122 converter. Mouser, Digikey, and Newark either have them out-of-stock or isn't a valid part number. Can anybody suggest equivilants that could be used in their place?
 
Dec 18, 2004 at 1:22 AM Post #403 of 519
Hehe, good luck with the Wima's. I got mine from Newark a few days ago, but there's a 17 day lead time on them; they have to ship them in from Farnell in England, and with the additional shipping cost, they ran me well over a buck apeice....
 
Dec 29, 2004 at 7:47 PM Post #404 of 519
Quote:

Does anyone have a part # (preferably Mouser or Digikey) for a pad/cushion for the crystal used in this DAC? I haven't been able to find one on either site. Or is there a more DIY method?

EDIT: Typical...right as I post it, I find the Mylar spacer in the catalog. For anyone else who is curious, the part # at Mouser is 520-700-9001.

I guess I have a side question...what discernable difference would I see with/without something?



Quote:

wait - is one even necessary?



Does anyone have an answer for this? I've looked at some computer motherboards and various pc cards and only half the time the crystals have plastic pads underneath.

What kind of performance benefits will damping pads bring to this specific application - or more aptly, how will the sonic performance be degraded by omitting it? I just built one without it and while I still can (haven't hooked it all up yet) I'd like to know if it's going to make a difference or not here.

Thanks in advance!
 
Dec 30, 2004 at 12:50 AM Post #405 of 519
Quote:

Originally Posted by doobooloo
Does anyone have an answer for this? I've looked at some computer motherboards and various pc cards and only half the time the crystals have plastic pads underneath.

What kind of performance benefits will damping pads bring to this specific application - or more aptly, how will the sonic performance be degraded by omitting it? I just built one without it and while I still can (haven't hooked it all up yet) I'd like to know if it's going to make a difference or not here.

Thanks in advance!



No, it's not necessary and you won't notice any difference in performance. The sole purpose is really to isolate the X-tal from too much heat, and from mechanical strain. Nothing to worry about on this front. The only thing is, they're very cheap so I thought I'd just include them anyway. Certainly not worth paying 8p for one, and £3.75 for shipping!

g
 

Users who are viewing this thread

Back
Top