[00940]

It is is quite likely that I'm going to have to move abroad again, for quite some time. Moving gear in between Belgium and France was easy but Toronto or Boston is a different business. It's not for tomorrow morning ( should be Fall 2015) but I'm already contemplating ways to simplify a lot my setup.
 
What I'm considering for now is a all-in-one USB DAC-amp for my HD650. On the digital side, I already have a pcb with pcm2707-src4192-pcm1798 I can hack (an old spare of my current DAC). I'll use a different USB receiver to have a 24bit input, to control the volume purely in the digital domain.
 
On the analog side, here's a possible solution. The I/V stage part of it works well, I'm using the same right now. The opamp feeding vcom is used to null the voltage at the input pins of the opa1632, or rather at the output of the DAC. The output section is nothing special but should be decent enough. It should hit 3Vrms when powered from +/-8v (to keep heat low, the opa1632 can get quite hot at higher voltages). It's enough in my experience with the HD650. With a src4192 upsampling everything to 96khz, in front of the pcm1798 with its 8x oversampling, filtering can be kept minimal.

 
I've drawn a preliminary pcb for a channel. It's 50*32mm, dual sided. All smd parts are 1206. It requires just one jumper running under the board, from the output of the DC bias opamp to vcom.
 

 
 
 
Obviously, comments on how to improve the pcb are more than welcome. 

In particular, comments on how to deal with the grounding of R5 would be useful. At first analysis, I'd run it by a dedicated wire to the star ground of the +/- power supply.

 

[nikongod]

R5 goes to ground on the board. Do not return it to the star ground point. 
 
Since the grounds will probably be tied together at the output jack (there I go assuming) perhaps you could save space by sharing caps and power & signal connectors between the 2 channels. 
 
I have never been a fan of the DIP8 BUF634P. The TO220 version BUF634T allows heatsinking, and IMO just works better for it, especially if the bandwidth pin is tied to V-. The DIP8 version kind of lives on the edge of thermal protection mode when run like this and I think it suffers for it. You *might* get away with gluing a small heatsink to the DIP8chip, but why bother when the TO220 is so much nicer? 

 

[00940]

  R5 goes to ground on the board. Do not return it to the star ground point. 
 
Since the grounds will probably be tied together at the output jack (there I go assuming) perhaps you could save space by sharing caps and power & signal connectors between the 2 channels. 
 
I have never been a fan of the DIP8 BUF634P. The TO220 version BUF634T allows heatsinking, and IMO just works better for it, especially if the bandwidth pin is tied to V-. The DIP8 version kind of lives on the edge of thermal protection mode when run like this and I think it suffers for it. You *might* get away with gluing a small heatsink to the DIP8chip, but why bother when the TO220 is so much nicer? 

 
Thanks for the comments 
 
 
- Well, R5 is the only ground on the board (the headphones return wire going directly to the star ground). Connecting it to the ground on board means connecting it to the bypass caps return wire. I guess I should draw a complete block schematic with the considered grounding scheme.
 
- Actually, you raise a very good point. I've been hesitating to recable the hd650 to a 4-pins xlr, more to get dual mono than balanced. If I did so, then it would make a lot of sense to return everything to the board rather than to the supply.
 
- While I like the buf634t, the dip version made the layout of the power supply quite simple and logical. At +/-8V, dissipation should be about 1/4W with full BW, and I could limit current a bit to reduce it. Blah, I will see tomorrow if I can get a to220 version of the board to work in Eagle.

 

[Avro_Arrow]

I made a custom part for the TO-220 BUF634 in EAGLE.
I spread the pins out a bit in-line. Both the zig-zag footprint
and the stock pin-in-line footprint leave little space
between the pins.
 
Like this:
 

 

[00940]

The default libraries in Eagle aren't that good indeed. I did something similar.
 
What I really wonder is how necessary is a heatsink with a buf634t on +/-8V, for 300r cans. If I avoid using one, all paths can be made very short.
 

 
Don't mind the top right components, they're part of a local reg.

 

[nikongod]

The TO220 version of the BUF634 has better heat dissipation than the DIP8 just for the exposed tab. 
 
It looks like the tab of the BUF634 is above IC1. You can probably add a small heatsink later. Even something as jerry-rigged as a small piece of copper sheet cut to fit.

 

[00940]

Thanks for the comments and the help. I decided to finally go for a more modular approach, based on small 50*50mm modules. I'll take my time going module by module. The target being to fit everything in 40mm high, 230*280mm galaxy case (a transportable setup).
 
First module is the same I/V but with a simple INA134  based bal-se converter at the output. The INA134 can be omitted and, with four small jumpers, you get a balanced output instead. Next module will be the headphones buffer output section. And then power supplies and maybe a redraw of the DAC section. For the usb receiver, I'll use the isolated cm6631a from diynhk, which is also 5*5cm.