USB powered dac-amp musings

[MisterX]

You may want to check the spec for capacitance between Vbus and GND.  

 

[Avro_Arrow]

  You may want to check the spec for capacitance between Vbus and GND.

Isn't it about 100uF?

 

[00940]

Duh, good catch MisterX.  The max is about 10uF. The first cap should be 4.7uF, there's a mistake in the schematic. The second cap with a 22R resistor in serie can be 47uF.

 

[Avro_Arrow]

I looked it up in the spec to be sure.
 
Upstream Facing Port Bypass Capacitance: Min - 1uF, Max - 10uF.

 

[Avro_Arrow]

I did a little checking up on what some of the manufacturers I have access to did
on their demo boards for their USB controllers.
 
None featured input capacitance anywhere as low as the 1 to 10 uF in the spec.
They all had anywhere from 470uF to nearly 1000uF across the USB power line.
They also used 330R ferrite.
 

 

[00940]

  I did a little checking up on what some of the manufacturers I have access to did
on their demo boards for their USB controllers.
 
None featured input capacitance anywhere as low as the 1 to 10 uF in the spec.
They all had anywhere from 470uF to nearly 1000uF across the USB power line.
 

 
Thanks for relaying the info. Still, I'd feel at ease with a 47uF with not so great ESR... 1000uF clearly not.
 

 They also used 330R ferrite.

 
That seems like a sweet spot from the sims I ran this evening. I'll be using BLM21PG331SN1D ferrites rather than AG for decoupling. The higher DC rating is actually an advantage. Sims also made me reconsider the bypassing btw. I'm gonna switch to 33uF polymer (nichicon NS) in // with 2.2uF X7R on all supply pins of the pcm5102. 
 
Time to rework the board now. While I'm at it, I'll switch to those Micrel regs everyone seems to love these days. And to the dlp31s filter, you convinced me on that one.

 

[MisterX]

It's not uncommon for people to ignore the spec.
I have a USB hard drive enclosure that has a bunch of big caps, no inrush limiting and it causes all of the other USB peripherals to freak out every time I plug it in.

 

[00940]

Well, a crude inrush limiting would be easy to add. The PCM2707 will work down to 4.35V. Even allowing for some voltage variations, we can safely drop 200-300mV for an input RC filter. Since the total current draw of the device is about 70ma, we can add a 3.3 resistor in line with the Vbus input. Doesn't look like much but it would certainly allow the use of a 47 to 100uF cap at the input.

 

[MisterX]

If you're going in that direction why not just add a load switch?
Usually only 1 IC and a couple of small caps gets you more protection features then a RC circuit, albeit with a higher cost.
 
http://www.ti.com/lsds/ti/power-management/power-mux-current-limited-switch-overview.page

 

[00940]

A ferrite, an OSCON and a MLCC walk into a bar...
 

 
Letting aside the parasitics to make the point clearer:
 
Top is 300R ferrite + Pana 20SVP22M (22uF/20V polymer)
Middle is 300R ferrite + Pana 20SVP22M + 100n X7R
Bottom is 300R ferrite + Pana 20SVP22M + 2.2uF X7R
 
Time to kiss those 100nF caps goodbye.

 

[DingoSmuggler]

  A ferrite, an OSCON and a MLCC walk into a bar...

The bartender doesn't react!

 

[00940]

Tempted, awfully tempted... 
 
This case is quite nice and flat : 
 

 
And if I fit everything onto a 85*60mm board (everything means adum4160, dc-dc converters, pcm2707, pcm5102 and amp section), panel mounting only the jack and a rk097, it works. Barely because there's only 10mm usable in between the pcb and the top of the case (which is why the rk097 won't fit on the pcb). But barely is enough. 

 

 
 
 
This wouldn't be "dragonfly small" but 72*100*19mm isn't too bad. It's more compact than an audinst hub-mini for example.

 

[00940]

The project isn't dead yet  

 
- The case above was really a bit too thin. I'll revert to the more conventionnal Hammond 1455C1201. Let's be a bit practical for once.
- Isolation is nice and all that stuff but I decided to keep it more simple. After all, it is made to be used with headphones so there's no risk of groundloop. I'll buy a Doodlebug for a home setup if I feel the need.
- The njm4556 in an opa134 loop is definitely a bit marginal. I'll go for the faster lm6171 (vfb) or lme49713 (cfb) as buffer. Those can push just as much current. The lme49713 has the edge for stability in sims.
- I also decided to go for a more modular approach. One board for the dac, one board for the amp. If I mess with one section, I won't waste all the pcb 

 Both boards are 50*50mm, with 3.5mm clearance on each side (to fit in the hammond enclosure)
- I found a more compact DC-DC converter, the Murata MTU2. It gives me +/-12V. I'll filter the lines heavily but, thanks to the good psrr of the opamps at audio frequencies (and especially LF), I don't feel the need to regulate the supply.
 
So where am I (after some help received on Diyaudio from jcx and Avro Arrow)...
 
Headphones amp section:
 

 

 
DAC section:
 

 

 
 
There might some last small changes on the pcb but now I'm almost ready to send the files to oshpark.

 

[Avro_Arrow]

Looking good!
If you have some spare boards I would be interested in a set.
I should have a Carrie Amp v3 prototype as well to compare
it with.

 

[00940]

I plan to order from Oshpark so I'll receive three boards of each. I can keep one set for you. In fact, the DRC works with 10mils clearances for both the DAC (at the exception of the I2S under the 0805 resistors) and headamp. So even Itead would probably be ok. That would be a bit cheaper than Oshpark but for 10 boards instead of 3. Still, I don't need 10 boards, 2 would be more than enough, and Oshpark is supposedly more reliable. And the ability to directly upload *.brd is nice too.