USB powered dac-amp musings

There are of course plenty of commercial dac-amp combo on the market right now, many of these powered by USB. Some are even quite cheap and decent. But since DIY isn't about saving money but about having fun, I'm starting to gather ideas to build my own. 
 
The aim would be to have a smallish box I could take away on vacation and the like (not really a portable amp thus) and which would be able to power HD650 (300r) and HD25-13 (600r) somewhat properly. I don't want batteries nor a power supply. Just usb in, headphones out.
 
Soooo:
 
- Case: the aim is to fit everything into a galaxy GX147 (124*170*40mm).
 
- Power supply: the digital boards required 5V, with regs onboard to take it down to 3.3V. I'll probably simply run the USB power through a lclc filter. For the amp section, a DC-DC converter is pretty much mandatory. Probably the NMH0509DC from Murata, followed again by LC filters. The datasheet promises 5mVp-p ripple with an additional filter. +/-9V rails should be ok for the hd650.
 
- USB receiver: if one wants to build his own board, only the pcm270* serie is easy to buy as parts in small quantities. Otoh, there's a lot of built boards for sale with xmos, te7022, cm6631a or sa9023. A bit less diy but those offer support for 24bits and higher sample rates than the good old TI parts. For now, I'm considering this board. The SA9023 doesn't require extra drivers and this particular pcb provides pads for a 5V external supply while having a ldo onboard. I'll remove the spdif section to save a tiny bit of  power.
 
- DAC: the es9023 has the great advantage to be able to run in async with its own master clock, reducing the reliance on the quality of the incoming I2S lines. Problem once again is that the es9023 isn't easy to buy in small quantities... built board time again. This one looks half decent for this application. A local clock is onboard. I'll have to unsolder those big rca. 
 
- Volume control: hesitating... in theory, with 24bits support, I should be able to control the volume from the computer, in the digital realm. But, from experience, it's easier to make mistakes with a digital volume control and a good old pot is safer. I might use a blue velvet in there.
 
- Amp: I was considering the tpa6120a2 in multiloop.. but current draw is a bit high at iddle and I don't really need its current abilities for my high impedance cans. I'm thinking about using a njm4556 inside a opa2134 loop. Why the opa2134 ? To interface the amp inputs easily with an analog potentiometer (jfet inputs...) and so that I can put RC filters in the opa134 power lines, to increase the power supply ripple rejection of the whole amp. Something like this:
 

 
 
What do you guys think ? Something I overlook, something you would do differently ?

Is the 4556 fast enough to be in a loop with the 2134 driving it?

avro_arrow said:

  Is the 4556 fast enough to be in a loop with the 2134 driving it?

Click to expand...

 
"The" question to ask indeed. Clearly not if the 2134 is run at unity gain and without compensation.
 
I had a quick look in ltspice, with the models provided by TI and NJR:
 
* with both opamps at unity gain and no compensation caps, we've got an oscillator.
* with a gain of 3 for the opa134, the njm4556 at unity and no compensation caps, the gain margin is only 45°
* with a gain of 3 for the opa134, the njm4556 at unity and a 22pf compensation cap, the gain margin is a more comfortable 65°
 
The frequency plot doesn't show any peaking either with the last config.
 
 
This being said, I should check how close the models actually are to the datasheets gain and phase plots.
 
 
A big advantage of the njm4556 over the tpa6120 in this case is that layout requirements are also very much relaxed.

The components are cheap...just build it and see what happens.
If you were using expensive components, you might want to check
more closely. But, as you said, part of the project is to have fun!

Indeed. I need to order a opa2134 though.
 
I'll etch a board tomorrow and try with a tl072 to check for basic functionnality while waiting on the opa2134.  
 

It's working without a hitch. No noise into sensitive earbuds, powered from a cheap wallwart and a simplistic passive rail splitter. Bad tl072 are giving me some highish offset (25mv) but otherwise, everything is fine.
 
I'll wait for the opa2134 to run true stability  tests.

Subbed. Interested in how your project will turn out.
 
If you want to try another differential IC amp, the MAX97220A works well in lieu of the TPA6120A2 as well.

tomscy2000 said:

 
If you want to try another differential IC amp, the MAX97220A works well in lieu of the TPA6120A2 as well.

Click to expand...

 
Thanks. I had a look. It indeed seems like a nice chip. 3Vrms into 600r is pretty good from a 5V supply and the rest of the specs is good too. I'll keep it in mind for a future project. It could be nice trying it at the output of a wm8740 or ak4396.
 
 
 
Of course, the more I kept looking for the best dac to use, the less I was satisfied with what I found. Which brings me to this: 
 

 
It's a quite simple pcm2707-pcm5102 board in the works. It's still missing the regulator section for the DAC. As it is, the pcm2707 is set for bus-powered. It's quite amazing to now have 2.2uF X7R caps in 0805 package.
 
The current plan is to hack a hifimediy USB isolator open and use it to power the DAC board. The amp section will be powered through the Murata nmh0509dc directly from the USB bus.
 
 
Power "budget":  
 
- The dac board will draw 70ma (25ma for the pcm2707, 35ma for the pcm5102, 10ma for various stuff). The hifimediy dc-dc converter (a b0505ls) is about 70% efficient. This will translate into a 100ma load on the USB bus.
- The amp will draw 20ma at iddle (10ma for the opa134, 10ma for the njm4556). This is at +/-9V. It translates into 72ma at 5V. The converter being only 80% efficient, almost 100ma in reality. 
 
Total: 200mA@5v, which is really fine, even for laptops. I've just to be careful with the inrush current.

I wondered: has anyone real world experience with the PCM5102 vs 5101 ? The extra db of THD isn't worth speaking about and the extra 6db of SNR risk of being lost anyway in an usb powered design.
 
 
Anyway, here's the board, kinda finished. It's 50*40mm and should fit easily in a small hammond with room to spare. The caps are all 0805, ferrite beads are 1206.
 

 

Hey, your USB input filter looks just like mine...I guess great minds think alike! 
Actually, I have a TVS diode in my filter as well...
 
Nice, compact design...Looks good!
 

I must confess I took the inspiration for the input filter layout from marce on Diyaudio and the schematic from browsing dozens of evaluation boards   The TVS is a good point. Soldering the murata DLW21SN common mode filter will be fun though... 

I used Murata DLP31S for my CM choke and NXP PRTR5V0U2X for my TVS.
 
Oh, and my L are BLM21.

The DLP31S should be easier to solder by hand, it's about 150% bigger and has pads going up the sides. It is also 3 to 4 times the price of a dlw21s (depending on qty). If you get boards soldered by a factory, there's no small savings  
 
In which variant do you take the blm21 ? I'm on the fence for the ferrites. Mouser has Laird LF ferrites , with an impedance starting to climb much earlier on. Those could be interesting for local decoupling.
 
Thx for the TVS reference, I'll look it up.

BLM21AG
 
I picked the DLP31S because I thought there was at least a chance that the average Joe could solder it on...
When designing for a kit, I have to consider other peoples soldering skills (or lack there of...)
Same with the TVS.
 
I'll have a look at the Laird...thanks for the tip.

avro_arrow said:

  I'll have a look at the Laird...thanks for the tip.

Click to expand...

 
Not sure about those... Well, when in doubt, let's ask .