Peer review of my design (USB DAC/Headphone AMP)

[JJalling]

Hi,
 
can you guys please give my design a critical review before I start the board layout - thanks.
Link: http://stuff.jalling.net/Headphone-TPA6120v3.PDF
 
Principle of the circuit:
The PCM2707 (U1) is a USB -> I2S interface. It has an integrated DAC and amp aswell, but they aren't used in my design.
The I2S audio data is passed through the SRC4193 to (hopefully) lower / remove jitter from the audio stream. I'm not sure if it is really needed, but I will keep it in the design for now.
Final step in the digital domain is the DAC PCM1792A, which is followed by a I/V stage and the final amplifier.

The PIC is used to configure the devices via SPI, as well as to control the volume by measuring the POT and set the volume register in the DAC accordingly.

The 5v from the USB is stepped up to +/- 11-12v in the LT1930, and then regulated to +/-9v and 5v by the three lm317 regulators. A forth lm317 is used to generate the 3.3v for the digital part of the circuit.

How would you seperate the analog and the digital part of the circuit?
Any tips on filtering the output from the boost converter? What about filtering the output from the lm317's - is it sufficient?

Thinking about it, I don't think I can use the LM317 to create 3.3v from the 5v rail because of the dropout voltage, but I haven't tested it yet.

Am I missing something obvious? Please comment.

BR Jonas

[JJalling]

Hi Pippy,

No i haven't seen that, but unfortunately my headphone impedance is 250 ohm, so I don't think it is powerful enough. But I'm very open to other solutions like that.

BR Jonas

[free_electron]

what is all that mucking about with opamps and end stages.. your headphone impedance is 250 ohm.. the pcm1792 can drive that directly !

amd why the mucking about with the i2s signals ? use tha analog output of the pcm2707

you are also violating a tin of USb specs. you never going to meet the 500uA standby current...

[JJalling]

You are right - I didn't even consider that. Basically I've just been following the reference design for the pcm2707 and tpa6120a2.

I will have a look at the USB spec. Maybe it's better to include an external psu.

Br Jonas

[senso]

You might already know, as your pull-up line in the D+ is already tied to the micro, but be sure not to pull it up to soon, or bam fried PCM for dinner.

[ve7xen]

You may be interested to look at the HPDAC designs here: http://www.ecpaudio.com/diy.shtml . The HPDAC 2 in particular is very similar to your design.

As far as comments on your posted schematic:

- LM317 has not very good high frequency rejection, and is kind of mediocre anyway. I would both use a modern low noise regulator instead and include some kind of choke-based filter after the switcher
- Similarly I would add ferrite chokes in the power lines of the digital stuff
- +5V regulator should probably be referenced and decoupled to AGND
- I don't mind the use of the datasheet I/V converter, it will perform well if you have room for it.
- This amp is apparently somewhat unstable, read carefully the application section and follow the recommendations for decoupling, feedback components etc. and regular best practice, and minimize stray capacitance in the feedback loop
- Obviously if you plan to sell these you should care about USB spec compliance, but if this is a one off I wouldn't waste your time, it will work fine

I don't see any obvious problems with the design.

[ovnr]

I agree with ve7xen - the design is pretty nice, but find new regulators - I know TI has some very nice (and unfortunately expensive) low-noise ones, couldn't find them at a quick glance tho. Ferrite chokes are also a fairly cheap way to give you some extra peace of mind in regards to noise.

As for the spec compliance, if you actively intend to follow the USB specs (esp. in regards to suspend current), you may want to ensure your system can run with split power domains, so you can safely power down the analog sections on suspend. Easiest solution would probably be to run the PCM2707 on a completely separate power supply and use a buffer chip in-between it and the rest of the system (so you can tristate the buffer outputs on suspend, to avoid driving input pins on unpowered chips. The PCM2707 may already have this built-in, didn't read the DS too closely.)

I'd personally just not give a damn about compliance as long as it works.

Good luck!

[ve7xen]

I often use LP2985 in designs like this. It's easy to use, cheap, multiple sourced, comes in a good selection of fixed and variable outputs and performance is pretty good. You can do better, but it's a good jellybean LDO.

[Zad]

There are some very good quality parts in that design, I have used them myself. The DAC and headphone amp are both capable of seriously high performance, and are used in equipment whose price is measured in thousands, not hundreds. I have had a play with the Sample Rate Converter (with the same aim as you - jitter reduction) but haven't really had time to get it working properly. It does have a good reputation. I would use a TCXO with as low a jitter spec as you can find. I realise it is getting into Audiophool territory, but if you are doing the job, do it properly. Similarly, I would use separate packages for each of the op-amp pairs rather than a single 4-in-1 package.

I wouldn't power it from USB, at least not the analogue section, it deserves better. The output amplifier is a current feedback amp, with a huge gain at wide bandwidth. This means it can tend to go into oscillation at RF frequencies, and so needs careful attention paid to the physical layout of the PCB.

[JJalling]

You might already know, as your pull-up line in the D+ is already tied to the micro, but be sure not to pull it up to soon, or bam fried PCM for dinner.

I haven't had any problems yet on the breadboard, but I will keep it in mind - thanks.

You may be interested to look at the HPDAC designs here: http://www.ecpaudio.com/diy.shtml . The HPDAC 2 in particular is very similar to your design.

Great, thanks. I have a look at it!

As far as comments on your posted schematic:

- LM317 has not very good high frequency rejection, and is kind of mediocre anyway. I would both use a modern low noise regulator instead and include some kind of choke-based filter after the switcher
- Similarly I would add ferrite chokes in the power lines of the digital stuff
- +5V regulator should probably be referenced and decoupled to AGND
- I don't mind the use of the datasheet I/V converter, it will perform well if you have room for it.
- This amp is apparently somewhat unstable, read carefully the application section and follow the recommendations for decoupling, feedback components etc. and regular best practice, and minimize stray capacitance in the feedback loop
- Obviously if you plan to sell these you should care about USB spec compliance, but if this is a one off I wouldn't waste your time, it will work fine

thank you for your input. I will optimize the filtering before I go any further - probably make a pcb for the PSU and optimize it before implementing it. Also I will find replacements for the LM317s.
I have made a small breakout board for the output stage, and I haven't had any major problems with it yet, but I have read that it can be difficult to keep it stable, as you say.
I don't plan to sell it - I just thought it could be a fun project to do

I have had a play with the Sample Rate Converter (with the same aim as you - jitter reduction) but haven't really had time to get it working properly. It does have a good reputation. I would use a TCXO with as low a jitter spec as you can find.

I have it running on the breadboard, but I don't know if it's worth it or not. Haven't played with all the configuration registers yet. Currently I'm just using a 25ppm crystal oscillator as the prices go up pretty fast with better specs.
As mentioned, I don't know if I want to go with it or not.

I realise it is getting into Audiophool territory, but if you are doing the job, do it properly. Similarly, I would use separate packages for each of the op-amp pairs rather than a single 4-in-1 package.

I'm trying to make a design with fairly high specs, but I still want to keep it out of the audiophool territory. Do you really think it will have a significant impact on the performance, if I use for single opamps? I'm trying to keep the size down, so I don't know if thats even possible yet.

I wouldn't power it from USB, at least not the analogue section, it deserves better. The output amplifier is a current feedback amp, with a huge gain at wide bandwidth. This means it can tend to go into oscillation at RF frequencies, and so needs careful attention paid to the physical layout of the PCB.

I see your point, but my initial requirements was an amplifier without external psu and if I can keep it that way, I will prefer that.

I will update the design, and upload the new schematic later. I really appreciate the inputs from all of you. Thanks!

BR Jonas

[JJalling]

You may be interested to look at the HPDAC designs here: http://www.ecpaudio.com/diy.shtml . The HPDAC 2 in particular is very similar to your design.

Great, thanks. I have a look at it!

That design is indeed very similar to mine - and the reference design 

While searching around I stumbled over this aswell: http://www.diyforums.org/PupDAC. When looking at the schematic you will find that the DAC is powered from a 4.75V regulator, powered from USB. Would you consider that a better solution than mine, where I'm powering it through a 5V regulator from the output of the switching regulator?
Considering that 4.75 is specified as the minimum supply voltage in the DAC datasheet i think it's a little on the low side ... or is it generally ok to do it like that?

How would you create a clean supply voltage (5V) for the DAC from USB?

BR Jonas

[Marco]

Do you think you will be able to hear the difference from resampling? The resampler still has to online adapt the resampling ratio ... this will introduce phase noise no matter how clean the output clock is.

From a high end audio perspective USB audio class 1 is broken to begin with ... it's simply infeasible to make good high end USB audio hardware yourself for windows, only the commercials have the money to get a good driver certified by Microsoft.

[JJalling]

Do you think you will be able to hear the difference from resampling? The resampler still has to online adapt the resampling ratio ... this will introduce phase noise no matter how clean the output clock is.

No, probably not! And the chip is expensive so for now I'm removing it from the design.

BR Jonas