Open-source USB interface: Audio Widget

[borges]

Demian and Oneoclock,

I will add your PSU to the AB-1.12 prototype board. This will be an unassembled board compatible with form factor and module from AB-1.1.

Before I can start with the layout I have a couple questions:
- Can you suggest an SMD version of the diode?
- Are you okay with 350mW SOT23 MMBT4403 and MMBT3905?
- Should eash XO have its own regulator, or share one?
- Should DAC also have a regulator like this?

Børge

Originally Posted by oneoclock

I tested a modified source for the 3.3 vol. analog AB1.1.

[oneoclock]

I checked with an oscilloscope, in every moment only works one regulator and the other is disable. May be worth the same power supply for two clock.

I had at home 100 uF. But regarding simulations, the analog circuit can have a value from 25 uF or higher.

Yes for me SMD diode or what's easiest.

[Turbon]

Would we benefit from a small resitor betweed the junction of R10 and the junction of C9 to AVCC - 2ohms or something?

Just a feeling to have semis without filtering right out...

Brgds

[1audio]

Originally Posted by Turbon

Would we benefit from a small resitor betweed the junction of R10 and the junction of C9 to AVCC - 2ohms or something?

Just a feeling to have semis without filtering right out...

Brgds

That would raise the source impedance at DC significantly. If the circuit is stable there is no downside to directly connecting it to the load.

[Turbon]

Originally Posted by 1audio

That would raise the source impedance at DC significantly. If the circuit is stable there is no downside to directly connecting it to the load.

Yes I agree Demian but if it isn't there is a big chance off oscillation. Does the fed circuits draw the same current whatever happens in the audiostream? What I see is that the circuit is expected to dance along in a predicted realtime manner with the load which is changing fast. A way to calm it down and not follow the load (woman) in every point in time is to give her a bit of resistance...

Brgds

[1audio]

Originally Posted by borges

Demian and Oneoclock,

I will add your PSU to the AB-1.12 prototype board. This will be an unassembled board compatible with form factor and module from AB-1.1.

Before I can start with the layout I have a couple questions:
- Can you suggest an SMD version of the diode?
- Are you okay with 350mW SOT23 MMBT4403 and MMBT3905?
- Should eash XO have its own regulator, or share one?
- Should DAC also have a regulator like this?

Børge

Lets look at the clocking as a critical subsystem. You have two oscillators with an enable that switches between them, a divide by 2 for the micro and the need to send really clean clocking to the DAC. The frequencies of the oscillators are high enough that they can be isolated with a ferrite bead and a .1 cap to probably 60 dB on the power supply.

I would, space and budget permitting, use a single regulator with isolation resistors on both supply and ground for the oscillators, the flip-flop and possibly an output buffer for the clock signals. I would transformer couple the outputs so ground related noise doesn't get into the oscillator system or cause modulation at the DAC input. Transformer isolation on the clock back to the USB subsystem will keep its noise away from the clock system as well.

Using another regulator on the output DAC brings up the question of supply requirements for the DAC. The ES9023 needs a single 3.3V supply. The PCM 5102 needs two separate supplies and a separate regulator for each makes sense.

Neither have much output drive and I think a headphone amp would enhance the usability enormously, making this an equivalent to a Proton with 192KHz support. I would suggest the TI TPA6130A2. There are others but most are harder to get and not better. The TI does need control to function. The TI has a differential input allowing good noise isolation if used carefully. It also has its own charge pump. It does not need a regulator and can run from the 5V supply.

[1audio]

Originally Posted by Turbon

Yes I agree Demian but if it isn't there is a big chance off oscillation. Does the fed circuits draw the same current whatever happens in the audiostream? What I see is that the circuit is expected to dance along in a predicted realtime manner with the load which is changing fast. A way to calm it down and not follow the load (woman) in every point in time is to give her a bit of resistance...

Brgds

Supplying power to an oscillator is like powering a resistor. The dynamics of the load are infintessimal if its working right. Powering a DAC has dynamics and that is where stability is critical. Most regulator circuits do connect the pass transistor directly to the load. Making it stable is not too difficult and with decent spice software its pretty easy to get close on the first pass. But you are right to ask the question and it should be bread-boarded and tested thoroughly.

[1audio]

Feature creep

Modularizing the system (I think I mentioned this earlier as well) and adding other capabilities is really interesting, and a good way to never finish even a part of the project. Constraints are a really good way to focus and really optimize a specific system for a specific task.

Taking the USB interface and mixing it with a premium DAC + SPDIF/AES interface is a different set of problems and also very interesting. If this SPDIF interface Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter can be "integrated' functionally with the USB interface it would be a pretty comprehensive input but a completely different project. if that is of interest the pieces are all here. They just need to be pulled together and someone needs to run with it. its very possible to add pieces and tweak to get tot he level of a Meitner or a Berkeley Audio with the project components on DIY. The work on this and on the fifo are as good as anything I have seen anywhere, better than most. The software quality is also very good.

[rsdio]

Originally Posted by 1audio

Taking the USB interface and mixing it with a premium DAC + SPDIF/AES interface is a different set of problems and also very interesting. If this SPDIF interface Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter can be "integrated' functionally with the USB interface it would be a pretty comprehensive input but a completely different project.

Very interesting consideration, but I would say SPDIF input and USB input are fundamentally different, and should not be combined in a minimalist project. SDPIF is a push clock system requiring reclocking at the output. USB is ideally a pull clock system that can more easily run with the DAC as the master. I'd say that adding SPDIF input would be a huge change, and I personally recommend against it unless you actually want to build a monster.

Edit: SPDIF doesn't necessarily require reclocking, but some mechanism must exist to account for the competing clocks. That means either a PLL to match the input clock or reclocking to match the data to the output rate. Either way, SPDIF requires a complex block that is not needed for USB. It's much more involved than simply adding minor electronics for the input jack. On the other hand, SPDIF output in parallel with the I2S output might be simple, but that might as well be treated the same as adding a different DAC.

[linuxworks]

if anyone's interested, I just got a pcm5102 DAC going, fed from the audio widget's i2s lines. forgive the 'informal construction' (lol):





I wanted an easy i2s source and this sure provided one for me

the ESS is offline (removed the 0R series R's on the pcb) and I'll listen to the TI dac for a while. like the ESS, its direct-out, internal charge-pump and very low cost/component count. I used all the bypasses the spec sheet said to use, which is why so many things are hanging off the chip-carrier

fwiw..