XMOS-based Asynchronous USB to I2S interface

[staki]

Originally Posted by Lorien

Well, I didnt thought at this when WaveIO was designed but we can find some alternatives for it:
1. You can try to connect all the I2S buses together (WaveIO I2S > DAC1 + DAC2). I'm not a big fan of this option but seems to be the faster I can think of right now.
2. For a "decent" approach, you can use fanout buffers, as you stated above but you have to implement it carefully (PSU, traces, connectors, etc.).
3. Use isolated I2S for one DAC while the nonisolated for the remaining one. Anyway please note that NVE isolator will add delays to all the signals of the isolated I2S port, delays that are NOT synchronized later on with Flip-Flops or other logic with the other I2S port thus this alternative is not the best approach still can be used for testing purposes!
L

For the "decent" approach, won't the fanout buffers increase jitter?
I think that I will use the first solution, provided that the target impedance is sufficiently high, like BrianDonegan says (I probably will use two PCM1794A dacs, eventually two AD1955). And of course make use of cables of the same length for both dac chips.
That means that I will be forced to use the isolated output, because I can't imagine a way to put two coax cables on a single U.FL connector.......

[barrows]

Originally Posted by staki

For the "decent" approach, won't the fanout buffers increase jitter?
I think that I will use the first solution, provided that the target impedance is sufficiently high, like BrianDonegan says (I probably will use two PCM1794A dacs, eventually two AD1955). And of course make use of cables of the same length for both dac chips.
That means that I will be forced to use the isolated output, because I can't imagine a way to put two coax cables on a single U.FL connector.......

Consider that Jeff Roland Designs decided against using multiple DAC chips in their Aeris DAC because they considered it basically impossible to insure that the two chips would be perfectly in sync. One might consider this marketing BS, but he price of the DAC, and the fact that they do use other expensive parts, seems to me to be an indication that they did not base this decision on the additional expense of adding another DAC chip.
Dual Mono DAC chips may create more problems than it solves...

[qusp]

yeah, like a µS difference left to right is going to make a big difference in a real room with real speakers with real filters and real ears. sitting 30cm to the right will cause a similar delay...

[1audio]

The point behind summing multiple DAC's is to get closer to the ideal output. Sync should not be an issue since the output (unless an NOS dac with no reconstruction filter) is low pass filtered octaves below any clock. The multiple DAC trick is nearly 20 years old and works. You should get about a 3 dB noise floor reduction for every doubling of DAC's.

I think Accuphase has a DAC with something like 16 dac chips summed. With more chips the individual errors drop as a proportion of the whole. I could argue that running one DAC on the back side of the master clock from another may be a good thing. With 16 you could make a sort of "ring DAC" with some intrinsic low pass functionality. Not too different from how a SAW filter is made.

Any fast CMOS buffer will have phase noise (jitter) way below any clock you can buy. At 1 KHz any 74AC family chip can easily be -170 dBC phase noise, if you have a reasonably quiet power supply and did not screw up the grounding etc.

[1audio]

Originally Posted by qusp

yeah, like a µS difference left to right is going to make a big difference in a real room with real speakers with real filters and real ears. sitting 30cm to the right will cause a similar delay...

Last I looked 1 uS in air was about 1/1000 of a foot or .02 inch or so. ( maybe 1 mm?). Perhaps gravity is less where you are and things go faster (relativity. . .)

[dahlberg]

Thank you for your kind words! Hard to choose because both are good! Unfortunately I do not own any of them so I just speak from "theory": I guess Placid (Shunts in general) are preferred against series regs. but my first thought would be to use both: TI for series regulation and after that Placid for "fine tuning". Anyway, this seems to be quite overkill for many of us so, the best judge will be your ears. Test your audio setup with both of them and choose the one you like it most.
Warm wishes,
L

Edit: now I see that you already have a LT-based pre-regulator so I would suggest to choose Placid as a starting point in your experiments.

Thanks, It's supposed to be used in a dsp/dac setup and todays brainstorming ended with this.


Maybe I'm skipping the "Placid's" for the dac's since they are already equipped with "tridents", but that's another story so......

[qusp]

Originally Posted by 1audio

Last I looked 1 uS in air was about 1/1000 of a foot or .02 inch or so. ( maybe 1 mm?). Perhaps gravity is less where you are and things go faster (relativity. . .)

I didnt do any crunching of numbers at all, and should have oviously read mS. I had a feeling the word similar would bite me . what I mean is there are bigger things to worry about.

[staki]

Originally Posted by barrows

Consider that Jeff Roland Designs decided against using multiple DAC chips in their Aeris DAC because they considered it basically impossible to insure that the two chips would be perfectly in sync. One might consider this marketing BS, but he price of the DAC, and the fact that they do use other expensive parts, seems to me to be an indication that they did not base this decision on the additional expense of adding another DAC chip.
Dual Mono DAC chips may create more problems than it solves...

Yet there are quiet a few examples of commercial or diy dacs that work with a dual mono implementation of two dac chips which share the same i2s signals without any buffers. And they seem to operate properly.......
Then where is the truth?

[staki]

Originally Posted by 1audio

Last I looked 1 uS in air was about 1/1000 of a foot or .02 inch or so. ( maybe 1 mm?). Perhaps gravity is less where you are and things go faster (relativity. . .)

it would be interesting to know what delay a difference of 1 or 2 mm in cable length causes, depending on the speed of signal propagation in copper......

[wlowes]

I believe it is 6ns per foot.