32 channel DAC @ 192kHz for testing: Recommendations?
For some experiments, we'd like a 32-channel DAC @ 192kHz. This is just for lab use so cost isn't a big factor. I was hoping to find some combination of eval boards to do this with relative ease.
There are a few good options here: https://freedsp.github.io/ but none are available to purchase due to chip availability, and I'd like to avoid building this by hand for a one-off.
Possible options:
Third party: https://www.ebay.com/itm/383654346294 - stuff like this exists, I have no experience with them.
4 x ADAU1452 eval boards - straightforward but due to the choice of CODECs I don't think 192kHz output is possible. Same issue with the ADAU1466 it seems, sadly.
1 x ADAU1452 EVB + 2 x ADAU1966 EVB. This might work?
It seems that only 16 channels would be supported, because the 1966 requires TDM4 for 192kHz, and there are only 4 output ports from the ADAU1452 with 4 channels each. Therefore, maybe we would need two 1452 EVB's, each driving 16 channels. Workable.
It also seems that I would have to tap the SDATA lines directly, since there doesn't seem to be a connector on the board. Rather irritating but I can live with it. I'm a little worried about signal integrity - anyone have experience with doing this?
Or third party, https://www.ebay.com/itm/202367496718 for a 1452 breakout (has all the data lines to a connector for $50!) going to the 1966 EVB.
Do I understand this correctly? Are there any other recommendations I'm overlooking? Can the 1452 support 16 x 192kHz via THD4 outputs? 16 bit is adequate, 24 bit would be nice.
I could get into the higher level DSP's perhaps but I'm very familiar with the sigmastudio environment, and it will otherwise do the job.
Any subtleties in connecting these evals together? I have done a lot in I2S with these chips but not TDM.
Thanks for any guidance!
Audio Signal Processors
adau1452
Audio Products
Audio D/A Converters
SigmaDSP Audio Processors
adau1966
adau1466
dac
Show More
Hello joe42,
Yes the ADAU1452 can do 16 channels of TDM-4 using four ports. When using wires to jump from one board to another signal integrity will be a issue. We have difficulty in the lab using the short Audio Precision cables from the eval board. You can increase the drive strength which helps but it still is tricky. You will have to run the DSP as a master to the DAC.
The EVAL-ADAU1452RevB board has the ability to be modified for control over the codec. Then you can set it for 192kHz fs. I have done it and the instructions for doing the modification is in the user guide. The older boards were VERY difficult to modify.
When doing this I would start at 48kHz to get it all up and working then start raising the rate. 96kHz and then 192kHz.
This way you can verify your hardware and software before trying to run at top speed.
Dave T
Thanks. I'm disappointed this isn't easier. If I follow the modification instructions, it looks like I can still use TDM-4 instead of I2S to control the DAC, and get my 8 channels @ 192kHz, is that true?
Looks like my options are:
1. 4xADAU1452 evals, hacked to 192kHz DAC.
2. 2x (ADAU1452+ADAU1966EVB) combinations, with hand-soldered TDM lines, and questionable signal integrity.
3. 2x Third party ADAU1452 board and a custom PCB with either DAC's or an interface to the ADAU1966EVB.
Too bad there isn't a proper port for the data lines. I'm leading to #3 but I'm not sure yet.
Hello Joe,
Option #1 I do not think is a good idea. With four DSP boards and two DAC boards the jumpers will have to be too long and so I see that being a problem.
Option #2 is better, you could use the four serial output ports with each setup for TDM4 and get 16 channels each. Still a lot of jumpers but it is more manageable.
Option #3 is the best of the choices you described.
My first thought here is that if you really need to run at 192kHz for your proof of concept, and I think you do need to, then I feel a custom PCB with two DSPs and one ADAU1966 DAC on it would be the best. Eval boards are designed to be flexible and have many options and headers etc. This all creates tradeoffs with the layout. To get a good transmission line with few reflections you would not be adding headers and you would keep the lines short. Along with other details.
However, you still may be able to prove your concept as long as some bit errors ( clicks and pops) are acceptable for your prototype. I have been trying to think of other creative options but I do not have any.
Dave T
Thanks Dave! Maybe designing and hand-building a complete board is worth the work. It's a bit of a pain but the extra time spent doing it "right" might make up for the extra work getting the OTS boards to play nicely together. And I get exactly what I want. I can probably start with some working designs anyway.
(I think my fantasy was that I could get this done in a couple of hours with off-the-shelf parts)
I think a 0.1" header with interspersed ground lines (every other wire) would be quite reliable for high speed connections over (reasonable length) ribbon cable, certainly for MCLK and I2S. Seems the third party boards don't do this either, which is unfortunate, and guides me back to a custom board.
Hello Joe,
I strongly agree about the 0.1" header with grounds on every other wire is the best design. We try to do this with all of our eval boards. We have two pin jumpers to connect to the Audio Precision serial interface with one being ground. So having either the odd or even pins on a header be all grounds makes it easy. Then if you use a ribbon cable for the entire header you end up with ground on every other pin. This keeps a defined impedance and a path for return currents for the signals.
Good luck with your project. Hope it goes smoothly.
Dave T
Please log in to reply.