Thank you for the response. If I understand correctly, WM8005 is TDM compatible and it "just" becomes a matter of finding the right implemention which converts I2S to SPDIF (in some delivery format) while allowing the user to chose which pair of TDM (8) channels should be output. Since @mdsimon2 has found a card with AES/EBU, I am curious of those or any other such alternatives (bear in mind that I very well may have it all wrong )
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Multichannel audio on a Pi will get a whole lot easier and cheaper!
- Thread starter voodooless
- Start date
Unfortunately, wm8805 is not TDM capable. You can chose between 8 stereo spdif inputs and one stereo i2s input.
But you could achieve TDM to spdif with other parts, for instance daisy chaining two es9080q
Last edited:
It does say "8 channels of I2S data output are routed from the RPi5 to 4 WM8805 based I2S to AES output boards. ", so at least somehow the choice btw the different channels is made by those boards, since they are all fed the same 8-channel signal, right?
The rpi5 outputs 8 channel i2s using 4 different 2 channel data lines, each of those goes to a different wm8805 board.
Ah I feel like Chat GPT lied to me a bit there again. An Rpi 5 indeed does have 4x2 I2S signals from pins and the only thing missing in order to directly get to them in spdif format is the correct and synchronized clock, same for each channel pair, as solved by the "black little board"! Also, that means that most any I2S -> SPDIF (whether coax 75 Ohm, Toslink or AES/EBY 110 Ohm) should be fine?
mdsimon2
Major Contributor
These are the specific boards I used -> https://www.yoycart.com/Product/540236665171/SupportxMOS_Amanero_USB_interface_to_RCA. They aren't anything special and there are many similar WM8805 boards available. They are I2S to AES and have nothing to do with TDM as the RPi5 outputs 4 x 2 channel I2S as mentioned by @MCH.
The issue with NOT using a SPDIF to I2S card on the input is that the WM8805 requires MCLK and the RPi5 does not have an MCLK output. It should be possible to output MCLK on a GPIO pin but to my knowledge no one has done it and the RPi developers have not provided any guidance on how to accomplish it.
Michael
The issue with NOT using a SPDIF to I2S card on the input is that the WM8805 requires MCLK and the RPi5 does not have an MCLK output. It should be possible to output MCLK on a GPIO pin but to my knowledge no one has done it and the RPi developers have not provided any guidance on how to accomplish it.
Michael
Thanks a lot! The lack of MCLK output on the RPi5, what consequences does that have for playback from "within" the Pi (i.e. Moode); one can't reasonably use the MCLK of the SPDIF-I2S board then? And the black buffer board only forwards MCLK, no master clock of its own that could be used?
Edit: just saw that one one of the purchasing options for the buffer board is to add an oscillator circuit, perhaps that would enable it to re-clock and be the master clock? If so, that seems to be what I need, in case I don't go for the spdif-Ì2S board too. (Would having that input card enable full DRC on realtime signals from old school CDs, Chromecast etc.?)
mdsimon2
Major Contributor
MCLK, BCLK, LRCLK and I2S data need to be sync'd.
When using a SPDIF to I2S board as input, the RPi is a clock consumer (slave) and everything is sync'd to the incoming SPDIF stream.
If you use the RPi as a clock producer (master), everything is sync'd to an oscillator internal to the RPi. Adding an oscillator to the Kaamos buffer board doesn't really help because it won't be sync'd to BLCK, LRCLK and I2S data.
There are products like the IanCanada McFIFO / McDualXO which can take BLCK, LRCLK and I2S data, and re-clock them using a separate oscillator, but by the nature of the FIFO a lot of delay (100+ ms) is added. The McFIFO / McDualXO also acts as a clock buffer so the Kaamos buffer board would be redundant.
I haven't tried it, but you may be able to output from Moode to a USB to I2S board, and then route BLCK, LRCLK and I2S data from that board to the RPi5 (acting as a clock consumer) and then take 8 channels I2S output from the RPi5.
Michael
IIUC it may only take an appropriate overlay.Interesting, looks trustworthy and reasonably priced: https://www.tindie.com/products/eclipsevl/multichannel-usb-to-i2s-uac2-interface-york/
Disclaimer: I am a newbie diyer so don't take my answer as an authoritative one, I hope others chime in.
I have seen plenty of boards that split the clocks without buffers and work fine, I think you could give it a try and if it works, just forget about the buffer board. Will probably depend on how tidy you manage to do it and what sample rate/MCLK frequency you are aiming at.
By coincidence today I was reading this (that is an interesting read anyways) where an Analog Devices engineer explains the advantages and disadvantages of splitting clock traces and buffering. When explaining one of his examples, he mentions that when splitting 8 times one might run into trouble but still doable, but I am pretty sure he is assuming a property done PCB, not a few jumpers laying around. I would think 4x can be fine if done properly. You don't lose anything by trying.
Forgot to mention: the guy that makes the york interface you linked has an active thread in diyaudio.com You could ask folks there how do they use it and what are potential issues. He also has an account here, but hasn't been active since last year.
Last edited:
Another idea which may have fatal caveats: what about using the Rpi5 pins to connect all 4 stereo data signals for I2S to a proper 8-channel DAC-chip, e.g. diyinhk's ESS9028PRO
www.diyinhk.com
The DAC card would act as master clock and the Rpi5 hence slave. Output in 8 channels would be perfectly time-synced. No need for XMOS USB->I2S-card and also no need for reclocker on the DAC card (since it is master anyhow).
Aim would be to surpass the performance of Hifiberry's DAC8X at similar cost. I "need" 8 channel output from my Rpi5 running Moode/CamillaDSP for an imagined 4-way active drive system
Fairly happy to spend some money on high value items like amps and tweeters etc. but not on a fancy studio USB interface that I otherwise don't need.
8 Channels 384kHz 32bit ES9028PRO PCM DXD DSD Audio DAC
4 layer PCB, two solid middle ground plane NDK NZ2520SD 80Mhz low phase noise oscillator USB Type-C socket, Gold plated RCA socket Three ultralow noise 1.2V linear regulator Free 0.96 OLED display until stock last
www.diyinhk.com
The DAC card would act as master clock and the Rpi5 hence slave. Output in 8 channels would be perfectly time-synced. No need for XMOS USB->I2S-card and also no need for reclocker on the DAC card (since it is master anyhow).
Aim would be to surpass the performance of Hifiberry's DAC8X at similar cost. I "need" 8 channel output from my Rpi5 running Moode/CamillaDSP for an imagined 4-way active drive system
Fairly happy to spend some money on high value items like amps and tweeters etc. but not on a fancy studio USB interface that I otherwise don't need.mdsimon2
Major Contributor
Another idea which may have fatal caveats: what about using the Rpi5 pins to connect all 4 stereo data signals for I2S to a proper 8-channel DAC-chip, e.g. diyinhk's ESS9028PRO
8 Channels 384kHz 32bit ES9028PRO PCM DXD DSD Audio DAC
4 layer PCB, two solid middle ground plane NDK NZ2520SD 80Mhz low phase noise oscillator USB Type-C socket, Gold plated RCA socket Three ultralow noise 1.2V linear regulator Free 0.96 OLED display until stock last
The DAC card would act as master clock and the Rpi5 hence slave. Output in 8 channels would be perfectly time-synced. No need for XMOS USB->I2S-card and also no need for reclocker on the DAC card (since it is master anyhow).
Aim would be to surpass the performance of Hifiberry's DAC8X at similar cost. I "need" 8 channel output from my Rpi5 running Moode/CamillaDSP for an imagined 4-way active drive system
In my experience those DIYINK boards are not very good, however I have only personally tried the ES9016 and 9038pro boards. Costs end up being quite high when you factor in required power supplies and stock I/V outputs limit performance. You can use external I/V but again that adds cost and requires more space.
One advantage of the DIYINHK board is that it has on board oscillator and operates in async mode by default, therefore you do not need to provide a MCLK. The RPi would still act as master (clock producer) in this case. You would not need a buffer as you are not splitting signals, just route BLCK, LRCLK and 4x I2S data from the RPi to the DAC.
Michael
mdsimon2
Major Contributor
Tough to beat the dac8x on price, but I don't love the form factor (headphone outputs are rather fragile) and personally prefer 8 channels of balanced output.
Unless you design your own I2S input DAC, I don't think you will be able to beat performance to price ratio of a commercial USB interface.
Michael
Unless you design your own I2S input DAC, I don't think you will be able to beat performance to price ratio of a commercial USB interface.
Michael
I see t
I see, thanks. I guess I would pay maybe €/$ 200 for 8 channels with a bit better performance than DAC8X but that seems very naive...
They have a balanced version now. Studio dac8x
