ASR Open Source Streamer Project

[buz]

That RockPI-S is limited to just 512Mb. Hopefully that'd be enough. If not, the RockPI 4 has up to 4GB.

Edit: RockPi 4 does not feature an 8ch I2S.

If I interpret it correctly, Pine64 Rock64 exposes all the I2S and is available at upto 4GB RAM (https://files.pine64.org/doc/rock64/ROCK64_V3_Pi-2_and_Pi-P5+_Bus.pdf). 512 is awfully low...

 

[phofman]

While 512MB RAM is low for a regular streamer (especially with some graphics output), it may be (and likely is) enough for a low-cost USB-UAC2 gadget -> 8ch I2S output (incl. I2S input for analog inputs) with USB explicit feedback control. Maybe even for 2->8ch DSP with CamillaDSP. Such a small cheap board could easily be incorporated into an enclosure of RPi streamer (when one of the RPi-specific audio distributions required). The longer boot time than XMOS would not matter as the RPi must boot too. Many options available, of course all taking some software and linux readiness.

 

[buz]

Having to maintain two different Linux platforms would be awkward at the very least... In case it has DSP duties you even need some interface to configure it :-(

 

[Skeptischism]

There wouldn't necessarily be much maintenance needed on the gadget. its essentially just acting as a peripheral, so once you lock in the build, its done, it doesnt need to be facing the internet, so no updates needed. it doesn't matter what device you are using, there will need to be an interface to configure it and you would do that headless in any case. I hadnt really thought of it like that, as a gadget that emulates a multichannel USB wth a further SBC in the same case, you could even have it in the same stack as the RPI. It doesn't even have to use the USB bus either @phofman it can be using ethernet as a 'USB gadget'?

 

[notabenem]

Rock64 have only 4.4 kernels (stable). Are there reasons to push for boards with newer kernels?

 

[phofman]

Having to maintain two different Linux platforms would be awkward at the very least...

If the I2S distribution is read-only (really no reason for writing on that device, even if running DSP, perhaps apart of the moment of storing DSP config), there is most likely no need for maintenance.

In case it has DSP duties you even need some interface to configure it :-(

Right, that's what the network websocket server in CamillaDSP is for. The python web server for GUI configuration in browser camilladspgui can run on the frontend machine, if needed, or on the backend, if enough memory etc.

The linux backend is just a very flexible option, should a multichannel properly programmed USB-audio controller be unavailable or not enough flexible for the particular situation.

 

[phofman]

it can be using ethernet as a 'USB gadget'?

That's a very good question. The USB-IP driver supports talking to the USB-gadget, I have not tested the combination, do not know the current status. The USB-IP layer for a regular USB device worked very well in my test (e.g. cheap galvanic isolation of USB highspeed connection). The USB-IP client is available even for windows, but that's a different story...

 

[phofman]

Rock64 have only 4.4 kernels (stable). Are there reasons to push for boards with newer kernels?

IMO having support in mainline kernel is a crucial criterium when selecting the technology. E.g. the USB-audio driver has received many important patches recently for UAC2 and implicit feedback, I wonder if anyone has ported them to the ancient android 4.4 kernel.

 

[notabenem]

there are some mainline kernels, but can't comment on their usability: https://github.com/ayufan-rock64/linux-mainline-kernel/releases

 

[buz]

Right, that's what the network websocket server in CamillaDSP is for. The python web server for GUI configuration in browser camilladspgui can run on the frontend machine, if needed, or on the backend, if enough memory etc.

The linux backend is just a very flexible option, should a multichannel properly programmed USB-audio controller be unavailable or not enough flexible for the particular situation.

Agreed, if a decent DAC can be married to it for a reasonable price, it may even have a place as USB DAC with 'hardware' eq, there is curious dearth of those. I also wonder if radxa could spin a version with more RAM, then it could presumably act as all in one streamer...

 

[phofman]

The Rock64 board by Pine64 https://wiki.pine64.org/wiki/ROCK64 has all the I2S pins required, DDR3 up to 4GB, quad cores, performance most likely enough for any streamer. Priced from 40USD (1GB) to 60USD (4GB) on aliexpress. Support for the latest kernel e.g. http://dl.fail.pp.ua/slackware/images/rock64/ .

It would all need to be tried, tested, wrinkles ironed out by reporting to developers and testing fixes. A procedure as usual, when pushing the envelope further. Except that RK3xxx SoCs are used for embedded linux in commercial projects (unlike the individually unobtainable and poorly documented Broadcom settop box SoCs), so probably no envelope to claim pushing Just my 2 cents...

 

[voodooless]

Looks like VolumeIO works on the Pine as well.

 

[Skeptischism]

yep. thats as far as I got with the pineA64 I had here last year. my mate threw it at me thinking I might like it. I looked on the header for i2s and saw the voliumo chit chat, with a thread that kinda just dropped off with not much activity. I only saw the 2 i2s pins exposed ... somewhere ... and put it aside, but if what I read last night is right, the i2s in on a different header, not on the rpi2 legacy GPIO connector, its on the '2x17 pins "Euler" GPIO Header', but it does not have all the extra SDIO1-3 pins exposed for multichannel like the 2 x 11pin on the Rock64.

the LT in LTS is coming up in 2023, but that alone at least is comforting: ie. that they have such products where they commit to supply and support for 5 years.

 

[Skeptischism]

So i'm gonna grab one and have a play I think

 

[Skeptischism]

I'll grab that A64 out and see if I can get volumio going with my miniDSP, for kicks.

 

[voodooless]

I only saw the 2 i2s pins exposed ...

According to this: https://files.pine64.org/doc/rock64/ROCK64_V3_Pi-2_and_Pi-P5+_Bus.pdf they are all on both headers. The smaller header also has SPDIF.

 

[Skeptischism]

yes, i'm talking about on the PINE A64 I have here, with having only the 2 channels exposed on the small header there. the links that would be explicit about the 2 x 20 pi2 header on the A64 are broken, but the second 'Euler' header, which is different again to the Pi-P5 connector on the ROCK64. it also has spdif on it in conjunction with the 2 channel i2s. it would be great if all were available on the pi2 header.

 

[Skeptischism]

thats a no ...

 

[voodooless]

The alwinner SOC seems to have two I2S ports, with only one data channel (well, one in, one out). It does claim to support TDM, but it’s unclear upto how many channels. The rock64 has a 4 channel I2S controller (and another 1 channel one).

Edit: It does seem to support 8 channel TDM over that single I2S interface. Clock upto 100 MHz. I think that gives quite some headroom for high sample rates.

 

[Skeptischism]

Yes, rock64 in my cart. TDM is great if youve not got any other options and/or want very high channel counts, but not widely supported out of the box in many dac chips, so some reciever hardware or glue logic would be needed to use with anything i'm currently working on.

having known the ROCK chips from DAPs, I was excited when he threw that to me; only to find it was a neutered model. I thought for sure it would have better audio capabilities than Pi. i'm going to have a play and maybe give it back to my mate with a little dac attached for his birthday.