ASR Open Source Streamer Project

[notabenem]

What have we been doing the last 13 pages then?

Good question. Does anyone know?
To me it seemed like physical design. Everybody was suggesting various, but very specific models of 8 channel DACs without it being clear whether:
1. we want to bundle a DAC with the streamer, or not.
2. we want to do also crossover duties outside the speaker cabinets
3. we want to have so many channels and thus making the product (often) unnecessarily expensive.

Amir himself mentioned 'multiple external dacs'. But regardless of his suggestion, i am not aware of any decision being taken on a conceptual level as for the roles, placement and integration points. If you recall my mindmap and open questions related to it, I think those are still open.

 

[Skeptischism]

Yup. I ask, because if you do bundle it with the DAC, you could focus funds without duplication. if you can do i2s input, that removes the need for an XMOS USB input. if you are happy with camilla DSP (and you should be) Henrik has designed a really powerful program and its free!! He is also quite responsive and its newish, so you may even be able to get features added, if he saw merit i them. that would remove the need for it to have onboard DSP processing and software and also, (and this is important IMO) it allows the exact delay induced by the particular filter you are using to be known by the host of the processing and video playback, so it would be fairly trivial to share that information with your video player, allowing it to delay video accordingly, even if you are running heavy FIR correction.

People are suggesting 8 channels, because really DAC chips mostly come in 2 or 8 channel variants. Its a pity, if it wasn't for the AKM fire, I would have a board that would be perfect for this, ready to go, based on AK4468 and opa1632. i'm aware the 1632 is EOL, but there's still plenty of stock around and anyway, there are plenty other suitable FDAs if it came to that.

Apart from more commodity chips, or ones built into Ti class D amp chips, ESS pretty much has the market cornered. crystal semi has a few, but ive never really been a fan of their parts TBH. I would suggest you may not need 8 channels for 2 way speakers ... but you would likely want to reserve the ability to add 2 subs, so 6 channels would be a minimum. For example, if you used an 8 channel chip, and possibly a lower performing one, like the 9080, or even the 'old' 9018S, you could parallel 2 channels for each midbass driver, increasing the performance for those channels in the most audible range.

As it is, i'm in the middle of designing a PCB for an 8 channel design, based on 2 x ES9038PRO, but it won't be ready for some time and, even if I end up going to market with it (its just high end DIY at this point) it won't be priced suitably for this project.

 

[notabenem]

... that removes the need for an XMOS USB input.

if we ever end up with a quality N channel DAC in this, i'd hate to see it not having a USB audio class feature so that you could simply connect it to your PC/laptop/whatever. I know, cherry on top.

 

[voodooless]

Did anyone ever hook up 8-channel audio to a Pi without USB, and with higher sample rates?

If you can get a 110 SINAD out of a 9080, is it really worth the additional effort and cost of implementing a higher-end model?

 

[Skeptischism]

No, it definitely does not work with 8 channel over the GPIO header. its 4 channel, possibly 6. its also bandwidth limited to about 15mb/s iirc.

if we ever end up with a quality N channel DAC in this, i'd hate to see it not having a USB audio class feature so that you could simply connect it to your PC/laptop/whatever. I know, cherry on top.

You want to be able to connect your computer to another computer?

Well, you see, that part in particular takes it from the realms of doable as a separate project, to the realms of, 'why dont I just go ahead and case it and sell it for 3-4x the money . lots of expense and time in designing an isolated, modern USB input and honestly that part isnt in my skillset, i'm not a software guy, thats my partner in the project. I can design that hardware, but that part is at least as much about software development.

if that were the case and you wanted to use it with something else, you could
A. buy a minidsp USB->i2s xmos card
B. use the raspberry pi in gadget mode. this allows it to appear as a usb peripheral. yes, this works and its cool AF

That being said, we might be able to work out the xmos, but it does add cost and complexity. if it can be avoided, you should avoid it IMO

 

[Skeptischism]

i'm not even certain about 4 channel. I need to look into it. that's how I remember it, but I need to double check that. XMOS is in my near future for my stuff, but i'm trying to cut a few corners for this project to potentially keep costs and development time down.

 

[voodooless]

i'm not even certain about 4 channel. I need to look into it. that's how I remember it, but I need to double check that. XMOS is in my near future for my stuff, but i'm trying to cut a few corners for this project to potentially keep costs and development time down.

So if you don’t wanna do USB, and more than 4 channels is a challenge, how can it be done then? I’m guessing the pci-e bus is not available on the header? Otherwise that would have been an easy target. Possibly misuse some other high speed bus on the thing? PDI interface? MIPI display interface? They would need some creative driver programming, as well as some programmable logic to convert to I2S. Adds a lot of complexity.

 

[Skeptischism]

Did anyone ever hook up 8-channel audio to a Pi without USB, and with higher sample rates?

If you can get a 110 SINAD out of a 9080, is it really worth the additional effort and cost of implementing a higher-end model?

Yeah, with a single number -108dB THD+N (probably just at 1khz) i'm not sure that tells us a great deal and there is no datasheet available anywhere for me to look at to see what other issues there are as far as interfacing it. if you look at their catalogue, the chip is only a few bucks, in a range where the top of the line is nearly $100 and it includes 8 x line drivers, so i'm a bit wary to suggest using it as part of a system of this quality. its possible it doesnt include internal volume control either, which would be a prerequisite IMO.

 

[Skeptischism]

Nah, those are the 2 realistic choices, GPIO (i2s) or USB. I just hadnt thought it through and cant remember definitely whether it will handle enough bandwidth over GPIO. if it came to needing USB, I was just going to suggest minidsp USB->i2s and that comes with the bonus of allowing people to optionally use their software instead of camilla (camilla is more powerful IMO) thats the cheapest, multichannel xmos I know of.

 

[voodooless]

its possible it doesnt include internal volume control either, which would be a prerequisite IMO.

It does appear to have volume control. But that is about all in info that is available without a proper datasheet.

 

[voodooless]

if it came to needing USB, I was just going to suggest minidsp USB->i2s

It isn’t very cheap though. Edit: $115. That is sort of doable I guess.

But all those separate components don’t really make a nice solution. Preferably you would like one PCB that houses the PI via a header, and all other stuff, like power supplies (just DC input), DAC, output buffers and connectors.

 

[Skeptischism]

No, it isnt. the XMOS XU216 series chips alone arent cheap and there is a fair bit of work involved.

See the multichannel i2s on Rpi has always been a bit vaguely described. its doable, but seems a bit of black magic is involved. I do know there is a bandwidth limitation that would have to be respected, so I couldn't see you going over 96khz for 4 channels.

a quick google results in this product, which is cheap .... may, or may not be aubibly transparent as it is, but it does suggest that its possible. https://www.amazon.com/Octo-channel-sound-card-Raspberry/dp/B06Y3PZ6MF

 

[Skeptischism]

but for example, in the i2s dac driver list in MoOde, there isnt a singe driver for a device that has more than 2 channels.

There may also be someone else on the forum that has a working xmos (or equivalent) program that they would allow us to use for the project, or reach a license agreement, given a run of production etc.. and I could include an xmos on the board and flash it with their firmware.

not really many XMOS equivalents; yet another virtual monopoly .

 

[notabenem]

You want to be able to connect your computer to another computer?

What is not a computer these days? Anyway, this is exactly the reason why I, in general, prefer the modular approach (regardless, if it's inside a single case or not, if the internal coupling is loose). If I need a good multichannel DAC with my PC (occasionally), i can use the one I bought and normally use with my networked streamer.

Also i believe modular solutions tend to do better with (or against) obsolescense. What are the chances that RPI will get obsolete in 5 years? Hefty. What are the chances that the DAC will still be doing well below the audability threshold? If those two are not tightly coupled together, I can simply change the processor and keep the DAC. Less waste this way and better long-term usability.

How many have you paid for a smart TV (Panasonic, anyone?) just to realized that first the video-conferencing apps stopped working (no use for the integrated camera) then most of the smart functions were gone (transition from TLS 1.1 to TLS 1.2) and then the dual-tuner got also useless (DVB-T1 -> DVB-T2, MPEG2-> H.265) to render the nicely integrated package to nothing more than a very expensive plasma 'monitor'?

I believe you if you say adding a USB port has its own challenges. One more reason to use an external DAC, imho.

 

[voodooless]

https://www.amazon.com/Octo-channel-sound-card-Raspberry/dp/B06Y3PZ6MF

That was a Kickstarter project a while ago. I think it only does 48kHz.

Ess also has some USB chips that can do multichannel I2S. The 9080 eval board also seems to have one. ES9280a could do it.

 

[Skeptischism]

Its not so much that it presents a huge challenge, is just not something I personally have ready to go on a board (although there may be someone here that has something usable; perhaps i'll shoot Amir a PM). well actually even that is fairly straightforward, it's more a matter of the firmware and thats not on the list for me/us until after ive prototyped the dac boards. My 2 channel test PCB goes off in the next couple weeks, then onto the multichannel 9038PRO board and clock distribution (for the ADC too) then the XMOS, so its not likely to be sooner than 4-6 months.

You'll have to pay for it somewhere if you want an external unit. btw, i'm aware of the dearth of multichannel dacs around of quality, so I wouldnt just say you'll get one, so casually . generally you have recording interfaces, or DEQX, OKTO Dac, Analogue precision UP2 (not out yet) and none of these are going to fall anywhere near your budget by the sounds of it, apart from the recording interfaces.

@voodooless yeah, I spotted that too and i'm saying thats the reason that datasheet for the 9080 part isn't available. Screw using ESS USB chips as well, its bad enough having one part of theres in what is preferred to be an open project.

if we can guarantee enough sales of the parts, its possible they would sell us some of the https://www.esstech.com/wp-content/uploads/2021/04/ES9281A_80A_PRO-Product-Brief-v0.2.4.pdf (perhaps go through ismosys rather than Shaw) and team it with as many channels as required. the internal ones, plus 1 x ES9038q2m would work. it has their new USB PHY embedded. but that odd 90 pin package looks like a pain. but if it was decided to use these, or any of the ones other than what is available at mouser, I and anyone involved would be bound by an NDA, so we couldn't discuss the project, the firmware or schematic could never be published, etc etc.

 

[notabenem]

You'll have to pay for it somewhere if you want an external unit. btw, i'm aware of the dearth of multichannel dacs around of quality, so I wouldnt just say you'll get one, so casually . generally you have recording interfaces, or DEQX, OKTO Dac, Analogue precision UP2 (not out yet) and none of these are going to fall anywhere near your budget by the sounds of it, apart from the recording interfaces.

That's also a point. Some of us don't need 8 channels, or 8 channels with the highest quality. We could start with 2 channels and swap later, when 8 channels of reasonable quality becomes more affordable. Each to their own, i'd say.

 

[Skeptischism]

I dont think anyone is suggesting an 8 channel dac (or any part of the project) is a prerequisite for building this project; so i'm not sure i'm following your point?

 

[voodooless]

That's also a point. Some of us don't need 8 channels, or 8 channels with the highest quality. We could start with 2 channels and swap later, when 8 channels of reasonable quality becomes more affordable. Each to their own, i'd say.

Well, a backplane board and places to stick in addition channels would be cool. Also probably complex. But would also give flexibility in boards. You could have cheap versions, high end versions, XLR, RCA versions, etc. One could also add input boards, like spdif or analog. Ultimate flexibility

While it is fun talking about all kinds of options, we actually still don’t know what this project actually is about..

 

[Skeptischism]

Lol. the additional cost of designing and making it modular like that, may well get you most of the way to having multiple channels to begin with. clock distribution, housekeeping circuits to manage the scalable power and inputs/outputs, connectors that maintain proper impedance for the i2s, a scalable low noise power supply etc etc. sounds like a nightmare to cater for people that dont have money to spend and may not ever spend the money i'm joking, but yeah you may end up making everyone else spend more money, to cater for people that aren't them, that may never take up the option. not something I would ever want to get involved with. I already have a fair bit on my plate. I just saw this the other day and thought, hmmm, i'm designing a multichannel dac, how can I offer something that

A. won't just be a massive time suck.
B. doesnt compete with my own potential product,
C. still offers good value to the project and fills the need.

basically I cant stand this tendency to turn everything into legoland to cater for every possible option. it makes the quality of every function worse. you cant possibly have clean, high performance design while doing that.