• 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!

New theory for cost-aware 4-way XO on Rpi5: get the cheapo Behringer UMC1820 and use its analogue outputs for sub bass and bass, but use external stereo DACS (present) for mids and highs, connected to the digital outputs of the Behringer (S/PDIF and ADAT). The Behringer should keep all channels in sync.

(Same principle could of course apply for other cheap but well equipped USB interfaces, perhaps there are better choices than the Behringer.)
 
I'm currently using a MiniDSP MCHStreamer with Kaamos buffer to output to two Ian Canada HDMIpi boards. An Ian Canada ShieldPi Pro III has 5v input from an IFi power supply (need to upgrade). This in turn provides filtered power to the Kaamos buffer (3.3v) and the 2x HDMIpi boards (via GPIO). Short UFL cabling from the MCHStreamer / Kaamos buffer to each of the HDMIpi boards which then send their respective outputs on to a couple of i2s DACs (one Sabaj D5 and the other an SMSL DO200).

Works great with 4 channels at 32bit / 384khz which is output from an rpi5 running Ubuntu and CamillaDSP.

-Jim

:
IMG_5599.jpeg
 

Attachments

  • IMG_5598.jpeg
    IMG_5598.jpeg
    63.7 KB · Views: 151
Last edited:
Hi thank you for the extremely interestin thread and sorry to jump in but i am very interested in volume control Ics
Speaking of the PGA2310 it seems to be good enough to be used in quite high end units like this one

" The volume control is implemented with two two-channel Burr-Brown PGA2310 programmable-gain chips, one per channel used as a differential volume control "
the result is quite spectacular

With the CP-800 in Analog Bypass mode and the volume control set to unity, the distortion harmonics are not much higher than the residual levels in the Audio Precision's output :oops:

912CP800fig15.jpg
 
Hello, smal question / Sugestion i have?
I read that an RME Digiface Dante is now runing with ALSA USB 2 or 3 Clascompilant on Linux with new Linux Kernel stable.

So now i thinking about using it for CamilaDSP on RPI5 with an DANTE to 16 or 32 Chanel DAC (Ferofisch A32pro Dante (3500€)
or like RME M-32 DA Pro II-D (3800€)) 16 Chanels OUT minimum required.

Also there is that little nice Device that makes HDMI to DANTE Audio Conversion, 48 and 96kHz Dante kompatible, Blustream SW42DA (ca. 1,260.00€ excl. VAT).

https://www.futureshop.co.uk/EUR/ne...SZLvFHdlrS_3KEME60Bo_NiX43VWiNljjiq4XG0K0S1uE

So both devices to one RME Digiface Dante (1200€) and you has 7,1 Digital in and 16 or 32 Analogue TRS Outs over Dante to USB 2 or 3 into an RPI 5 to use CamilaDSP!

OK OK i know thats the most expensive solution but i gues also the most High End solution.

Routing you need to do with an Windows PC and than it will save it on RME Digiface Dante once running in STandalone or CC Mode (Clascompilant)

MINIDSP TIDE 16 is not an option for me only 48kHz internal and no FIR filter :( thats sucks.

More Cheeper option (without HDMI in option) is Using RME Digiface USB (420€) with 4x4 ADAT outs with 96kHz into an Ferofish PULSE16 AE AD/DA Converter (1200€).

So than 16 TRS Outs aviable with one Volumeknob with Monitor Controll. for 1620€ Price is neare OCTO 8 DAC but with 16 Chanels

Or waiting til MOTU release the new MOTU 24AO (Stil no Info about) to have 24 Analogue Chanels on USB 2 the old 24AO you can only buy used if you has luck and the DAC's are not that great like on MOTU 16A (2025) or Ultralite 2.

Not sure how good the new MOTU 16A will work on Linux with USB, also not cheep and also no 4x4 ADAT in possible if USB don't work well on it.

But you need still to use this HDMI to I2S Board on RPI 5, Not Sure if there still an Chanelswap problem with it, use it still Stereo only :(

Onlly loud thinking here lol

Need to find an new Apartment for me and my CNC right now, so no time for DIY at the Moment LOL.

Still use my option with 2 MOTU Ultralights conected with ADAT and SPDIF together for 16 Chanels Out and it works so far.
But thats also more that 1400€ for 2 MOTU UL 2 ......right!
 
Last edited:
Noob here just food for thought

Going higher than 48 is IMO a waste of money, but if using a PC as convolver NP.

If using RPi instead for CamillaDSP go "Modularised DSP ™ " for the speaker building functions with more than one, e.g. one per LR side?
Raspiaudio 8xIN+8xOUT, or HiFiBerry DAC8x + ADC8x

Could still use a stereo DSP for the ad-hoc user controls and DRC layer.

Avoid HDMI AMAP for audio, stick to ADAT and SPDIF

What benefits do Dante bring you?
 
Last edited:
Noob here just foid for thought

Going higher than 48 is IMO a waste of money, but if using a PC as convolver NP.

If using RPi instead for CamillaDSP go "Modularised DSP ™ " for the speaker building functions with more than one, e.g. one per LR side?
Raspiaudio 8xIN+8xOUT, or HiFiBerry DAC8x + ADC8x

Could still use a stereo DSP for the ad-hoc user controls and DRC layer.

Avoid HDMI AMAP for audio, stick to ADAT and SPDIF

What benefits do Dante bring you?
With onlly 8 Analogue outs you can't handle 7 2 Way Speaker and 2 Subs full aktiv Surround, at least 16 Analogue Chanels you need for 7.2 Suround full aktiv

I use right now 2 MOTU Ultralite on an RPI with Camila DSP and steep FIR Filter for the Main Speaker for that but i have 2 Volumenobs on 2 different devices to handle Master volume of all 16 Analogue Chanels thats not that bad but not strait vorward also you can use only LPCM on the HDMI to I2S Board into the RPI 5
No Dolby or Atmos at all.

With that Bluestream SW42DA you has all codecs and sampleconversion to feed into DANTE (8 CHanels or more) also HDCP compatible with 96kHz and with RME Digface DANTE also in to RPI 5 and out to an DANTE 16 or 32 Chanel DAC with Master Volume in sabre DACS strait before the AMPS. So any Blueray with any Suround format would play to DANTE.

More and More AV Processors will become DANTE Compatible right now but they absolutly Expensive like 4000€ full Digital with Dante
https://www.hyperionprofessional.com/products/dpr-16-digital-immersive-processor

JBL Synthesis – SDP-55 6000€

You can try to buy an Trinov for 18000 $ ;)
 
Last edited:
Using interfacii is sure expensive when using so many channels - Do you have a PC doing convolving?

With onlly 8 Analogue outs you can't handle 7 2 Way Speaker and 2 Subs full aktiv Surround, at least 16 Analogue Chanels you need for 7.2 Suround full aktiv
As I said, multiple RPi units.

If you only need 16ch for the "speaker building" functions, then 2x might be enough, call one Left and the other Right

Maybe one more DSP unit upstream either another RPi, or a 2x4 miniDSP just for stereo level, can act as preamp, do DRC.

You did not answer my Q about Dante.
 
Using interfacii is sure expensive when using so many channels - Do you have a PC doing convolving?

As I said, multiple RPi units.

If you only need 16ch for the "speaker building" functions, then 2x might be enough, call one Left and the other Right

Maybe one more DSP unit upstream either another RPi, or a 2x4 miniDSP just for stereo level, can act as preamp, do DRC.

You did not answer my Q about Dante.
You don't need multiple Camila DSP processors i tried it already with one RPI 5 and Camila DSP on 16 Analogue OUTS 96kHz Fir Filter on al Outputs and need onlly 20% Power off that one RPI 5!!! With HDMI in to I2S 8 Chanels At this time ALSA had the Bandwith only for 24 Chanels OUT at once but with an new Linux Kernel they solved it i read and RME DIGIFACE on USB 3 is Capabel to deliver 32 Chanels in and Out AT 96 kHz if i read right max 128 x 128 at 48 kHz DANTE and MADI together i gues.
But i need ony 8 in and 16 Out ;)

I have 2 Bluerayplayer, 2 because only the older Cambridge CXU HD is capabel to run my SACD's the 4K Panasonic didn't do that. And HDMI is the only possability to feed 8 Chanels Digitaly into an DSP Software

Next problem you need to sync more than one RPI or you get an Phaseshift between the Devices on Outputs, Already using 2 USB Dacs at the same time on RPI with Camila was not working without Phaseshifts !!!
 
Last edited:
I think I still need multiple RPi not for CPU power but to save money on interfacii, to go past 8 ports cheaply.

Yes RPi5 is overkill I guess can save more, thinking RPi4b is the sweet spot.

Please answer my Dante question.

You don't need multiple Camila DSP processors i tried it already with one RPI 5 and Camila DSP on 16 Analogue OUTS 96kHz Fir Filter on al Outputs and need onlly 20% Power off that one RPI 5!!! With HDMI in to I2S 8 Chanels At this time ALSA had the Bandwith only for 24 Chanels OUT at once but with an new Linux Kernel the solved it i read and RME DIGIFACE on USB 3 is Capabel to deliver 32 Chanels in and Out AT 96 kHz if i read right max 128 x 128 at 48 kHz DANTE and MADI together i gues.
 
I think I still need multiple RPi not for CPU power but to save money on interfacii, to go past 8 ports cheaply.

Yes RPi5 is overkill I guess can save more, thinking RPi4b is the sweet spot.

Please answer my Dante question.
?? Did already The Bluestreem SW42DA has 8 Analogue OUTS or DANTE if you want to use digital 7.1 decoded signals to feed into an RPI with that RME Digiface DANTE over USB 3

I use an cheep HDMI to i2s Board but need to use the LPCM converted Signal from my Bluerayplayer no way to get Atmos or Dolby Digital nativ in to the RPI right now.

Not sure how much Difference that made tho. It works that way for me.

How to get it Decoded than before CamilaDSP

You can also use that Hyperion fully digital Surroundprocessor feed into RME Digiface put an Camiladsp between the Processor and an DANTE DAC with AMPS on it

And you can use a lot of Chanels in and out not only 7.1 there are up to 16 Chanels possible with Atmos but thats overkill for me i am already Happy with normal 5 or 7.2 , don't have more room for much more Speaker and the Nighbor get obsessed with me already LOL

Thats by far the most expensive but still cheepper way than using Trinov for 18000€

But i think about an idea to made an Chair with a lot of Speakers (Smal may be Widerange) around the Head canceled from Room around, like on ealrlier days and feed in Surroundsound to watch movie or listen Suround music withou geting the nighbor angry lol on the earlier days the didn't have surround sound at all i guess Quadro sound only or so so, they had that nice Ball kind chair but only for Stereo Sound. Only loud thinkink ! LOL
The Chair don't need to be round at all but placing smal Speaker around the Head only 50 cm or an Meter away may be with canceling it from the room with only an front open space for TV Watching could be nice to test like the Big Hair Dryer from our Mothers 1970 LOL
 
Last edited:
Sorry, language barrier here.

I have no interest in proprietary decoding myself, but my choice would be, output the channels to the first RPi AFTER decoding.

My objective is avoiding expensive interfaces, using two or more RPi units at 8 channels I/O each is how. Spend hundreds not thousands.

I am not asking about your current system, that approach is unaffordable for me.

What I want is:

Avoiding HDMI for audio, avoiding I2S.

Stick to normal cheap connections - analog or S/PDIF or USB only, maybe ADAT if needed.

and I guess avoiding Dante, since I do not see (you have not clearly stated) what its advantage is, other than "these crazy pricey interfaces require it".


?? Did already The Bluestreem SW42DA has 8 Analogue OUTS or DANTE if you want to use digital 7.1 decoded signals to feed into an RPI with that RME Digiface DANTE over USB 3

I use an cheep HDMI to i2s Board but need to use the LPCM converted Signal from my Bluerayplayer no way to get Atmos or Dolby Digital nativ in to the RPI right now.

Not sure how much Difference that made tho. It works that way for me.

How to get it Decoded than before CamilaDSP

You can also use that Hyperion fully digital Surroundprocessor feed into RME Digiface put an Camiladsp between the Processor and an DANTE DAC with AMPS on it

And you can use a lot of Chanels in and out not only 7.1 there are up to 16 Chanels possible with Atmos but thats overkill for me i am already Happy with normal 5 or 7.2 , don't have more room for much more Speaker and the Nighbor get obsessed with me already LOL

Thats by far the most expensive but still cheepper way than using Trinov for 18000€
 
How you will get different RPI syncronised in Digital Domain than??? If 8 chanels in one and 8 Chanles in the Other RPI need to work together???

Or You want only to use them Analogue in and out ???? Thats totaly the oposite of what i want to do. Thtas double conversion than.

You us an Digital source converting signal from digital to Analogue than again Analogue to Digital using DSP and after that again Digital to Analogue ???

That is the last thing i would ever do, by sure. More noise more conversion errors in between every DAC and ADC add his own sound into the signal.

Avoiding I2S ...so you are only working Analogue than??? I2S is used on most Audioequipment for Signaltransport between different Chips, DACS, DSPS and so on

No I2S no digital Comunication in between i guess lol

May be i didn't understand what youre target is for me is to have the possability using FIR FIlter in Digital Domain befor first DAC to avoid Phaseshifts, Noise and reaching perfekt Timing betwen Chassis on a Speaker

IIR is neare the same as using Passive Crossover only with the possability Timealigning the Chassis much easier. Stil Phaseshifts possible between Chassis and sound coloration overlab between two Chassis

On Analogue side neare inpossible to do so and also needs lot of Time to develope.

The only problem is processing delay per Chanel if you are using much channels with lots of Taps on DSP Side it costs processing time and using 16 Chanels didn't make it better.

But thats only my opinion every one could do what he want to do if he is still happy with it ;)
 
How you will get different RPI syncronised in Digital Domain than??? If 8 chanels in one and 8 Chanles in the Other RPI need to work together???
Well I'm not sure that will in reality be an issue. So long as any latency is deterministic / predictable there may not actually be any "drift" between them.

Just thinking out loud, if the upstream stereo RPi is doing the DRC as "master in Stereo" of the downstream nodes handling the "speaker building"

Linux does have "network clocking" protocols, I have yet to learn the details

...

> double conversion...

> You us an Digital source converting signal from digital to Analogue than again Analogue to Digital using DSP and after that again Digital to Analogue

TBD but not a concern for now as I plan for each stage to be transparent way past human audibility.

Say at the SQ level of miniDSP Pocket ADC as a benchmark. No "own sound" objectively, just psychoacoustic from unconscious biases.

Even 10x conversions back and forth, not detectable via human ears!

...

> using FIR FIlter in Digital Domain befor first DAC to avoid Phaseshifts, Noise and reaching perfekt Timing betwen Chassis on a Speaker

> IIR ... Timealigning ... Stil Phaseshifts

The DSPing can happen anywhere in the signal chain.

> possible between Chassis and sound coloration overlab between two Chassis

wut?

> processing delay per Chanel if you are using much channels with lots of Taps on DSP Side it costs processing time and using 16 Chanels didn't make it better.

RPi has plenty of processing power for all that. One RPi5 has more FIR convolving power than lots of small / quiet PCs

But RPi4 is plenty too, if each is only handling 8 channels.

...

> Avoiding I2S ...so you are only working Analogue than?

No but so far S/PDIF and USB are fine. If I see a real need to use I2S outside of an assembled component OK, but there aren't even defined standard connectors. And HDMI for audio I avoid like the plague.

Relevant
 
Well I'm not sure that will in reality be an issue. So long as any latency is deterministic / predictable there may not actually be any "drift" between them.
Frunse is right, it is a major issue.
Just thinking out loud, if the upstream stereo RPi is doing the DRC as "master in Stereo" of the downstream nodes handling the "speaker building"
How does that help with the audio output syncing?
Linux does have "network clocking" protocols, I have yet to learn the details
It does, but again - how does it solve the problem?
 
How does that help with the audio output syncing?
As I said I dunno yet, I'm not even 5% up a decent learning curve.

I plan to TEST, and adjust my plans as needed.

Researching PipeWire GStreamer RTP / PTP, ROC, Netjack2 etc will be my kind of fun

AES67?

Worst case, might try doing it in Windoze


Or, maybe I get lucky, and it just works ™

or even luckier, and no DSP is required at all!

 
Every DAC may be also ADC (not sure yet) has an Filter thats add ripple on any way and Destortion if the full Headroom isn't use.

Depends on what kind of filter let say on Sabre Dac is used.

I heare different sound if i try different dacs like i did blind test it already longer Time ago this difference isn't that huge like comparing 2 different Speaker but it is there ;)

I tested Dacs around 400 to 1500€ with an friend and we both could hear the difference, tide Bass or not deteils in the Hight there or not and so on....

So double conversion is always an bad decission for me you add something to the Sound the one or other way ....its what i have learned the time i did this Hobby

But i am far away from being an Ingeneur in Accustics so i learn also every day something new ;)

Sad i cant do more testing right now with my Surround Setup and present Measurings but i will one day ;)

Need to find an new Home and than i am not sure how nice the Neighbor would be if i do all that acustic Measurings lol ....so much to do.....

Also the FIlter Math in an DSP add rounding errors but at 64 Bit like Camilla DSP i didn't think thats an huge Problem or you can her it, 32Bit has already 16Milions steps in Voltage Level..... if yuo lose one step in short time periode how you will ever hear it.... Onlly thomething happen on Most significant Bit could drop the Level half way thats gif you probably an short pop or noise but thats would be an real huge Error.

I use Computers for long Time and my Files i save still have the Same Data after years if nothing extreme is happening but it could flip an Bit over long Time sure.....That why the Most stuff use CRC or anthing like this to check Data while transporting it digitaly....

Thats waht Google say ... ;)

1. Digital Filter Passband Ripple
Sabre DACs feature built-in digital FIR filters. When these filters upsample digital audio (e.g., standard 44.1 kHz Redbook audio), they can introduce a tiny amount of sinusoidal ripple in the passband. [1, 2, 3]
  • Typical Spec: \(< \pm 0.005 \, \text{dB}\) to \(< \pm 0.001 \, \text{dB}\) depending on the selected filter preset (e.g., Fast Roll-Off or Linear Phase).
  • Audibility: This sub-fraction of a decibel ripple is considered inaudible to human ears but is mathematically tied to tiny pre- and post-echos in the time domain.
  • Customization: Flagship Sabre chips allow users to bypass the internal filter or load custom filter coefficients to eliminate this ripple. [1, 2, 3]

Different typ of Filters Ripple

5a28b5d2a1737208db4e26d4e8393c8c561db082.png
 
Last edited:
Back
Top Bottom