If you call RF TX/RX transmission pseudo isolation, then, well, be it so.
I have used that successfully in several designs.
One other question:
How do you plan to handle the collision of i2s4out and INT4 coming from the INTB of MCP23017 on U2 of Zynthian V5, V5.1 and Mini V2 boards, which handles the rotary encoders?
Are you proposing using radio frequency transmission and reception within the soundcard? I donāt understand your proposal for isolating each input and output.
Hi @fussl , as the thread title suggests, this is a ālightā codec meaning with no high end components and affordable.
And, as you have noticed, thereās not much products on the market with such a setup, from my knowledge they are the AudioInjector isolated and an IanCanada accessory
Anyway, Iām quiet currious about this. What type of ICās would you use for this ?
The INTA and INTB can be changed and configured from webconf so far I remember (I do not have my Zynthian under my hands right now to verify).
@le51 explained it well. This is cost effective concept card developed to test new rpi5 features for multi channel i2s.
If the goal was to develop high end, fully isolated audio card that utilise all 4 channels we would use different dac adac chips.
V5 canāt use this card because it does not have hat. I am working on Mini v3 that will incorporate this design on the board but not for all 4 inputs and outputs. There will be two inputs and 2 outputs on mini v3 so interrupt pins will still be available. I am also thinking to remove 16 switches (second mcp23107) and go back to 4 encoders with switches and additional 4 switches). This decision will depend on successful adoption of @wanthalf implementation of onscreen buttons that are currently merging to development branch. In this case additional cards would be possible to add to mini v3 hat and utilise all 4 i2s channels.
Yes, and if this is set, the WSLEDS on the keypad will no longer work, as there is some false logic in zyngui, which makes them work only for V5 wiring. There is a similar thing with touch rotation, btw.
The additional buttons of V5.x are such a benefit to the workflow, as are the LEDs showing actual functionality, it would be a shame to drop them.
The Ian Canada accessory is more of an audiophile nonsense and due to a delay of a full second in audio it will be unsuitable for a synth, anyway.
So, for all goodness, make it as cheap as possible, if you want to.
The good things will come on another train.
edit: sorry. I have no right to criticise anything here.
Meaning leds to button assignation are hard coded ?
Yes and quiet expensive if you go with this kind of setup 
Do not apologise, it isnāt criticism itās observation and itās all welcome.
With only positive feedback any engineer will tell you you are going to dissolve into perpetual howling feedback or slam into either PSU rail.
Negative feedback is where control lies.
Many of our political leaders have built careers on ignoring this simple premise.
Donāt worry.
Yes, configurable audio card is not for everyone. It will come with default config that works out of box. Configuration only need to be changed for additional stacked card(s). First card always goes with default and any additional stacked card need to be switched to slave mode, oscillator disabled and different i2s out in channel selected based on need (available i2s channels pins). That is not that hard to explain in documentation. So people can have just one, two, three or four of them.
Are those audio isolators on a purely electronic device? 
no and yes 
these are digital signals isolators used between a noisy RBPi and the codec
like the H11L1 āelectronicā device between Midi In and RBPi UART in Zynthian HW
Heh, I meant the gold spikes at the corners 
For the picture I presume it looks smart 
Where the world of sensory perception meets technical ignorance with a mystical whiff of snake oil and rook soup . . .
May my copper be more oxygen free than yours . . .
Look here:
https://www.skyworksinc.com/en/Products/Isolation/Si86xx-Digital-Isolators
They are fast, reliable, made for the purpose and work with RF transmission @riban and no, not pseudo. Look at the isolation properties and speed, and the default state in case of input HiZ/undefined. It is important to use a single IC for signals that have a timing relation for an isolated branch and to not allow the isolated parts to couple too much into a device āgoundā-plane (or call it noise plate with switching power supplies). Distance matters, as little circuit mass and surface as possible ā matters.
Snake oil and gold spikes ā nonsense.
Thank you @fussl. It satisfies my curiosity. 
Available chips in this series at JLCPCB (4 different right now, but I donāt know if they are accurate though) are between 2 and 5$ piece. Some, that are not in stock, are charged more than 15$.
Two of them would be needed for isolating all the RBPi5 I2S data lanes, so yes, it add costs and complexity.
Is it worth the effort ? I canāt really tell
To isolate a branch, an isolated DC-DC converter and regulators, filters, etc, wil be necessary for each to isolate power rails as well.
For a really good result, Iād isolate each ADC and each DAC separately.
The transmission through the isolators is unidirectional, so for each selectable master / slave config DAC, a 3+1 and a 4+0 would be necessary for each possible master, so itās three chips per card, plus two DC-DC, plus voltage regulators, capacitors, etc. Just isolating the entire card would still give some benefit and save one DC-DC, one set of regulators, etc, and a 4+0 or 3+1, dependent on whether ADC or DAC is master.
You get what you pay for.
Si860x canāt be used to isolate internal clock because it only support the speed up to 1.7MHz while internal clock on this card is 24.4MHz.
Also Si860x is a i2c isolators, can it be used for i2s ?
This audio card does not use i2c protocol. Only i2s with following lines: BCK, WCK, and four pairs of DataIN/DataOUT. It also has internal clock. So if the same IC can support i2s and then just to isolate i2s participant for one i2s channel we will need 7 isolators:
Is my assumption correct?
No. Sorry.
Si8662BC, look for the numbering scheme, 6 of which 2 are reverse channels (leaving 4 forward channels). 150MHz.
Okay - I have looked at the datasheet for the Si8662BC and understand that it is a real, air-gap electrical isolator that uses radio frequency amplitude modulation to transfer data. This is both ingeneous and crazy! I wonder how effective the reduction of digital noise is when we are introducing lots of radio frequency transmitters in the heart of our audio board!
I also query what problem is being solved? It is important to decouple the digital noise from the audio and there is significant advantage to decouple the audio sender and receiver to reduce ground loop hum. These are different requirements and neither seem to relate to the digital signal decoupling that the Si8662BC provides.
@fussl please describe the problem that you are trying to solve by using these devices?
Regarding analogue audio decoupling, this can be done by passing the audio through a balancing output, ideally an audio transformer but more likely electronic differential amplifier and doing similar on the input so that you have a differential (balanced) transmission line with common mode noise rejection.
Regarding digital noise decoupling, this can be done by creating a clean analouge supply. Most good DAC/ADC support an analogue supply and digital supply for this reason and do the decoupling internally, i.e. the chip already does this job. We just need a clean supply which the LDO is supposed to do.
