I'm planning my next project which is to implement 3 digital filters to drive a 3 way speaker without crossovers, in effect simulating Linn Aktiv but digitally instead of using analogue filters.
The digital filtering is to be done on one or more Raspberry Pi pico devices, circa £5 each.
These have a 32 bit cpu running at 133
The rough idea so far is to have one pico supplied with s/pdif from a cd player/transport and split it into 2 parallel data streams for left and right channels.
Then there will be 6 more picos, 3 for each channel to perform LF, mid, and HF filtering. The output from each, probably converted back to s/pdif, feeding a DAC (6 of).
I need to investigate filtering algorithms but a rough estimate for a 44.1 kHz 16 bit data stream, a 133MHz pico be able to perform 1500 instructions (assuming 2 cycles per instructions, most instructions take 1 cycles) so I'd have thought that is more than enough.
Has something like this been done before?
I'm looking for C source for decoding the s/pdif signal and for performing filtering on a data stream.
Note this is purely for fun to see if I can do it so not intending it to be a commercial product.
The digital filtering is to be done on one or more Raspberry Pi pico devices, circa £5 each.
These have a 32 bit cpu running at 133
The rough idea so far is to have one pico supplied with s/pdif from a cd player/transport and split it into 2 parallel data streams for left and right channels.
Then there will be 6 more picos, 3 for each channel to perform LF, mid, and HF filtering. The output from each, probably converted back to s/pdif, feeding a DAC (6 of).
I need to investigate filtering algorithms but a rough estimate for a 44.1 kHz 16 bit data stream, a 133MHz pico be able to perform 1500 instructions (assuming 2 cycles per instructions, most instructions take 1 cycles) so I'd have thought that is more than enough.
Has something like this been done before?
I'm looking for C source for decoding the s/pdif signal and for performing filtering on a data stream.
Note this is purely for fun to see if I can do it so not intending it to be a commercial product.
I've implemented digital FIR filtering on Cortex M0 and M3. There's plenty enough horsepower on a 133MHz M0+ to do a hundred or more tap FIR, stereo with 16bit data and coefficients, when coded in assembler. And you have two cores on the Pico. For C, I don't know.
One of the biggest headaches is in the realm of clocking - do you want the CPU to run synchronously with the audio? Normally that gives the most flexibility in peripheral clocking but when the audio signal goes away or maybe just glitches, what happens to the CPU? Perhaps if you implement S/PDIF decoding in software this problem might go away, but with a software decoder you'll still need a hardware oscillator to give yourself an MCK (or BCK if you prefer). Software oscillators (NCO for example) are just too jittery for audio use unless you go 'the full monty' with a DAC, filter and comparator in hardware.
Over on the RPi forum there is a short thread on S/PDIF decoding : https://forums.raspberrypi.com/viewtopic.php?t=318134. As you'll need an MCK/BCK anyway though I would go for a hardware decoder such as DIR9001 or AK4113 and focus your energies on the filtering stuff.
I am not aware of the device you mention, it may do what I require, I chose the Raspberry Pi pico due to cost. If the project fails, I won't have wasted much money, time is a different issue if course.