I would like to use 32 bits fixed point (1.31) on an adau1451 instead 8.24. Is it possible on this target ? How do I have to do to configure all flux audio in Sigma Studio ?
In fact, I made a filter IIR under the canonic form (IIR+FIR) with low level function, to reduce rounding error at the maximum and maybe also to reduce the cycle number. I don't know how the IIR module of Analog Team are coded.
Thank you in advance for your answer.
Yes, you can bypass the truncators and saturators and process 32 bit data into and out of the core.
Mouse over the input cells and the output cells, right-click and then select "Data Format" and you can set it to 32 bit mode. The serial port must be set such that 32 bit data is being transmitted. So standard I2S is fine or TDM with 32 bits per slot is required. If it is packed 24 bit or 32 bit then you will not get 32 bits.
Here is a screenshot to see what it looks like:
Then your question about how the IIR filters are processed I think can be answered in the datasheet:
The DSP core is 32-bit fixed-point, with an 8.24 data format for audio. The four multipliers are 64-bit double precision, capable of multiplying an 8.56 format number by an 8.24 number. The multiply accumulators consist of 16 registers, with a depth of 80 bits.
The general 2nd order filters we use are Direct Form 1.
Thanks a lot.