I've been experimenting with concepts of using the serial data outs for SPI control of other ICs, where an ADAU1701 is used in self-boot mode and there is thus no MCU. This is because my designs are so small I usually don't have space for anything but a handful of other components, and I can't fit an ADAU larger than -1701.
My applications are too specific to be of general use, but I had a concept for external delay using an SPI SRAM, as the newer ADAUs can do natively. This requires four serial data lines; BCLK, one data in & out, and CS (I think this can be a GPIO).
Some time ago KJBob helped me describe a method for using downsampling with -1701's internal delay to extend the initial delay for a reasonable reverb. See here for an example of downsampling with the internal buffer:
Using an extension of this technique, an external delay has the advantage of a delay time limited only by the external SRAM addresses (64K, for 16K sample delay). This is a fourfold increase!
The SigmaStudio program would involve a counter that is used to encode the memory address for the read/write commands, and transmitting that command as an audio sample, followed by several samples of actual audio data. It would have to use -1701's internal buffer to read & write a page of bytes back and forth to the SRAM--this would be a short buffer that has to multiplex & demultiplex the mono audio from the stereo data stream, using the technique in the link above.
It takes -1701 five sample periods to read or write a full page, which is eight (mono) samples of data (because -1701 only addresses 24 bits in a 32 bit word), plus another period to toggle CS between commands, so an eight sample roundtrip takes twelve sample periods. Probably easiest to downsample for six samples in that period.