The native resolution of the LMS7002M is 12 bits. However we currently represent this as a Complex{Int16} for convenience. Once the IQ and channel interlacing and ordering is verified, bitpacking to a ComplexInt12 should provide more efficient use of bandwidth. Of course this adds some processing on the GPU or CPU, but it may provide a nice opportunity for multiple dispatch to shine.
I think transferring raw 12-bit samples as 24-bit (3-byte) ComplexInt12 complexes should be relatively straightforward. It should give us a cool 25% transfer and storage savings.
Transferring @sjkelly's issue: https://github.com/JuliaComputing/DARPA-TRIAD/issues/29
I think transferring raw 12-bit samples as 24-bit (3-byte) ComplexInt12 complexes should be relatively straightforward. It should give us a cool 25% transfer and storage savings.