I am using SHARC 21469 with AD1939 for my audio application.
I want to play and record a stimulus sampled at 1kHz for 20s. As the minimum sampling rate supported by AD1939 is 8kHz, I upsampled my stimulus so now it has 160,000 data points. The problem is I can't read the whole stimulus file into data memory as it doesn't have enough space. There are 2 routes I can think of -
1) Split my stimulus file into 32 parts of 5000 samples each and then read all the files and store the stimulus into DDR2 ram.
2) Split my original file at 1kHz into 4 parts of 5000 points each and then use FIR filter to extrapolate to 8 times. I am not sure this will work as the SHARC document says that the FIR accelerators only work with data memory.
What is the best way to solve this problem?