This was recently discussed on the SDR# yahoo group, a few people claim to have this working. The common denominator seems to be fast multi-core processors and tons of RAM. Personally, I've tried for weeks with no success. One problem is that DSD requires a pretty stable 48000 sample rate. SDR# puts out 32000 with most configurations.
There is a setting in the SDR# config file that sets the minimum output sample rate that can be set to 48000. Unfortunately, this sets the minimum output rate but it usually is not 48000. The only way I could get a 48000 was to modify the source to add a dongle sample rate equally divisible by 32 (seems to be the default decimation factor) of 1536000. This gave me 48000 output sample rate but still DSD had problems decoding P25.
The only near error free decoding I could get while testing was with the default 32k output rate recorded to a wav file, loaded into an audio editor and re-sampled to 48k. This made me think the root problem was the SDR# output rate. I created a real-time re-sampler that took SDR# 32k input on one virtual cable re-sampled to 48k and output on a second vc. Still no joy. Then I hacked the SDR# source to get the 48k output, still no joy.
This was with Intel dual core 2Ghz processors and 2G of RAM.
