vax71

Hello

I'm new with kx and DSP, but I had an idea/question.
I browsed many threads but did not find any related topic.

I know the DSP only has 16bits inputs/outputs, but i think 24 (or 32bits!) samples might be handled.

1. The driver splits each 24bit sample : 12bit less significant bits and 12bit most significant bits
2. The LSB part is send to DSP port 1, the MSB part is sent to DSP port 2
3. A DSP plugin read both ports and rebuild the 24bit sample
4. etc...

A 96kHz stream might be handled in a similar way :

1. The driver splits the 24bit stream in two streams : odd samples and even samples
2. The odd stream is sent to DSP port 1, even stream is sent to DSP port 2
3. The DSP plugin handles both stream and mix them
4. etc...

In both case, the result would be split by the DSP and send to the CPU. If needed, the CPU might send the rebuild stream to p16v 24 DAC.
Of Course, all plugins would need to be rewritten, and would be fatter!

Is all this stupid ?

Thanks

Max M.

strange, there're a few topics of that here and there. This is how actually kX supports 24 bits
(~30 bits actually) w/o p16v (although initiated a while ago, this feature was finally implemented in v3545).
Regarding the 96kHz - this is how Emu cards support it (the "splitted" streams are merged by FPGA before DACs).
There're too few benefits of such complicated flow, as it renders the DSP to be sort of useless (if it is not already) - it is much more effective/flexible to do whatever 96kHz processing in software and route things directly to p16v. (The CPU->DSP->CPU->p16v flow would give neither less CPU load, neither better latency, nor better stability)

Absolulty no, it is not. Just late a bit

__________________

vax71

Sorry This forum is quite huge

Everything is said

Thanks!