[hpsdr] FIR filtering and HPSDR

[Alberto I2PHD]

On 9/3/2014 9:08 AM, Erik Anderson wrote:

> /I'd still like to know about how many taps people end up using when doing convolution, mostly to know //
> //whether there's any value in adapting the 16-tap pure-register SSE implementation I found./

IMHO 16-tap FIR filters are quite on the low side, at least for a bandpass filter ahead of the demodulator...
as an example in Winrad I used a fast convolution filtering, whose mask was computed starting
from a 1527-tap kernel (1537 = 2048 - 512 + 1)

In the firmware of the new Elad DUO, running on the Cortex ARM M4F, I used a fast convolution
whose mask started from a 769-tap kernel (769 = 1024 - 256 + 1).

Both implementations have very good skirts, as shown by the Matlab simulations done when computing
the kernels.

-- 
/*73 Alberto I2PHD*
Credo Ut Intelligam



/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20140905/a993e1e3/attachment-0003.htm>