[Elecraft] K3S DSP filter plots

[David Woolley]

Having a constant slope makes a lot of engineering sense.  I believe, at 
least at some bandwidths, the K3 offers both IIR and FIR (infinite and 
finite impulse response filters).  For data modes, FIR filters are 
better, as they are linear phase filters.  The skirt slope of such a 
filter is directly proportional to the number of taps used in the filter 
(the number of old samples added together).  Having a large number of 
taps increases the processing load on the DSP, so will be limited by the 
need to work within the processing power of the DSP device.  Also longer 
filters mean larger group delays.

IIR's have variable group delays and ring, but can produce responses 
that could only be approximated by very long FIR filters.

Incidentally, merging the roofing and DSP filter edges is likely to 
compromise the good group delay, and lack of ringing, characteristics of 
the FIR filters.

-- 
David Woolley
Owner K2 06123


On 16/11/15 15:59, charles at k5ua.com wrote:
> I'm hoping that the constant slope you described is not the way the DSP
> filters are implemented, because that would mean the 50hz filter has a
> 5.625:1 shape factor, the the 100hz filter has a 5:1 shape factor, the
> 200 hz filter has a 3:1 shape factor and the 400hz filter would have a
> 2:1 shape factor. To achieve better than 2:1 shape factor at 30dB per