[Elecraft] Re: roofing filter loss and method of compensation

[Bill Tippett]

KK7P:
 >With the IF DSP in the K3, the issues that are 
of concern in a fully analog radio are less important.

 >As long as the crystal filter, which is being 
used as a roofing filter, has reasonably good 
skirts and ultimate rejection, the DSP will take 
care of the rest. The first 20 dB or so of skirt 
selectivity, the passband ripple, and the 
ultimate rejection at the DSP IF image frequency are the most critical factors.

         Below W2VJN (ex-Inrad) says the first ~30 dB of
skirt selectivity is key for IMD issues.  It will be important
for you to specify actual IMD and BDR performance with
each of the various CW filters installed.  So far I count 4
that are 500 Hz or less, not counting the variable BW
option.  I hope you can post this information soon so
that those planning to order will have information to
properly configure a unit.

                         73,  Bill  W4ZV

http://www.qth.com/inrad/roofing-filters.pdf (page 6)

5. If 6 poles work so well, why not 8 poles?
The most important part of the filter 
characteristic is from the pass-band on down
to about –30 dB on either side of center. Eight poles would provide much better
stop-band isolation, but it’s not required in a 
roofing filter and would make no
noticeable improvement in IMD performance.