[Elecraft] Roofing filters and DSP bandwidth tuning

[Jim Brown]

On 1/27/2021 3:06 PM, David Woolley wrote:
> Having the roofing filter too close to the DSP filter is not necessarily 
> a good thing, as the roofing filters are likely to have worse passband 
> ripples and will have non-linear phase responses, which can compromise 
> digital modes.  At least some of the DSP filters are finite impulse 
> response, meaning they are also linear phase, which means that pulses 
> will not get smeared out.

K1JT strongly urges 3kHz or greater IF bandwidths for his modes, and for 
exactly that reason. And it's why top RTTY contesters have abandoned the 
K3's dual-peak filter in favor of 500 Hz IF bandwidth. My professional 
life in audio system design taught me that speech intelligibility is 
degraded by time/phase distortion and suspected the result would be the 
same with RTTY, but I was derided when I started preaching that to RTTY 
guys. Several years later, author of the 2Tone RTTY software G3YYD said 
the same thing, and folks started believing it. It's also why I find 
that the 2.1 kHz 8-pole provides better speech intelligibility than the 
1.8 kHz filter.

Years ago, I tried using narrow SSB realignments of the K2's CW crystal 
filters in contests. I had carefully tweaked them per the build 
instructions, noting that their amplitude response looked like the 
profile of a mountain range. I wasn't surprised that those settings made 
signals much harder to copy. The SSB TX filter sounded fine on RX.

When I switched from the 2.7 kHz 5-pole filters to the 2.8 kHz filters 
for TX in one of my K3s I noticed considerably less incidental AM on 
RTTY, and immediately converted the other two. I didn't do that in the 
2nd RX, because I only use it for weak signal CW work on the lower bands.

73, Jim K9YC