[Elecraft] What crystal filters for the K3 are linear phase delay (constant group delay) ?

[Jim Brown]

On 4/23/2014 10:13 PM, Larry Lopez wrote:
> Some of the wider filters in that paper are pretty good.
> The narrowest filters are not so good.
> The flat part of the passband has constant group delay.
> The skirts are awful

Nothing at all surprising. In general, phase distortion (group delay) 
goes along with variation in amplitude response, steeper skirts results 
in more phase distortion in the passband. That means it's greatest with 
steep filter skirts, AND extends within the passband from the filter 
skirts. DSP signal processing is a simulation of analog circuits -- when 
you do filtering in DSP, it's a simulation of an analog filter, and it's 
far easier to build a filter in DSP with low phase distortion than it is 
with physical components.

In general, when phase shift is an issue, we want to use DSP for the 
narrow filtering and fairly wide analog filters (the crystal roofing 
filters) to protect the DSP from overload. And when phase shift is an 
issue, wider bandwidth is better.

The current wisdom from those who "get it" is to use a 400 Hz roofer for 
most narrow digital modes (RTTY, JT65) and set DSP for about the same 
bandwidth. K1JT, the Nobel laureate who gave us the WSJT modes, says to 
run the RX broad and let his software do the filtering.

73, Jim K9YC