[Elecraft] K3 filter performance

[Don Wilhelm]

Larry,

You are quite correct that all filters are not created equal.  For 
digital modes, the group delay is just as important (if not more 
important) than a flat passband and steep filter skirts.  In fact, those 
filters with steep skirts often sacrifice group delay to achieve the 
steep skirts - but I am generalizing here, and that may not be a 
universal truth.

Unfortunately, group delay plots are not common for filters since steep 
skirts seem to be the 'criteria of choice' for most amateurs.  A 
Gaussian to 6 dB filter has a nice rounded nose and gentle skirts, but 
has a great group delay characteristic, OTOH, Cohn filters usually have 
great skirts and poor group delay in the passband.  Group delay will 
make a difference in the ability to decode digital signals.

73,
Don W3FPR

Larry Molitor wrote:
> --- "Joe Subich, W4TV" <lists at microham-usa.com> wrote:
>  For a died in the
>   
>> wool 
>> RTTY DXer, a steep sided 270 - 300 Hz filter would
>> be nice
>> instead of 370 Hz from the "250 Hz" filter.
>>     
>
> Over the years I've spent a lot of time looking at
> this issue. My interest is primarily digging weak RTTY
> DX out of the noise but I do occasionally get into a
> contest. 
>
> I've played with a lot of radios and filters but most
> of my real testing was done with a FT-980, a TS-2000,
> and a FT-2000. 
>
> While I have seen good weak signal performance with a
> 250 Hz filter, it was the exception not the rule. As
> an example, the improvement on the FT-980 going from
> the stock SSB filter to the dual CW filter was
> dramatic. But going to the 250 Hz CW filter lost about
> 6 dB in ability to properly decode weak RTTY signals.
> It was even worse on signals that had polar flutter. I
> no longer have the plots of these FT-980 filters but
> as I recall, the 250 Hz filter was a bit peaky in the
> middle and had poor group delay characteristics
> extending well in from the corners.
>
> As you say Joe, the determining factor is the passband
> ripple/group delay. In a typical bandpass filter the
> group delay goes to heck at the corners. But the shape
> of the "corners" varies from filter to filter. I've
> run a number of "ham filters" on a network analyzer
> and plotted group delay. Seems like no two filters are
> the same even if the same part number. This I believe
> is due to ham filters being so cheap and manufacturing
> process control being minimal to keep the sell price
> down to what we can afford.
>
> So I would say, if you have a 250 Hz filter, no matter
> where it's placed in any radio, give it a try on very
> weak signals. Switch between a wider filter and the
> narrow filter and see if there is any degradation. If
> not, use the narrow filter.
>
> Other wise, plan on not using any filter less than 300
> Hz in passband width (note - this is NOT the 6 dB down
> width!!! very important!!) if you want optimum weak
> signal performance.
>
>