[R-390] Why use a Roofing Filter?

[Charles Steinmetz]

Roy wrote:

>I suspect that there are some of us who: - are not all that sure 
>what a roofing filter is - wonder why the R-390A needs to be 
>"improved" by adding one - think that maybe the conditions under 
>which we use our radios at our places, does or does not warrant the 
>improvement.  So, a short description of what the thing is, where it 
>goes in the radio, and why it might be an improvement would be welcome.

Basically, it is an IF filter that sets the maximum bandwidth of the 
system.  To be effective, it needs to go as far "upstream" as 
possible in the radio, to keep out-of-band energy out of as many 
stages as possible.

These days, it is very common to make general coverage radios by 
upconverting received signals to a VHF first IF frequency (70 MHz or 
thereabouts is common).  This gives good image rejection, but exposes 
the radio to out-of-band energy at VHF frequencies.  Most radios do 
not have sharp RF filtering (because it is hard to get the RF filters 
to track), so this is a problem.  Enter the roofing filter -- 
installed at the output of the first mixer, it limits the frequencies 
that can enter the VHF first IF.  Typical BW is 20 kHz -- wider than 
the widest filter bandwidth in the final IF.  (These radios often 
have no RF amplification, and have "brute strength" first mixers and 
post-mixer amplifiers with 3rd order intercepts in the +40 dBm range 
to handle the strong out-of-band energy they will receive.)

Contesters (people who spend their radio lives trying to pick weak 
signals out of pile-ups) frequently install much narrower roofing 
filters, to improve the closer-in overload performance of their 
radios.  It is not uncommon for these folks to install roofing 
filters that are only a few kHz wide.  Again, this filter needs to go 
as far upstream as you can get it -- at the output of the first 
mixer.  (Note, however, that trying to design VHF filters that narrow 
is a losing proposition.  If that sort of performance is what one 
wants, better to start with a single-band, downconverting rx 
architecture instead of a general-coverage upconverting rx.  That 
also allows you to make the RF filters much narrower, too, which 
further improves close-in IMD performance.)

The retrofitted "roofing filters" for boatanchors (and, in 
particular, for the 390/390A) are typically installed much farther 
downstream for convenience, thereby pretty much nullifying most of 
the benefit by leaving all of the preceding IF circuitry 
unprotected.  In the case of the 390/390A, because the VFO feeds the 
last mixer, the preceding IFs must be wide to accommodate a whole 
band -- so any roofing filter placed where it really needs to be to 
do its job would need to be a tracking filter.  In practice, people 
put them after the 3rd mixer, generally ahead of the existing 455 kHz 
IF filters (the mechanical filters, in the case of the 390A).  Placed 
there, the "roofing filters" can clean up the stop band of the 
narrower mechanical filters, but that's it.  And since the real IMD 
limitations in a 390A are the RF Amp (V201) and the First Mixer 
(V202), the retrofitted "roofing filter" can't do anything to improve 
the weakest links of these radios.

In sum, the overall architecture of a 390A does not accommodate a 
real roofing filter.  People add what they think are roofing filters 
anyway, because they've heard that it is a good idea.

A 6 kHz filter added to a 390A DOES reduce the close in IMD -- but 
ONLY compared to the existing 8 or 16 kHz filters.  The existing 4 
kHz filter is better than the added 6 kHz filter.  So the improvement 
is not a matter of "roofing," it is simply a reflection of the fact 
that narrower IF bandwidths have better close in IMD performance that 
wider IF bandwidths -- it is inherent in the nature of close in IMD 
measurements.  Replacing the existing 8 kHz filter with a 6 kHz 
filter (or using the 4 kHz filter) would do the same thing.

I concur that in today's band conditions, the existing 8 and 16 kHz 
filters have no practical use, and that a 5 to 6 KHz filter is 
optimal.  If I were choosing a filter array from scratch today, I'd 
probably choose 1.5 kHz, 2.1 kHz, 3 kHz, and 6 kHz.

Best regards,

Charles