[TenTec] Roofing Filters

[Dr. Gerald N. Johnson]

On Wed, 2006-05-31 at 10:23 -1000, Ken Brown wrote:
> I have received quite a few responses to my request for a clear 
> definition of the term "roofing filter". Most of the respondents seem to 
> think that they know exactly what a roofing filter is, yet they don't 
> all agree exactly with each other. The best (in my opinion) and most 
> thorough response was from Bill W4ZV, who also included some reference 
> links.
> 
> I got a chuckle from the opening sentence of one link, from Inrad: "  
> "roofing filter" is the current buzzword in high-end radios.  "
> 
> I agree with that, though there is no doubt that a better first IF 
> filter can make a big improvement in receiver performance.
> 
> The way the term is being used in some advertisements sort of reminds me 
> of  "Solid State", "Transistorized", and "Digital". And remember "Color" 
> TV antennas?
> 
> Most everyone agrees that (lately) a roofing filter is the first 
> bandpass filter after the first mixer in a superheterodyne receiver. 
> (Once upon a time a roofing filter was a lowpass filter, not necessarily 
> in the IF of a receiver.) There is some consensus that a roofing filter 
> is not the narrowest or final filter (or filtering process in the case 
> of a DSP receiver) in the IF. Perhaps the later, narrower filters are 
> "ceiling filters." It seems that we have always had roofing filters, we 
> just didn't call them that until a few years ago.
> 
> There is also an implication that a roofing filter must be a crystal 
> filter with a fairly narrow bandwidth, on the order of  a few times the 
> bandwidth needed for the mode of operation. I have not yet heard anyone 
> refer to an LC BPF or IF transformer as a "roofing filter". Or perhaps 
> the first filter after the mixer is always a roofing filter, regardless 
> of type, or bandwidth?
> 
> Is the first IF transformer in my Hammarlund Super-Pro a roofing filter?

Yes. But not very narrow.
> 
> DE N6KB
> 

The classic receivers developed all the selectivity in the last IF
through a gang of IF transformers. That allowed adjacent frequency
strong signals to overload mixers (and mixers are delicate things though
noisy they are easily overloaded). Getting good selectivity at 1.6 MHz
or even 455 KHz took high quality (expen$ive) IF transformers. But the
single conversion receiver with 455 KHz IF has image rejection problems
about 15 or 20 MHz, it being inconvenient (co$tly) to make tracking RF
stages to achieve image rejection, so the multiple conversion receiver
was created. That made the RF image rejection easier and initially the
only need for selectivity at the first IF was to reject enough of the
image of its conversion to 455 KHz. That was a roofing filter. Long
about 1959, I think it was Byron Goodman wrote about Modern Receiver
design in QST where he exposed a single conversion receiver with
filtering right after the first mixer, getting the filtering as close to
the antenna as possible. Both it and the conventional (ala Hammarlund)
receiver depended on a band switched local oscillator which tended to
wander at best, sometimes it ran away. Collins came out with the A line
and then the S line where the first conversion was crystal controlled
and the VFO had no switches because it always covered the same range.
Then the first IF had to either track the VFO (as in the 51J and R390())
or be as broad as the band segment covered per crystal (as in the S-
line). While the Collins scheme drastically improved stability it left
the mixers, especially the second mixer open to overload from strong
unwanted signals.

I improved my 75S-3B when used with a 2m converter by replacing the
noisy 6AU8 second mixer with an Amperex 7788 and then I lowered the
front end gain drastically improving the intermod response of the 2m
receiving system.

Today most all receivers to make use of synthesizers (that have far more
phase noise than crystals and Collins PTO, but that's another topic) and
multiple conversions. Each stage of conversion adds noise and limits
strong signals because of the delicacy of the mixers. There is a roofing
filter after the first mixer, but its often far wider than the final SSB
filter to allow for receiving FM or AM and for passing noise with a
spectrum wide enough for good noise blanking. There is that trade off,
that the pulse width of a noise pulse (ignition, lighting, or power
line) gets wider each time the filter gets narrower. And in some filter
designs it gets a lot wider (longer in time) because of undamped filter
ringing.

Because FAST ADC are expen$ive, its cheaper to mix down to a low IF
frequency at the cost of more limitations of filters and of mixers
before the digital portion of the digital signal processed radio. And
that front end roofing filter sets the intermod bandwidth as well as the
limits on how far the DSP can tune without having to move the
synthesizer, or on how many independent signals the DSP can detect at
the same time.

Byron Goodman had it mostly right. He demanded a square cornered
amplitude response curve that is responsible for the worst of the filter
ringing and he ignored the IF stage noise that is broad band after that
first filter. The better receiver needs to include a time response
specification on the filters (and Tentec's are good about their time
response compared to the S-line vintage mechanical filter) and needs to
include another filter right before each mixer (whether to a new IF or
to audio as the product detector) to clear away IF stage noise on the
image frequency of that mixing stage, and then there needs to be a good
low pass filter at the speaker to take away audio distortion harmonics
and audio stage noise. That audio analog filter is needed even more in
the DSP radio. But few have it.

When the roofing filter is wider than the last filter, its common for a
signal in the roofing filter pass band but not in the output to pump the
AGC confusing the user about the qualities of the desired signal. This
has been true for all audio filters outside the AGC loop since such
filters were invented (by Armstrong perhaps, if not earlier) whether
digital or analogue.

There is a long history of receiver design with some decisions made
arbitrarily, some driven by available components, some that proved poor,
some that proved to be good, and some that should have been forgotten.
The SSB filter at 15 KHz, typical of the Orion DSP radio was also one
technique used with analog filters back about 1956, and then abandoned
because the extra mixers hurt the receiver performance too much. Modern
mixers if chosen and applied properly can do significantly better than
the tube mixers of yore, but still are the limits of receiver dynamic
range at both extremes.

-- 
73, Jerry, K0CQ,
All content copyright Dr. Gerald N. Johnson, electrical engineer