[HBR] The long, SLOW HBR project

[Walt Hutchens]

> Before continuing with coil work, however, I'm going to see if I can improve
> the shape of the crystal filter a bit. One peak is higher than the other which
> causes uneven voice frequency response.
> 
> Tuning the 1st IFT gives you some shape control but the choices are a LOT of
> difference between the peaks and more equal peaks with a big valley in
> between. 
> 
The half lattice filter is two crystals, nominally 1.5 kcs apart, custom
made at the same time.   However in the filter circuit the lower frequency
crystal passes somewhat more signal than the other, making the response
lopsided.   (Generally true of such filters.)  The response of a crystal
isn't symmetrical and the overlap between the two thus affects them
differently.

Sometimes you can get rid of this by tuning the IFT, sometimes not, however
for a two-crystal filter this will be a compromise between balancing of
peaks and skirt selectivity.  The details depend on the peak spacing, IFT Q
and similar things.

For starters I did several experiments with shunt and series capacitors that
reminded me of how half lattice filters work:  The two crystals must see
exactly equal signals (but of opposite phase) so the holder capacitance
doesn't couple any net signal past the filter.  And of course any shunt
capacitance passes so much signal that the filter skirts are ruined.

That left a change in the load resistance.   I cut that down from 33k to
3.9k and that's a whole lot better, though the gain is somewhat lower.

This got the audio to 'okay.'   The gain on 80M is about right and
adjustments are possible for other bands if needed.   I expect 20M is a bit
low but that's due to lower RF stage gain coming from the (unavoidably)
lower Q of those coils and should be fixable in the coils.

Well, fixable if I wind another mixer coil ... lots of coil winding practice
with this project!

A half lattice isn't a perfect filter for a communications receiver: It is
way better than most of us had as novices, roughly the selectivity equal of
the double conversion to a low IF scheme (worse far out skirts but better
ability to handle close in strong signals), and plenty good enough for a
simple but very effective set like an HBR.   However there are better
filters out there -- the HR-20 has one -- and if we want to beat a half
lattice, something like that is the way to go.

Another possibility would be a higher IF -- say 5 to 9 Mcs -- and a ladder
filter.   Those things are EASY to make -- four to six identical crystals
and a handful of tiny caps -- and will equal just about anything out there.
The higher frequency crystals would be easier for an HBR because you can
more easily get the 2.8 or a bit more kcs of bandwidth needed to copy AM
decently and one of the crystals probably can be pulled to act as a BFO.

You can't really be sure about warm up drift of a coil on a plastic form
until it is a few weeks old as evaporation of solvents and squirming of the
windings on a new one happen more when it warms up than at other times.
(And we wondered why those military sets all used porcelain and machine
wound #16 wire ...)  Rechecking the 80M coil showed less than 200 cps down
over about 20 minutes and then within a 50 cps range.   I expect this coil
to require less temperature compensation than the others so I'll put that in
the chassis.   

Listening on 80M after doing the filter tweak revealed that I again had
warble on CW signals and of course that means distortion on SSB voice.   It
didn't take long to discover a tiny bit of ripple on the new separate (12AL5
rectifier) plate supply for the LO.   I added a second section to the filter
for these two voltages.   (Now 100 mf -- 1000 ohms -- 100 mf on both
+/- 140V lines.)   One disadvantage of this receiver design is that it has
more circuits requiring absolutely pure DC than most:  Both AGC and audio
detectors, the LO, and the BFO.

Better, but still some warble: Improved bypassing of the oscillator cathode
cut that down.  However the first pass at this -- with an 0.1 mfd metalized
paper cap -- more than doubled the warm-up drift.   Replacing that with
several .01 ceramic disks in parallel got it back to the 200-300 cps range.
You can buy non-inductive capacitors and that might be a better way to go.

It STILL wasn't a T-9 note.  Sudden flash of inspiration:  Perhaps the
bypass from the cold end of the feedback winding should go direct to the
cathode, thus keeping H-K leakage completely out of this circuit?  (Rather
than going to the chassis at both ends ...)   Another step better, but still
not perfect.

So, I spent three hours rewiring the 6BH6 LO filament from about 1/3 of the
way up the series string to the cold end.   (Undo about 3/4 of the cabling,
new wires go under a bunch of stuff ...) THAT got rid of the last trace of
warble.   T9 CW and voice quality is now fine -- as good as any receiver
here.  

Filed for the future: Local oscillators belong at the cold end of a series
string, regardless of what the commercial boys did with the TVs and FM sets.

Entering the home stretch, I think.  Tomorrow I will check the BFO for drift
-- I know there is some -- and deal with that.   Pretty soon I need to wind
a whole flock of coils, but I want to get the boxes (coils and cabinet) made
while we have nice weather so maybe that should come first.

Walt
KJ4KV