[HBR] The long, SLOW HBR project

[Walt Hutchens]

In the previous episode I realized that the much too low sensitivity on 20M
was due to weak LO injection and that this was due to the unavoidably lower
'Q' of the 20M tank circuit.   The only solution was to raise the feedback
on this and higher bands.

(Q must go down on higher bands of a ham band only receiver because with a
roughly constant bandwidth -- ~500 kcs -- and using the same tuning
capacitor, band changing is primarily by changing the inductance.   With XL
going down faster than circuit resistance, XL/R declines.)

Adjusting feedback on a per-band basis could only be done with a circuit
that separated the feedback level from the tuning arrangement: Although any
circuit can be adapted, the Colpitts winds up using too many capacitors and
most of them are inside the coil.  As a practical matter the choices are
either a Hartley circuit (feedback controlled by the location of the cathode
tap) or a tickler feedback circuit (feedback controlled by tickler winding
size and coupling).

Either circuit would have to be mapped into my specific situation:
Oscillator RF grounding is to the chassis but DC grounds return to line
neutral and that introduces a few kinks.   Additionally, it's best to avoid
RF on the cathode of a line-operated tube.   And finally, my strong desire
to track the oscillator to the existing dial calibration required a three
point fit which translates to a padder cap in series with the tuning cap.

The tickler feedback circuit proved the more adaptable.   The three point
fit (to the dial calibration) comes from:

1. The usual APC trimmer in parallel with the tuning capacitor and located
in the plug in coil; this most affects the high end of the tuning range.

2. A padder capacitor in series with the 'cold' end of the coil and thus in
series with the tuning capacitor; this most affects the LOW end of the
range.  And,

3. The inductance of the coil -- number of turns and spacing -- which
affects the entire range.

Because my dial and capacitor came from the same receiver the three point
fit would not require extreme values but finding the right combination is a
challenge when you start with guesswork for all three.   The guesswork isn't
uninformed: I knew that the series padder cap implied more turns on the coil
than in the HBR designs and from previous circuits I had a rough idea of
both the padder cap size and the number of turns.   But when turned on for
the first time neither of the two coils was within a megacycle of the right
frequency or a factor of two for tuning range.

I installed an APC 75 mmf series padder in the receiver chassis and brought
out the 'cold' end of the main coil to a separate pin.   I'm thus able to
adjust the padder with a screwdriver during the initial tune up.   Once
everything works, I select a fixed padder, install it in the coil form, and
clip the connection to the base pin so that coil can be used while the next
band is being tested with different settings of the chassis padder.

Ultimately this padder will be unneeded but I'll probably leave it in place
against the possibility of winding a coil for some other frequency range in
the future -- maybe 30M?

The tracking process works like this:

1.  Adjust everything including the number of turns if needed so the tuning
range is roughly centered on the proper band.

2. Alternately adjust the padder at the low end of the band and the trimmer
at the high end until these points correspond to the end dial markings.

3. Check the fit to the dial at 100 kcs points through the band.   There
will be either a sag -- oscillator frequency too low across the middle -- or
the reverse. 

For a sag, you RAISE the whole band by reducing the number of turns, then
readjust the ends to lower them by increasing both trimmer and padder.   For
the first lowering I do roughly half with each, then adjust alternately as
above.   After the first couple of go-rounds it speeds things considerably
if you overshoot each adjustment.   (The narrower the band, the more
overshoot is helpful.)

Repeat step 3 until satisfied with the fit throughout the band.   Since
these dials can't be casually read more closely than about 1 kcs,  a sub-kc
error throughout the band is okay.

(My dial is definitely a one-off.   It appears to have been factory
calibrated at the 100 kcs points because those are accurate.   Subsequently,
someone added 10 kcs points using India ink and a ruling pen.   This is a
nice clear job but the points are a bit scattered -- like a few 2 kcs
errors? -- and the variation of line width gives away the use of a not too
carefully adjusted ruling pen.   It is a thing of beauty nonetheless ...)

When calibration is okay from end to end, then I estimate the proper fixed
padder capacitor based on the setting of the variable and tack that in in
place of the variable.   These estimates are inexact and I always have to
change the fixed cap at least once but with consideration of the sign of any
remaining error, the nearest standard size is good enough.   Finally, with
the cap installed in the coil and the base connection clipped I tweak the
coil inductance to get the center frequency right again -- that will be a
small change of spacing for an end turn -- and then reset the ends in the
usual manner.   

The new 80M coil started at 32 turns and had a 4 kcs sag in the middle.
Removing two turns reduced the sag to 2 kcs and removing two more brought it
to the spec throughout the band.

The process took about six hours,  including making and winding the coil.
The three point fit process, however, takes most of the time and is entirely
a self-inflicted wound -- the result of wanting to track the receiver to an
existing dial calibration.   With a blank dial to be calibrated, a two point
fit (to pre-marked end points) would be fine and any of the usually
recommended tuning caps will give satisfactory straight line frequency
calibration.

The tickler specs for the HBR series worked fine on 80M.   For 20M, I had
severe 'squegging' -- a blocking oscillation caused by excessive feedback.
This was likely due to the fact that I'm running the oscillator at higher
voltage than in the original designs AND wound the 20M coil for fundamental
rather than half-frequency operation as in the original.   When I reduced
the number of tickler turns and increased the spacing, the problem went
away.   

The sensitivity on 20M is dramatically improved: As on 80M very strong
signals can be copied with no antenna and the 4th harmonic of the 3502 kcs
band edge marker (still not properly adjusted!) gives 10% meter deflection.
I will make measurements.   Both mixer gain and oscillator output probably
can be increased if necessary, ditto the calibrator output, and there are
spare pins on all the coils that could be used to make adjustments per band
if needed, though only at the price of remaking those coils.

Unfortunately just about the time I got the 20M oscillator coil roughed out,
I discovered I had some pickup of the BFO in the IF stages.  This had been
hidden by incorrect adjustment of the AGC back apron control; it should be
zeroed with the BFO off, then the BFO turned ON and tuned through the range.
Any S-meter movement indicates leakage into the IFs which must be corrected.

This appears to be an easy one: It's going into the crystal filter area, as
indicated by the fact that tuning the BFO displays the crystal filter shape
and subsequent IFT adjustments do peak the signal.   About half of it went
away with improved bypassing on the 1st IF cathode (suggesting that there
may be a bit of BFO on the filaments); the other half will most likely be
eliminated by a tinplate shield over the 1st IFT and crystal filter wiring.
Unfortunately, this requires moving several other parts.

The previous version of this design needed a shield over this area for
exactly this reason; I had hoped that the major chassis reorganization would
eliminate that but should nonetheless have allowed for easy installation.
Such shielding is good practice anyhow: There is a LOT of gain from the 1st
IF grid onward, 1700 kcs isn't the quietest frequency, and in a receiver
that will most likely have a wooden cabinet, noise and spurious signal
pickup is possible.

The mixer tube might also benefit from shielding.  Although pickup here may
not be a problem due to lower Z (plates of triode mixer) and the attenuation
of the crystal filter that immediately follows, I will do more checks once
the known problem is eliminated.

Another very odd problem may be in the Heath IM-2410 frequency counter I
use.  Turning on a particular CFL light in the room (one with a solid state
switch) causes the counter reading to DECREASE by several kcs.   This isn't
an effect on the radio because it isn't detuned; I'm assuming it's trash
getting into the counter via the power line and causing desensitization.
I'll see about a line filter, maybe at the lamp end first.

I hope to get back to the coils by the weekend, finish the 20M coil, measure
sensitivity and drive level, and so on.   One great advantage of the
push-push dual triode mixer is that the LO signal appears on the cathode in
a well-buffered form so you can connect a scope or frequency counter without
disturbance.  

I've thought some about how this receiver would be to duplicate -- assuming,
of course, that it is one day finished.  That picture seems reasonable.

Building today, the parts problems would be simpler than for the original
designs.   I think just about everything critical could be extracted from an
HR-10 junker.   It might even be possible to put a new (perhaps just
partial) top plate on the chassis and use that which would bring along what
looks like a perfectly usable dial.

An HR-10 adaptation could just stay with the power transformer supply,
rather than going series string/line HV.  That would provide some of the
tubes as well.   

There are some uncommon circuits but the only one of them that's the least
bit tricky is the AGC detector and the AGC itself.   This requires a back
apron adjustment and the setting interacts with the S-meter zero.  However
the procedure can be done step-by-step.

This circuit is more subject to problems with small amounts of tube gas than
the usual diode AGC.   However once you have a set of tubes that work,
you're good to go: I haven't had one develop gas while in use, yet.

The rare 19J6 tube is gone now; all tubes are generally available and mostly
cheap because nobody's doing much with 150 mA series string radios if you do
go that way.  

Walt
KJ4KV