[HBR] Re: HBR -- Part 6

[Walt Hutchens]

The local oscillator is wired and the 80M (3500-4000 kcs) band
oscillator coil wound and tracking the dial. This was a BUNCH of work,
but seems to be successful.

The oscillator is a Vackar, although not a great one, due mainly to
the small value of the tuning cap. The really good Vackars for the 5
Mcs range swamp both grid and plate with 100s-1000s of mmf caps; this
one has 100 mmf on the grid and 126 mmf on the screen grid (=plate)

But standard Vackar circuits for this frequency range use tuning caps
in the 50 mmf range; mine is a 15 mmf delta-C unit as commonly chosen
for the W6TC and similar ham-band-only designs. Using such a small cap
and with the lower gain of an ECO, you can't use an extremely high-C
circuit.

The coil goes from the screen grid to a capacitive voltage divider.
The tap of the divider goes to the control grid. A fixed cap ('padder
cap') goes from screen to ground. The tuning cap goes from the control
grid end of the coil to ground and is paralleled by an air variable
glued in the top of the coil ('trimmer cap') and a very small air
variable on the panel, for calibration trimming purposes.

In traditional ham work a two point fit to the dial calibration is
generally used.  A trimmer cap that's across the tuning cap is used to
set the top end of the band to something like 95 (on an 0-100 dial
scale) and the coil is adjusted to put the bottom end at 5 or so on
the scale.  Then a stable frequency source or frequency counter is
used to mark the dial for in-between frequencies.

However my Eddystone 898 is already calibrated for the G3RKK-design
receiver. It's a nice, professional calibration job for six bands,
160-10 meters, and nearly perfectly linear. I'm using the tuning cap
that came with the dial in this set. Rather than recalibrate the thing
(when I could not do nearly so nice a job) the plan is to fit the
receiver tuning to the scales that are already there. Getting an
accurate match requires a three point fit.

The typical three point scheme uses the fact that changing the coil
inductance moves all frequencies by the same percentage, the
(parallel) trimmer mainly moves the top end of the dial, and the
(series) padder mainly moves the bottom end. While you can calculate
the values for a scale for which you have precise percentage numbers
-- there are programs on the web to do it for a linear scale and cap
-- fitting to a not-perfectly-linear scale on an existing dial with a
cap that isn't straight-line-capacitance is a trial-and-error process.

The best approach depends on how the various adjustments are done. In
this case I wanted to select a fixed value of padder cap, making that
the hardest variable to change. So I started with a wild guess about
the padder (several times the max value of the tuning cap plus trimmer
is a good start) and used the trimmer and inductance value (removing
and/or spacing turns on the coil) to get the end points right.

With the ends right, I checked the mid point. With the first try, the
mid point was 8 kcs low in frequency: when the dial was set for an
oscillator frequency of 5450 kcs (3750 + 1700 kcs IF) the actual
frequency was 5442 kcs. Over several hours of work I increased the
padder cap (mainly lowering the low end frequency), corrected the low
end by decreasing the inductance, and brought the high end back down
by increasing the trimmer value. Because the tuning range is small
(500 kcs at over 5 Mcs -- less than 10%) the adjustments interact a
lot, so you have to overshoot each adjustment at first and go back and
forth many times.

Reducing the coil from the initial 30 turns to 26 and increasing the
padder from 100 to 126 mmf (measured value of cap marked 120 mmf) and the
trimmer from 2 mmf to about 10 mmf did the job.

The W6TC coil designs are beautiful to look at and very practical for
use but they're impossible to work with for a cut and try design. I
did the initial work with the chassis upside down and the coil form
(no base plug) and other parts tacked to the tube and coil sockets.
Only when I had the padder (fixed) value right did I assemble the
padder, the air trimmer, and the coil form onto the tube base. The
APC-25 trimmer is glued in the top of the coil with PVC pipe cement;
the coil is epoxied to the salvaged base from a metal octal tube.

A two point fit will do for the final tweak (when everything is wired
up) bringing the mid point error to less than 1 kcs. That is about the
width of the lines on the frequency dial and rather less than the
draftsman's errors in placing a couple of them.

Yeah, making the coil fit the dial calibration is kind of a ship in
the bottle exercise. And I get to do it again, for five more bands.
When I get all done, I'll be able to say whoopee, I got the ship in
the bottle.

The good news is that the results from 80 will be some help with (can
be roughly scaled to) the other bands and since 80 is the widest band
(percentagewise) the fit won't be as critical, anyway.

As assembled it drifted down about three kcs during well over an hour
of warm-up. Fairly standard for a 5.2-5.7 Mcs oscillator and this type
of construction. I replaced the Western Electric socket (with those
oh-so-convenient extra terminals) with a socket mounted on spacers
about 3/16" above the chassis. The socket pins stick down just far
enough that you can wire them in the usual way and there's ample space
for air to flow up around the pins, taking heat from them and the
socket along.

With this arrangement, not only is less heat transferred to the
chassis, but in about 3/4 hour, the drift nearly stops, showing that
the tube temperature has stabilized. This result is logical, because
the hotter the tube (base, socket, pins, wiring ...) the greater the
air flow. This process reaches equilibrium much faster (at a lower
temp) than the one in which heat is dissipated only from the tube
envelope and by conduction into the chassis and wiring.

LO (and VFO) tubes generally should NOT be shielded: not only do most
tube shields make the tube run hotter, but more of the heat will be
conducted back into the chassis.  When a tube will not be shielded, it
should NOT have a shield base socket, since that alone significantly
increases the temperature and heat transfer to the chassis.

With the drift this low and so well-behaved, the right temperature
compensating caps on the tube socket and either the coil socket or
tuning cap should easily get it to the 100 cps/hour range on the lower
bands.

There's no need for a three point fit for the RF and antenna coils.
With the maximum coil Q being around 150, being within a few kcs
across the band is plenty good enough. A two point fit at (say) 10%
and 90% of the dial will do.

I'll post new pictures in a day or so.

Plate voltage is often fed to Vackar oscillators through a choke or
resistor to the plate. That's an error: the resistor loads the coil
(there's RF voltage there, right?) and the choke is another reactance
in the oscillator circuit, hence another source of instability.
Feeding through a resistor at a tap near the zero RF voltage point
eliminates both issues, giving the highest possible coil Q, hence best
possible rejection of frequency change due to voltage fluctuations and
tube characteristic changes. There is no bypass cap at this point --
it floats.

The tap is twisted and soldered into the length of wire at the right
point, then the two ends are fed through two holes from the inside of
the coil so the loop is inside. The coil is wound both ways from the
tap. The resistor is thus part of the coil assembly.

I'll probably do the 2nd IFT and 1st IF stage next.  Ugh ... another
100 turns of #40 wire ...

Walt
KJ4KV