[HBR] HBR -- Part 9

[Walt Hutchens]

Well ... THAT was exciting.

The mixer worked straight off, and the half lattice crystal filter is
fine.  Two slight humps, adjusting the 1st IFT (which feeds the
half lattice) lets you balance them, and if I want to flatten it a bit
I can increase the load resistor somewhat.  I think it's 33k now.

No problems wiring the RF stage or winding the mixer coil.

Then I wound the Antenna coil ... hummm ... no dip. After a few trials
of this and that, it developed that the 1st gang of the tuning cap was
shorted. This was a bit hard to believe because it's somewhere between
double and triple spaced and the stators are VERY solidly mounted.
However after a few minutes study I noticed that the rearmost of the
six hairpin-type contacts that pinch the shaft to ground the rotor was
touching the stator. Furthermore, I don't see how it could ever have
worked on this cap OR been reliable on any of those caps: There simply
isn't room enough between the stator and the capacitor end plate.

Six contacts is more than enough: I unsoldered and removed the
offending one. The antenna coil then dipped correctly. Yep, the
MG/Jaguar experience -- with a radio.

So with everything dipped, the RF stage oscillated.  MAN, did it
oscillate.  I always hate that problem because you (or at least, I)
can't just work X hours and fix it: It requires study and flashes of
insight.

It was obvious at the start that this wasn't a minor issue: the RF
stage was producing VOLTS of RF.

The first thing to look for is places where the grid circuit can 'see'
the plate. I did find one that was obvious: the tuning cap stator for
the grid (antenna) circuit was in view of the plate (mixer) coil. A
tinplate shield over that gang of the tuning cap ended that. The
oscillation got better, but was still strong.

(Having the S-meter working is a big help with this testing because it
tells you which way you're going with the strength of the
oscillation.)

The plate and cathode leads between the two halves of the two-triode
mixer (which are the triode halves of the RF amp and oscillator
19JN8's) ran through the RF amp stage: There is certainly some RF at
the plate of the mixer, so I ran these leads with RG-174/U cable. The
change wasn't significant.

A day or two passed, looking at the circuit and parts layout once in a
while, and chewing on the situation. Then another Ah HA! moment: For
physical convenience I had mounted shared cathode resistor for the two
triodes of the mixer on the RF stage socket, returning it to the
common bus lead of that socket. But that caused the RF stage ground
(common) currents to flow in the same 4" of wire that carried the
current from the mixer which is basically a cathode follower for the
RF stage output.

Gee, that seemed like a bad idea. I moved the resistor to the
oscillator tube socket and the RF stage oscillation nearly
disappeared.

Writing this, I realize that I need to give that resistor its own
ground bus lead back to the central tie point.  Returning a cathode
follower ground lead in common with an oscillator ground lead means
that the CF is part of the oscillator circuit and changes there will
affect the osc. frequency.  That's probably not an issue on 80M, but
on 10M it almost certainly would be.

With a separate common line, a separate resistor/capacitor filter can
be used. Only the mixer bias will be changed and that circuit is quite
tolerant.

This illustrates the main issue in designing 'transformerless' HF
radios: Since the common line returns to the AC line directly, it must
be isolated from the chassis. But avoiding feedback paths when doing
this can be challenging, and the problems are not always obvious.
Fixing the cathode resistor problem involved no change in the DC
circuit: It was just moving a resistor that went between one end of
two wires to the other end of the same two wires -- but the difference
in results is dramatic.

The same issue, of course, comes up in every set in the B+ circuit.
But resistances there are much higher, so filtering is simpler.
Cathode resistors for high-gain RF and IF amplifiers are often less
than 100 ohms and the reactance of a .01 mfd cap at 80M is ~5 ohms so
you have to use unusually large bypasses to get adequate attenuation.

Additional complication: Modern ground fault current interrupter
outlet circuits don't like to see ANY current on the neutral line. But
if you bypass the neutral line (that's B- in transformerless sets) to
the grounded chassis (always a 3-wire cord for these sets, with green
directly to the chassis) then those capacitors will charge when the
set is turned on and the GFCI will trip, at least sometimes. So it's
necessary to keep the total bypass capacitance down: you cannot just
fling a handful of ceramic disk caps at one of these sets.

The way that has worked best for me is to put the B supply filtering
near the center of the chassis and run a cap of 0.47 mfd or so from B-
to chassis ground right from the last filter cap negative. Then common
(B-) leads go in star fashion to each stage from that point.

Each stage that has an RF signal has its B- bypassed at the stage:
for RF and IF stages the cathode resistor and cathode bypass cap does
this job.  The remarks above about the probable inadequacy of 0.01
bypasses and the need to minimize total neutral-to-ground capacitance,
apply.

Another consideration is that you can't do an unbypassed cathode in
order to get degeneration unless you're willing to add a separate R-C
decoupling section after the cathode resistor. But that will add to
the the cathode bias and reduce the AGC action.  A choke MIGHT be
used.

>From the handful of these sets I've built, I'd say that the problems
can always be solved, but there's a lot more head scratching,
compromise, and partial redesign than I really like. It is getting
easier: On this set the RF stage is the only place I've had a problem.

Then I noticed that the grid pin on the socket isn't all THAT far from
the plate pin and plate wiring.  I made a 'U' of tinplate and soldered
it on a ground lug so that it shielded the grid connection from
everything else.  That DEFINITELY helped.

Adding an 0.1 mfd cap in parallel with the 0.01 cathode bypass on the
stage helped a bit more.

At this point it was stable without a load on the antenna connection.
So I lowered the cathode resistor from the 270 ohms I'd been using, to
150 ohms, bringing the oscillation back.

Then I realized that the layout caused the ground currents in the two
tuned circuits to go through the same section of chassis: that too is
coupling. So I ran separate leads from the hairpin contacts on the RF
stage grid and plate sections of the tuning cap to the ground end of
the coils, rather than just grounding the coil end. Each of those
helped a lot; it was now barely unstable even with no antenna load.

There are more things to be tried: The chassis ground of the bypass of
the common negative point probably should be closer to the RF stage,
the mixer cathode resistor should be given a separate common lead, I
should again test shielding of the oscillator/mixer tube (which is
near the RF stage grid coil) and by the time I've done those things,
I'll probably see others.

I will try a choke in the cathode lead with no bypass at the cathode
itself.  If that works (and doesn't give trouble with the parasitic
diode) then it will improve RF stage linearity.

HF receivers don't need much RF stage gain unless the mixer noise is
very high -- say a 6BE6 or something like that.

There's considerable hum on signals: that's from the audio detector
cathode parasitic diode problem noted a few days ago which I haven't
fixed yet. And there may be some noise from a heater-cathode partial
short in the RF or mixer stage: there's been enough other stuff going on
that mere noise wasn't worth trying to pin down.

There's quite a bit to do, but I think it'll all shake out. And the
BFO, band edge marker, and standby circuits have yet to be wired.

Oh ... I don't have an HF antenna since the last ice storm last
winter, so I need to spend a day re-stringing my 160 meter full-wave
loop.

Walt
KJ4KV