[R-390] The other favorite signal generator

[David Wise]

Some of you like the HP 8640B signal
generator.  Here's my story.  But since
some may not read to the end :-) I'll
ask my question first.  Does anybody
have a manual newer than the
one on logsa?  Logsa's manual for the
8460B Option 001 has change notes covering
S/Ns up to 15-something.  They also have
a manual for 8460B Option 004, which
goes up to 16-something.  I wonder maybe
the two variants shared the same
serial number stream, but in any case,
my instrument is 17-something and I've
already found discrepancies.

On to the report.

I looked for 606Bs on *that place* for
a few weeks and never saw any.  Eventually
I cast my net wider and immediately found
an 8640B, "as-is".  Looked slightly rough.
I got it for $215 + $25 shipping and insurance.
Options 001 (variable frequency modulation
oscillator) and 003 (reverse power protection).
As advertised, it lit up, but it was
not usable.  The main tuning knob
was spinning loose on the shaft, a button
and a knob were missing, and another was
a generic replacement.  Every skirt was
loose, an important thing because they
all have numbers!

PROBLEM 1: GUM

When I tightened the tuning setscrews
(get spare .050 screws and wrenches),
I found the real problem: the shaft was
frozen.  The oil had gone sticky.
The guy had worked the knob until
the setscrews scored a groove.  Lucky he
didn't break off the crank.  I filed off
the burrs and added a flat for each
setscrew.  I removed the oscillator,
worked off the frozen-solid stack of
turn-stop washers, and, ignoring the "Factory
Repair Only" tag, put it in a vise and
applied a spanner and almost my entire
supply of elbow grease to the large locking
ring.  Just before the pain in my fingers
would have made me desist (wear gloves!),
it gave.  I unscrewed it, and the
oscillator fell apart in two halves.
One half carries a shaft which is threaded
similarly to (obR390) our favorite PTO.
It slides an aluminum slug that occupies the
entire housing, with just a one-inch hole down
the center.  The other half is a gold-plated
snout that pokes into this hole.
I cleaned out the bearing with brake cleaner.
The shaft and the plastic 8-turn stop too.
I washed the stop washers in soap and water
afterward so the cleaner wouldn't melt them
or something.
The guy had damaged the stop, but it wasn't
fatal: one tab had rounded-off corners.
There was enough material left that I could
trim it back square with an X-Acto knife.
This increased the stop range, but it's
adjustable.

Got it back together, lubricated the shaft
with one small drop of Mobil 1 on a
toothpick.  I did not lubricate the stop
at all.  Ahh, smooth tuning.

But when I went from the 8MHz band
to 16, the display went from 8 to 2.
Someone recently posted this same
problem.  Here's the diagnosis:

PROBLEM 2: BANDSWITCH

The bandswitch proper was ok; that's
a bunch of cams inside a casting that operate
lever switches in the Divider/Filter module.
The 8640B's oscillator runs in the VHF, from
256MHz to 512MHz.  This gets digitally divided
down and filtered back to a sine wave.
The cam-operated switches set the divider ratio
for the output but the counter hears the
oscillator directly, and a different switch
tells it what to divide it by before displaying.
This is a planar rotary PCB switch, where
the stator contacts are pads on a circuit
board and the moving contacts are tiny
bifurcated fingers projecting from a
plastic disk.  There are four in the 8640B,
three on the Bandswitch/Deviation shafts and
one on the attenuator.  The fingers came off.
There are also stress cracks all around the shaft
holes.  Most gears are cracked as if they
shrank.  One gear was bad enough to stop the
show.  The Bandswitch/Deviation gear assembly can
be removed without opening up the bandswitch
casting.  That shaft is driven off the front
shaft via an oldham coupler (more R-390).
After removing the problem gear (in a
differential -- still more R-390!),
I epoxied a washer to the back after forcing
out the brass insert, ground down the insert
slightly, and epoxied it back in.
JB-Weld is thick enough that you can just
plunk the fingers into it and they won't
drift around while it sets.  The main
trick is getting them located right, since
JB-Weld is opaque.  I used The Force.
Got it together next day and it worked.
I played with other controls, and noticed
that if I turned on FM the frequency shifted.

PROBLEM 3: FM

Using a manual I tracked this down to
the FM deviation amplifier, which drives
a varactor in the oscillator.  It was pinned.
It turned out to be an open resistor with
no evidence of violence.  Almost all the
resistors around here are 1%, but I cobbled
up a replacement out of the 1% junk box.

PROBLEM 4: KNOBS

On the existing knobs, I just glued the
skirts back on with contact cement.
The FM Deviation knob was missing.
Thank goodness the skirt was still
dangling there.  The knobs at Radio Shack
are too small or too large.
But someone had previously lost the Band
knob and replaced it with one just right
for Deviation.  Since it was already loose
from its HP skirt, I moved it to Deviation
and used a 1-inch RS knob for Band.
Since Deviation has a concentric vernier,
I had to drill the ex-Band knob.  I also
had to cut down both foreign knobs with my
Dremel to make room for the skirts.
Works great, but they don't match their
neighbors.  OTOH I don't see how anyone
could have turned the originals.  I
removed two of the four (!) detent leaf
springs and they're still kind of tight.

That's it for now.  I know there's other
stuff lurking, including some things that
I threw off by disassembling the oscillator.

The 8640B is advertised to cover 500kHz to
512MHz, but the spec includes enough overlap
to do at least 450kHz to 550MHz.
I've seen posts where the unit went down
to 440-something and faded out.  Mine keeps
going right to the absolute mechanical limit
of the oscillator, with the 8-turn stop
removed.  At that point it's around 427.
With the stop, it does 445.  Just
enough for my R-390A...

I'm a happy camper,
Dave Wise