[HBR] The long, SLOW HBR project

[Walt Hutchens]

Don said:

> I want to thank you for your posts concerning your receiver. Your
> progress has been an education, and I look forward to hearing more!

Thanks for the encouragement,  to both you and other list members who have
written privately.  I sometimes feel I take up too much bandwidth but the
list is mostly quiet and I am now summarizing completed steps which requires
fewer words than the blow-by-blow.

> I am in a similar situation with my receiver, it seems that unexpected stuff,
> always rises up to add to the learning curve.

If you find a way to avoid this, PLEASE post it,  hi!

The goal this time was to raise the mixer gain by better matching to the
crystal filter.  However, sometimes you have to do last night's unfinished
homework before you can start on today's assignment ...

I had noticed that the S-meter was off zero ... trivial fix, since there's a
control for that.  Oops ... out of range.   A few minutes changing resistors
to get that right.

Then noted that the S-meter sometimes doesn't return to zero but hangs at
about 3% of F.S.  This is a mechanical  problem in the meter (a 1"
International 100 uA unit) so another half hour went to checking stock for a
replacement meter:  Nothing found.   I will live with this minor issue.

Then I thought there was a really easy way to get more of the gain the RF
and IF stages are capable of by returning the AGC line to a tube cathode
rather than ground.  (At these supply voltages the tubes would still be well
within limits at zero bias.)   So I checked the actual AGC line zero
voltage: The maximum (with the AGC detector triode cut off) was about -0.5
volt which seemed a more negative than it should be.   I couldn't remember
ever checking for a slightly bad RF or IF tube, so I started substituting
... sure enough, replacing the second IF tube raised the voltage to about
-0.1 volt ... a more reasonable point for these tubes to start to flow a
tiny bit of grid current.

A significant fraction -- 1/4th?  -- of today's NOS pentodes are just gassy
enough that they won't work in critical circuits.   Because this (plate
detector) AGC circuit can't deliver as much current as the usual diode, it
is more sensitive to this issue.

(Part of the design of this receiver is the use of sharp cutoff pentodes in
place of the usual remote or semi-remote tubes.  This choice was desirable
because there are no semi- or remote cutoff  9-pin miniature 150 mA
triode-pentode tubes that I'm aware of, but it means that IF signals should
be kept smaller than usual.   Cutoff voltage for the 6BJ6, -20, for the
19JN8 pentode, -8. 

(This in turn means that diode detectors aren't a good choice -- they become
seriously non-linear with small signals.   Plate detectors work much better
-- with smaller signals linearity is good -- but reasonable AGC time
constants for SSB and CW require higher AGC load resistors ... and there we
arrive at the sensitivity to even slightly excessive grid current.

(This set has about the function of the HBR 13 (and higher) sets with only 9
sockets ... use of two-section tubes for everything but the LO  was part of
how it got there.) 

THEN I was able to change the AGC return from ground to the cathode of the
1st IF.   This raised the RF/IF stage cathode currents from ~ 7 mA to ~10
mA; the 'typical operation' value is 16 mA so I could still do more by
raising the screen voltage.

(The currents that are being raised by these changes vary dramatically with
AGC action, so increasing them complicates voltage regulation for the LO and
could force regulation for the BFO, as well.  Plus operating somewhat below
the maximum means longer tube life.)

With the meter readjusted once again (but this time the rear-of-chassis
controls did the trick) I checked the gain on 20M; definitely better.
Before getting serious I'd probably better check the alignment, right?
Ooops -- the mixer plate was WAY off ... With the IFT way out of resonance
it's no WONDER the thing was mismatched.

Donno how that happened, either.  I did check the alignment after replacing
the mixer (19J6 -> 19JN8) but obviously I got it wrong.

So that raised the gain even more, as well as improving the shape of the IF
passband considerably.   The gain on 20M is still too low but I was now
ready to tackle the remaining mismatch problem by adding turns to the 1st
IFT primary.   However the higher gain made the audio roughness problem
intolerable, so that needed to be done first!

This fix was actually simple: Changing to a two-time constant AGC scheme
(think of two R-C combinations in series, though that's not exactly the
circuit) with 0.1 and 1 second TC's respectively made a dramatic
improvement.   Basically the (new) 0.1 second TC allows the AGC to respond
quickly enough to prevent overload on voice peaks (and static pulses) while
the 1 second TC keeps background noise down between words.

The key thing here is the attack time constant which is the plate resistance
of the 19JN8 triode (~5000 ohms) in series with a 1 mf cap in the original
but the same in series with 1 mf in series with 0.33 mf or about 0.25 mf in
the revised circuit.   So the attack time has been reduced from about 5 mS
to a bit over 1 mS -- a worthwhile improvement.   Then the two separate
decay times take over -- 0.1 S, then 1.0 S the rest of the way.

You can see this easily on very strong signals: The meter jumps to close to
full scale and bounces between ~1/3 scale and FS during a sentence, then
falls more slowly from 1/3 scale at the end of the sentence.

(These meters are slightly underdamped -- good for observing things like
this, but annoying in general!)

Hang AGC is another way to go and does a better job on a reasonably well
behaved band.   However static crashes and very choppy round tables can
cause you to miss multiple words and these circuits require a couple more
triodes.   The two time constant scheme doesn't suppress the background
quite as well during a transmission but never gets fooled into killing a
short comment from a weaker station right after a strong one or shutting off
a moderate strength station due to frequent static crashes.

The AGC threshold control on this set can be used when all stations being
copied are of good strength.   This sets the level to which the AGC decays:
If everyone is S9+ you might set this control so the S-meter drops only to
where an S8 signal would put it, meaning that noise (or signals) below this
level are much reduced.

I believe Collins pioneered the use of two time constant AGC; I know I've
seen it in one of their fancy military sets.

Adding a 100 mfd bypass cap to the cathode of the AGC detector improved the
performance still further.   Audio quality is now good enough but detector
linearity should be checked further at a later time.

Once I figure out where to mount the new parts permanently,  I think I can
look at the mixer to crystal filter mismatch issue!

(Something about remembering that your original objective was to drain the
swamp ...)

Walt 
KJ4KV