[HBR] Another Receiver Project -- HBR-4, Part 17

[waltah at earthlink.net]

"Good grief, Charlie Brown" edition ...

So after a few months of pretty hard work, I measured the 
interference-free dynamic range.  That's the range from the weakest 
signal you can copy to the strength of two strong equally spaced 
signals off to one side required to produce a same-frequency signal 
of strength equal to the one you're copying.   On the lower bands --
80 and 40 at least -- IFDR is often the determining factor in what 
you can copy and even more frequently determines when signals 
sound 'clean' and when they don't.   

 It was a rather disappointing 75 db.  Well above the early solid-
state rigs, on a par with the better Collins ham sets, not too close 
to the 90 db range of the R-390 and certainly not close to what I 
expected with a beam tube mixer and no RF stage.

I took stock.   The mixer is a 6JH8 operated at 140 volts on the 
plate; nearly all the published designs used 7360's.   Maybe that's 
a better tube, but the 6JH8 and 6ME8 came later; one wouldn't 
expect them to be markedly less good.   

The manufacturer's data sheets 'typical operation' info is 150 plate 
volts for the 7360 and 250 V for the 6JH8 and 6ME8 but I have seen 
comments that the 7360 requires 300 volts or so -- that's the 
manufacturers absolute maximum -- when used as a receiver 
mixer.  No doubt higher plate voltage would improve the linearity.  

I looked again for wiring errors and reconsidered test setup 
problems.   Didn't find any.  

(You can sniff out test setup problems by doing the testing at 
various signal levels.   Since this distortion follows a known law, 
IFDR can be calculated over a considerable range of input test 
signals and the calculated results should be in general agreement --
 say +/- a couple db, at worst.  If the receiver tests 'better' at low 
signal levels one suspects test signal leakage into the input 
circuits leading to a seemingly better figure for sensitivity or (less 
likely) cancellation of a signal produced by receiver non-linearity by 
a test setup problem; if 'worse' at low levels then the oscillators or 
hybrid combiner are at fault -- the oscillators may have excessive 
2nd harmonic distortion, they may crossmodulate each other, or 
the hybrid combiner may be overloaded.

I considered non-linearity of the coils in the receiver input circuits.   
I can't rule it out completely, but it doesn't seem too likely, either.   
The only practical way to reduce that while keeping this 
configuration is removing a turn from the antenna winding on the 
input coil to reduce the field strength in the coil.  But those cores 
are *large* (cup-and-core tunable part in series with a T50 toroid on 
40 and below); saturation at levels of ~ -13 dbm and less seems 
out of the question.   

A partial check on the location of the problem was to reduce the 
mixer cathode resistor.   I got more noise and less gain -- but the 
IFDR seemed to go up a couple db.  That seemed to finger the 
mixer.   

Fork in the road time.   75 db isn't that bad an IFDR and it can 
almost certainly be improved by raising the plate voltage to the 
maximum -- 250V for the 6JH8.   Increasing the plate voltage would 
require another transformer but it's certainly possible to do it.  But 
these tubes were supposed to be not just 'good', but outstanding in 
the linearity department.   Heck the R-390's get 90 db with 6C4 
mixers.   Why the hoop-la over beam tubes if it's a struggle to 
equal a 6C4?   Why not just duplicate the R-390 circuit with a far, 
far cheaper tube?   Might a 12AU7 with the halves in parallel be 
even better?   If we're doing that, what about a 5814?

Haven't I seen speculation about the properties of an 833A as a 
receiver RF stage?   

Such ideas certainly might be tried, but I still wanted to know why 
the big deal about beam tubes.   The only tube anybody used for 
the receiver mixer job was the 7360 and as it happened that was 
the only tube for which I had never printed a manufacturer's data 
sheet.   So I printed it.   Then I refilled the printer cartridge with ink 
and printed it again.

I think that was a light that just went on above my head.   RCA's 
data sheet shows two circuits using the 7360; one is a balanced 
modulator, the other is a high-level SSB mixer as would be used in 
a transmitter.   

Neither one of the two circuits was designed to give a rip about the 
linearity of reproduction of the signal on the tube's G1.   The 
balanced modulator used it for the carrier input, and the mixer, for 
conversion oscillator injection.   

The sign illuminated by that light bulb up there says "seen from 
G1, beam deflection tubes are just *pentodes*; there's no reason to 
expect remarkable linearity there.  What's highly linear is the path 
via the *deflection plate* input."

(From the curves in the tube data sheets the 7360 shows pretty 
good linearity over about 10V p-p on the deflection plates, while the 
6JH8 looks good to around 40V.)

In a receiver, acheivement of the remarkable performance of which 
these tubes are said to be capable seems to require driving one or 
both deflection grids with the antenna signal and using G1 for the 
local oscillator.   Since the single ended G1 signal is strongly 
rejected at the push-pull output that should have the added 
advantage of taking out most spurious responses, probably at the 
price of requiring a 9 Mcs trap in the antenna circuit.

There could even be instability problems, though it seems unlikely 
with the deflection plates having relatively low sensitivity and 
always tuned fairly well away from 9 Mcs.

I don't know about the deflection plate gain.   With input signals of 
10V on G1 and 8V on the deflection plates, RCA got 40V p-p on 
the ouput plates.  That's a conversion gain of >12db, but I probably 
can't put a 10V single-ended signal on G1 and the deflection plates 
draw a little bit of current -- maybe 10 uA for the 6JH8.   

However, unless the noise level is dramatically higher (why?) the 
sensitivity problem seemingly could be easily made up with an RF 
stage.  If the mixer input can truly reach levels of ~10V or more, 
there's room for lots of gain before there's a mixer overload issue.   
The 6EH7 is a great RF tube at these frequencies and running at 
very low gain should give an outstanding performance.

My first reaction was to cringe at the thought of rearranging the 
connections to that stage to reverse G1 and the deflection plates.   
Maybe a different tube would be better?  But it turns out that 
simply turning the 6JH8 socket 180 degrees will make it close 
enough.  The single ended premixer drive can be obtained with a 
link on the premixer plate coil.   The change shouldn't take over a 
day.   Well, that's not counting the (almost certainly needed) 
installation of an RF stage.   But at least I already have the socket 
hole center punched.

The remaining question is, if this speculation turns out to be right, 
why did everyone else get it wrong?   Such things happen -- there 
were whole decades in which we all told each other that good 
VFOs were wired with #8 wire.   But the reason for that error was 
simple -- we were looking at the excellent VFOs of WW-II which 
were expected to keep working in bombers with an engine shot 
away or ships providing shore fire at Omaha beach.  Taking the 
best VFOs of that period and simply rewiring them without all those 
internal struts (#8 wire is a pretty strong structural member) and 
heat pipes (#8 is pretty good at moving heat, too) would 
dramatically improve their performance under the more favorable 
conditions of our hamshack desks.

The gain's almost certainly higher with the input on G1 and when 
the HB-65, HB-67, Junior Miser's Dream, and the W1OMX designs 
were developed, nobody that I know of in the ham world was doing 
intermod testing, though obviously engineers like Ernie Pappenfus 
did it.   There were a couple of other QST articles in the 50s-60s 
timeframe by people who were aware of the issue, I believe -- By 
Goodman, right?   But weren't the designers of the SS-1R in the 
same league?  I don't know.

With a new tube type, there were no prexisting designs to follow.  
Once someone stuck the antenna signal in the logical place and 
got good results, why wouldn't others do the same?   All those 
designs were built in the last five years (1965-70, roughly) of the 
vacuum tube's ham radio lifespan; there wasn't the time for the 
hashing things over that occured with issues like VFO design, let 
alone linear amps.  

I'll know more in a couple of days.

Does anyone know of any receiver design using a 7360 (or other 
beam deflection tube) mixer with the input signal on the deflection 
plate(s)?

Walt 
KJ4KV