[HBR] HBR2K -- Chapter 14 -- Large Signal Performance, Part 6

[hbr@mailman.qth.net]

(In the previous part I got the IFDR up to about 70 db at the antenna 
and considered the choice between working on the mixers to try to 
carry the 84 db IFDR number at the filter driver back to the antenna 
and going the other way to improve the 84 db number by reducing the 
noise at the 1st IF stage.)

I didn't wanna do it, but in the end the 1st IF stage changes won.   
There was just too much wrong there.  The need for high gain 
seemed to give me inescapable problems with pickup of stray 
signals (3180 kcs IF = 2x 1590 = 3x 1060 ...) and subtle feedback 
issues.   Even if it could be tamed, a 'touchy' stage has no place in a 
design others might try to duplicate.   

I ripped out the tuned circuit and as a first step wound a 4:1 (bifilar) 
transformer to couple the filter to the cathode of the 6EH7 1st fixed IF 
stage.   It worked (and didn't work) in exactly the way you might 
suppose: All the weirdness was gone and such signals as I could 
hear, sounded very clean.   However the gain was way too low and 
the receiver noise was now dominated by  that stage.   The gain 
might be somewhat improved by a better match from filter to cathode 
-- 500 ohms to an 80 ohm (about) cathode isn't 4:1 -- but the problem 
was larger than that.

Next stop a grounded-grid triode stage -- or rather, two G-G stages (a 
6ES8).   Matching a 500 ohm filter to a nominal 80 ohm cathode 
(1/12,500 uMhos) 6ES8 cathode called for tapping a 30 turn toroid at 
12 turns. But the filter passband wasn't quite the right shape, 
indicating a mismatch.   Digging further I discovered what I might 
have calculated at the start: my untuned transformer didn't have 
enough inductance to work right at 3180 Kcs.   In the audio world 
you'd say the transformer didn't have enough 'low end.'    The easy 
answer was to tune it and doing so allowed me to do the impedance 
match by adjusting the ratio of two capacitors in series rather than 
by adjusting the tap on a coil -- a bunch simpler for testing and 
simpler to duplicate, as well.

The overall gain was better but still too low.  I increased the gain of 
the 2nd IF to the maximum and boosted the RF stage slightly: still 
not enough.   

Reluctantly, I decided to try to feed the stage at the grid, giving a 
genuine cascode setup.  In the process, I learned another one of 
those obvious facts -- for any given unneutralized amplifier tube there 
is an input impedance above which it will oscillate.  Getting the 
correct ratio to step up 500 ohms to the grid with no trace of 
regeneration with the tube operating at the desired (near maximum) 
gain took about two weeks and maybe a dozen tries?   But it was 
worth it -- perfect clean signals, rock solid stability, a nice, flat filter 
passband, no pickup of stray sigs.   And there was now plenty of 
gain.  

(Lowering the grid impedance from 33 to 12k helped cut the stray 
pickup but the main thing was probably that with the resulting lower 
Q I was able to eliminate the trimmer cap, and so cut the total lead 
length in the grid circuit severalfold.)

Unfortunately ... the final measurements at the filter driver were noise 
floor 107dbm, IP3, 19.5 dbm,  and IFDR 84.5 db -- not significantly 
better numbers than before.   

Since the distortion that set the IP3 was presumably in the filters 
themselves, better sensitivity had to mean a better IFDR.   And the 
improved function said the measurements should be better.   
Something was wrong ...

Fortunately, my course in Receiver Design Self Taught In Only A 
Few Dozen Long Exasperating Lessons had now equipped me to 
deal with this. I had to push the URM-25 signal generators pretty 
hard to test this stage; could they (again) be the problem?   I 
repeated the IP3 measurement at two lower signal levels; by the 
theory, IP3 does not depend on the level at which the measurement 
is done but I got three very different answers.   Since the answer got 
better with increasing test power levels, the conclusion was that 
there was indeed distortion in the signal generators.   

The good news is that measuring at successively higher levels gives 
results that approach the right answer.   So the section from the filter 
driver on has an IP3 of better than 19.5 dbm and an IFDR better than 
84.5 db.   That's good enough for now.   

(A way to think of this:  The signal generators operate at constant 
conditions, hence the strength of the false signals from distortion 
there is a constant fraction of their output.   Changing the attenuation 
ahead of the receiver subjects it to a varying input; its 3rd-order 
distortion changes at three times the rate at which the input 
changes.  At a sufficiently high input, the receiver distortion swamps 
the false signal from the generator so the measured IP3 approaches 
the correct answer.)

What next?   The overall receiver gain is about right as indicated by 
S9 (i.e., AGC voltage about half cutoff for the controlled stages) being 
close to 50 uV.   But the noise floor measuring at the antenna is 
around -116 dBm (0.23 uV would be -120 dBm) -- that's marginal at 
best.  And a 70 dB IFDR at the antenna probably means that one or 
both mixers enter distortion before the filters do.  Since the noise is 
now mainly from the mixers, the performance probably can be 
improved just by reducing the mixer gain and increasing the gain of 
the RF stage.

Sounds so simple, doesn't it?

Walt Hutchens
KJ4KV