[HBR] HBR2K -- Chapter 14 -- Large Signal Performance, Part 5

[hbr@mailman.qth.net]

In the last part, I had the 20 kcs intermodulation-free dynamic range 
(IFDR) up to 82.5 db measured at the input to the crystal filter driver --
 the last point at which the measurement can be done before the 
filters mostly wipe out the test signals.   At the antenna, however, 
the number was in the low 60s -- better than the (solid state) FT-101 
but only on a par with the KWM-2A transceiver.

I first looked at the the 2nd mixer performance and the possiblity of 
noise from the VFO.   Making a long story short, while the VFO 
design isn't ideal from a noise viewpoint, I don't think VFO noise is 
the issue now.   I then turned to the tube operating conditions; 
increasing the cathode resistor from the usual published value of 
about 500 ohms to 2.2k increased the IFDR at that point from mid-
60's to 76 db -- good enough for a first pass.

Gee ... That was easy.   I moved back to the 1st mixer and by 
increasing the cathode resistor there to 1.5k and changing the 
voltage divider to reduce the mixer tube current, I got an IFDR of 71db 
there and 72.5 db at the antenna.  

(I made a mistake last time: IFDR *could* improve somewhat as you 
move backward through the stages if the added selectivity takes out 
the interfering signals faster than the stages add noise.   Since the 
HBR2K uses a tunable 2nd IF and about the highest-Q coils 
practical, this slight improvement is plausable.)

So, things are improving.   The best of the well known vacuum tube 
receivers are the R-390A at 81 db and the Drake R4C at 85 db.  
There's no obvious reason not to be able to hit or even better those 
numbers.   But doing so will require another pass, starting with 
improving the performance at the filter driver.   

The 82.5 db figure at the filter driver is composed of an okay IP3 
(about 16dbm) and a crummy noise floor (about 108 db).  The noise 
floor here is set mainly by the (ballpark) 10 db attenuation of the 
signal in the crystal filters and the noise floor of the 1st 3180 kcs IF 
stage.  

I have never felt I had gotten the filter matching precisely right so that 
was next: using a scope and resistive dummy loads I tweeked the 
filter input resistor (in series with the driver to filter connection) and 
output (shunting the matching transformer).   That improved the IP3 
at the filter driver to 17.5 dbm and the IFDR to 84 db.   (A couple of 
db is about the minimum you can be really sure is a change in either 
number, due to error in reading the S-meter, ripple in the filter 
passband, and so on.)

 I don't think anything else can be done about the filter attenuation -- I 
already changed to relay switching so unnecessary losses are gone.

There are two ways to go:  back through the mixers again, trying to 
really optimize the parameters in each stage and thus carry the 84 
db IFDR back to the antenna, or forward to the 1st IF stage where a 
change to a triode circuit might make sense.  Since the second 
choice is really a job (it requires a complete rebuild of the stage 
including removing and redesigning a shield across the socket) and 
even if it worked wouldn't likely improve results at the antenna until 
the mixers were redone, I will probably hit the mixers again first.

I believe that the reason the published Pullen designs aren't optimum 
for this set is that other parameters are different.   I've got higher Q 
and higher impedence coils in the mixer plates and the oscillator 
injection in the 1st mixer is somewhat wimpy.

But another visit to the 1st IF stage is clearly in the future.   For a 
given state of tube technology almost any triode is quieter than a 
pentode (right?) and I don't need so much gain -- the 2nd IF (6EH7 
pentode) is running at less than half throttle.   Might be able to drive 
the cathode of a triode and get rid of the tuned matching transformer? 
Input impedence = 1/transconductance shunted by cathode resistor? 
So for 500 ohms and modern tubes that would be about 5:1 step 
down.

Matching a grid is the pits -- the impedence is so high that pickup of 
stray signals and feedback is a problem.   But a mismatch is very 
costly because of the filter attenuation and noise floor issue.

I'm now using the coils of the two reed relays for each filter as the 
cathode resistor for the audio output tubes -- simple, and minimum 
parts count.   I'll probably try using a switch there as a muting circuit; 
I am doing that in the AGC circuit now but fixing the long time 
constant there will require some work.

The thing is gradually shaping up.   It sounds fine, doesn't drift, 
redoing the filter switching and adding a shield there gave about 80 
db of rejection 10 kcs off frequency which is as much as the filters 
are capable of.  

Walt Hutchens
KJ4KV