[HBR] HBR2K -- Chapter 15 -- Large Signal Performance, Part 2

[hbr@mailman.qth.net]

Subtitle: The education continues.

As I increased the overall gain and cleaned up assorted problems the 
HBR2K's third order IMD dynamic range wound up 55 db.    
Reference points:  FT-101 (early transistor set, 1970) -- 55 db, KWM-
2A -- 62 db,  R-390 -- 80 db,  current state of the art -- 100+ db.   

The receiver sounds outstanding -- way better than the second best 
receiver here, a KWM-2A.   But you'd sooner tell your mother you 
make adult toys for a living than that your receiver has a 55 db IMD 
dynamic range.   

Several days of thrashing about followed.   The usual problem is in 
the first mixer but that didn't seem to be the case here.   

Shoot.   Finally I figured out how to do the measurement on 
successive stages and began to pin things down.   It was equally 
bad (55 db) at the grid of the second mixer but went up to the over-60 
db range when measured at various other points from the second 
mixer plate to the filter input.   Clearly the second mixer is a 
problem.  

But why only 63 db at the filter input?   This is the place to start work 
because there's no obvious reason it should be so low and the 
receiver's overall performance can never be better than the worst 
stage.

At the filter there's sort of nothing left that ought to be non-linear 
before the filter throws away the two unwanted signals.   To clarify:  
Feeding signals at 3200 and 3220 kcs into a 3180/2.4 kcs filter can 
only produce a spurious 3180 signal if mixing happens *in or before* 
the filter -- after the filter the 3200 and 3220 kcs signals are so much 
attenuated that they're not going to cause a problem.

I disconnected all the filter selection diode switching stuff and wired 
the SSB filter directly.   The dynamic range went up to 68 db.   The 
filter switching is done with diodes, chokes and resistors; at first I 
suspected the diodes and the chokes.   The signal required to 
produce an S-1 spurious signal is 12,500 uV and the filter impedence 
is 500 ohms; maybe the diode switching current is not high enough 
to avoid distortion?  

But no such luck -- doubling the switching current made *no* 
difference in the dynamic range.   There followed several more days 
of poking at the problem with a sharp stick that eliminated non-
linearity in the filter switching parts as the limiting factor.

Dynamic range is the ratio of two numbers:  The largest signal the 
circuit can handle without distortion, seemingly set in this case by 
the filter itself, divided by the smallest, set by internal noise or gain.   
With circuits ahead of the filter disconnected (and no input) the noise 
from the 1st 3180kcs IF amp is not sufficient to cause S-meter 
deflection so the limit is probably gain, however the noise is easily 
audible so a noise limit isn't far off.  

The upper limit could be hurt if the filter is mismatched, however I'm 
satisfied that that isn't the case.   The passband looks perfect and it 
sounds just fine -- you don't get that with a mismatched filter.

Most likely the filter switching circuitry hurts the performance (68 db 
with one filter hardwired, down to 63 with switching for three filters) by 
increasing losses.   Probably this can be improved by changing the 
switching circuit.   It certainly seems that relays would be better than 
diodes and resistors but I'll try some other things first -- I'm not sure 
there's space for relays.   

(In a complete redesign one could consider using the MODE switch 
directly although the mechanical challenges might be considerable 
and the FT-101 parts set would not have the necessary switch.)

The FT-101 was the first mass-production modern HF transceiver, ca 
1970.   With a 55 db overall dynamic range set by the available 
transistor technology there's no reason it should have filter circuitry 
that will do better than 60-some db.

Next steps:

1.  Measure the noise floor at the 1st 3180kcs IF stage -- a 6EH7.    
Improve this number if possible.   (Maybe a cascode amp instead of 
a pentode?)   Look at increasing the detector sensitivity.   Increase 
the IF gain.  I may try 6KT6's -- half again the Gm but still good AGC 
characteristics.

2.  Try to reduce switching losses in the filter circuits.   

3.  When performance at the filter is as good as possible, tackle the 
second mixer.

Those who want to know a lot more about IMD testing of high 
performance receivers will find a very interesting article on the AOR 
(UK) web site at:

www.aoruk.com/comments.htm

The comments are by John Thorpe, the designer of the AR 7030.  
Way, way the best writing I've seen on the subject.  

Walt Hutchens
KJ4KV