[HBR] Another Receiver Project -- HBR-4, Part 18

[waltah at earthlink.net]

'Some progress' edition.

With the antenna signal applied to one of the 6JH8 deflection 
plates (and assorted tweeking) the sensitivity is about back to 
where it was and the receiver sounds really great -- better than 
anything else I've got.    Testing time ... I hooked up my two 
homebrew signal generators through the homebrew hybrid 
combiner, dialed up 3800 and 3820 and the strongest signals 
they're able to deliver:  IFDR 77 db.   Not great, but another notch 
better.   

Unfortunately there's not enough gain to use these generators to 
test at higher levels so it would be possible to determine if the 
problem was in the receiver or was due to bad signals from the test 
setup.   

One way the test signals can be bad is if they contain harmonics.   
I built a 7-element lowpass filter (think: two pi nets hooked end to 
end) designed to pass the 75 meter band and kill 40 (where the 
second harmonic of my test sigs would be.   Gee, when I hook an 
antenna through it to the receiver, 80 is unaffected and 40 is gone.  
 These are already *very good* signals; running through that, 
there'll be no significant harmonic content at all.

That got the measured IFDR up to 81 db -- slightly the best figure 
for any receiver here, but I don't have anything really good.   

The way to pin down the bad guy is to do the test again at higher 
power levels.   Three or four tests, say, one at the noise floor, one 
at S-1, and one each at two S-unit intervals to S-9 and in general 
agreement, confirm the limit as being in the receiver.   If these 
measurements do *not* agree, the trouble is somewhere in the test 
setup or (for complex receivers) possibly in a receiver oddity.

A low pass filter gets rid only of harmonic issues in the test 
signals.  (This could be caused by a non-linear amplifier in the test 
set, in fact, a little non-linearity is basically certain to be present, 
even in a class 'A' amp with a tuned output circuit.)   If, however, 
the basic problem is non-linearity of the combiner or 
crossmodulation of the signal generators, then we're looking at the 
two intended signals at 3800, 3820, being accompanied by 
spurious signals at 3780 and 3840 as they come into the set.  
There's no way to get rid of those with a low-pass filter!

Higher receiver gain will allow testing at higher levels of mixer 
overload and since the receiver does need more gain, that's the 
next step.   However, the needed gain is in the range 10-20 db and 
the max increase in IFDR I can measure is only as great as the 
increase of the gain.   To get beyond that, I'll need to modify the 
generators for higher output -- say 1 volt p-p signals instead of 
100,000 uV.   That's tougher than it sounds because the 
generators must also have very high isolation between the non-
linear part of the generator (the tubes) and the output jack.   

Because tubes are non-linear, if you feed a signal in at the output 
jack of a sig gen, it will modulate the signal produced by the 
generator and in IFDR testing, that modulated signal looks exactly 
like the bad signal produced by a non-linear mixer in the receiver.   
So the tubes in two signal generators used for this kind of testing 
must be unable to 'see' each other.   

Part of the job is done by the hybrid combiner -- a weird little circuit 
using a small transformer and a couple of resistors that (ideally) 
takes 1/4 the power from each generator and adds those two sigs 
together at the output jack (a hybrid combiner has two input jacks 
and one output) and throws all the rest of the power away, so 
neither input jack (signal generator) sees anything from the other 
one.

However, practical combiners don't provide much more than 40 db 
isolation between input ports.   So the signal generators 
themselves need to provide isolation.   It's conventionally done with 
a voltage divider.    My homebrew generators produce a 30-volt p-p 
signal and a chain of resistors takes that down to 100,000 uV 
across 50 ohms.   That's most of 50 db additional, giving (with the 
combiner) maybe 90 db isolation beween each generator and the 
output signal of the other. 

You can see the practical problem of increasing generator power.  
If I don't change anything else, then stepping up to 1V p-p output 
requires a 300-volt p-p signal on the voltage divider.   Among other 
nasty problems, what kind of sheilding will it take to keep that 
bottled up inside the sig gen case, so it won't go right past the 
attenuators?   I'm using well-shielded military signal gen cases, but 
they're not *that* well shielded!

Even deciding to use less isolation won't work very well because a 
lower series resistance in the voltage divider chain will lower the Q 
of the amplifier tuned circuit, increasing any waveform distortion.

I'm not eager to start on signal generator power increases.

Another problem is that the receiver crystal filter is right behind the 
first mixer.   With a highly linear mixer, what's to stop a powerful 
input signal from destroying the CF?   Adding an AGC-controlled 
RF stage will help for in-channel signals, however nearby out-of-
channel signals don't produce AGC voltage.   Very high 
performance receivers deal with this by putting a local AGC loop in 
the front end so that *any* extremely strong signal getting through 
the tuned circuits reduces the RF stage gain.   Typically these 
circuits start to take effect around S-9 to S-9+40 and they 
introduce their own set of new issues.  

Fortunately the filter I'm using is stolen from a Tempo-One which is 
a single conversion set.   So it should be a fairly robust filter.

Next step: adding an RF stage.

Walt 
KJ4KV