[HBR] HBR2K -- Chapter 14 -- Large Signal Performance, Part 4

[hbr@mailman.qth.net]

(In previous updates I struggled to figure out how to do proper 
measurements of intermod free dynamic range {IFDR} and began 
sorting out the problems.)

After perhaps a month of receiver design kindergarten revisited (note 
to self -- do not skip this grade next time) I homed in on the last 
stage before the crystal filter -- a cathode follower used as a driver 
mainly for the purpose of impedence matching.   I found that by 
changing my initial design to operate the c.f. at higher current and 
higher cathode voltage I could get the same results at the driver input 
as at the filter itself:  a third order intercept (IP3) of about 11.5 dbm 
(db above 1 milliwatt) and an IFDR of about 82 db.   By increasing 
the gain of the 1st IF stage (and decreasing that of the second to 
maintain the same overall IF gain) I pushed the IP3 up to 15.5 dbm 
without hurting the IFDR.  

An IP3 of 15.5 dbm and IFDR of 82 db is fairly respectable but of 
course there are three stages between these measurements and the 
antenna so I'm still a long, long way from a good receiver. 

Next stop, one stage earlier -- the 2nd mixer.   This stage combines 
a tunable IF signal 5520-6020 kcs with the VFO signal 8700-9200 to 
yield the fixed IF of 3180 kcs.   Measuring at the signal grid of the 
mixer (Pullen, 6ES8), I obtained an IP3 of +13 dbm (tolerable) and an 
IFDR of 65 db.   In other words, I lost nearly 20 db of dynamic range.  
Looking at what goes into the calculation of IFDR, the noise floor was 
up from -107 dbm to -84 dbm -- a 23 db increase in the receiver 
noise. 

Gee ... I think I found the next problem to work on.   Noise in this 
type of mixer can arrive by three routes:

1.  The tube itself -- some is unavoidable.

2.  The filament circuit -- since the cathode is 'hot', if the filament line 
is noisy, noise will be picked up.   This seems very unlikely because 
there's a line filter and a pi filter in the filament line to the front end 
tubes.   And since the cathode resistor is 470 ohms, the cathode is 
not extremely sensitive.

3.   The VFO.   No VFO generates a perfectly pure sine wave signal.  
For conventional oscillator circuits (PLL's have a whole new set of 
problems) there's a spectrum peaking at the nominal frequency and 
sloping away rapidly on both sides.   However, when you're looking at 
3 volts of peak (desired oscillator injection) you have to slope down 
pretty far to get below the few-microvolt level that's the tolerable 
maximum at the 3180 kcs IF.   (One volt to one microvolt = 120 db!)   

It helps that the VFO frequency is so far away from the IF -- about 9 
Mcs to about 3 Mcs -- but 120 db is still a lot.

The VFO is likely a problem.   If you disable the signal input to a 
mixer then when you also disable the oscillator input there should be 
only a slight change in the mixer output noise.   But as things stand 
now, the change is significant.   

(Even a pure oscillator signal would cause tube noise at frequencies 
3180 kcs above and below the oscillator frequency to be mixed to the 
IF so it will produce some increase.)

You can think of a VFO as a regenerative detector operating in 
oscillating mode.   We all know how a regen amplifies the tube noise 
-- that's the famous regenerative hiss.   But that hiss also modulates 
the pure sine wave you'd like to get from a VFO.   The problem can 
be minimized at the source by using a high Q tank and indeed a 
crystal oscillator causes much less trouble.   

Here we see the bite of a design compromise:  In order to get the 
oscillator to track the dial, I was forced to use a higher L/lower C 
combination than I wanted.   This is by design a relatively noisy 
oscillator.

There's yet another potential oscillator issue.   By the nature of the 
design, the VFO is about 4" from the 2nd mixer.   Each circuit is 
grounded in its own location -- the usual good practice.   But that 
means that 4" of chassis is in series with the oscillator injection.   
Bypassing the injection grid to the cathode of the stage does much 
more to reduce the output noise than does bypassing it to the 
chassis nearby, suggesting that noise voltages on the chassis are 
an issue.

Since the signal input to and output from the 2nd mixer are 
transformer coupled I can return the mixer signal grounds to the VFO 
-- the cleanest way would likely be via the shield of the RG-174 
injection cable.   At the same time I'll move the buffer plate tank 
circuit to the oscillator assembly (temporarily mounted at the mixer 
but radiation from it causes some humongous spurious responses) 
and provide a place to add a 3180 kcs trap on the oscillator output if 
needed to reduce oscillator spectrum noise. 

(Can't use an untuned buffer at 9 mcs because the stray 
capacitances are too large to allow a high enough plate impedance 
for the necessary gain unless circuit is tuned.   Another possibility is 
a filter with a peak centered on the VFO range and a notch at 3180 
kcs.   This is complicated by the need for a substantial bandwidth on 
the peak.   Note to self: study some filter design next time, too.)

This means, of course, a substantial rewiring of the 2nd mixer plus 
removing the VFO assembly: I sure could have provided better 
access to the screws that hold the darn thing in place.

I'm beginning to understand why there were few good receivers 
produced in the 1970 time frame.   Of course some of those 
designers knew a lot more about this stuff than I do so they'd have 
wasted a lot less time falling through holes in the learning curve.   
But most of the best of them were working for Collins ...

One exception -- the Drake R4C is at the top of at least one list of 
high performance receivers.   With a noise floor of -138 dbm and an 
IP3 (@20kcs spacing) of -10.5 dbm it has an IFDR of 85 db -- truly 
excellent performance for its day.   

Anyhow, it's a hobby.   I don't have a deadline, although I'd like to 
have the project sufficiently wrapped up to set it aside over the 
summer.

Walt Hutchens
KJ4KV