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

[hbr@mailman.qth.net]

Mike Feher wrote:
> This implies that your receiver's NF (noise figure) is 31 dB. Does
> that make sense? I would be surprised if it was over 15 dB. So, to
> me, right now things just do not add up. 

In other words, on fundamental theory it ought to be a heck of a lot 
quieter (more sensitive) than it is.   I agree with that, and that's why 
I'm still working on it!

Let's say a kid gets some bicycle wheels, a small block Chevy 
engine, the transmission from a riding lawnmower, a couple of V-
belts, um..., a 4'x8' sheet of plywood and a few 2x4's.   He has 
ridden in some cars and once made a soapbox racer so he decides 
to build a real car.   An engineer friend calculates that with the gear 
ratio he has, an engine of that displacement (etc.) the top speed 
ought to be 72 mph but the kid can only get 'er up to 30.   

The engineer's theory fails to take into account Designer Error.   
There's simply a lot more to stitching together a complete car (or 
receiver) than the fundamentals tell you.   In this case the 
fundamentals apply only to a receiver in which a reasonable input 
stage dominates the noise.   Well, of course that's how it *ought* to 
be.   But I am not there yet, and going by the direct route -- simply 
cranking up the gain of the RF stage -- would (at the current state of 
things) compromise the dynamic range.

One disadvantage I have is that when a design like this is done 
professionally the component parts are fairly well characterized by 
the manufacturers.   Custom parts are designed to specific targets.   
For example, the impedence of an IFT and the gain of a stage using 
a certain tube with that IFT would have been known before the stage 
was wired in a prototype.   Ditto the drive level for a certain crystal 
and the output a particular oscillator might produce with that crystal.  
Of course the optimum drive for a mixer would be known be before 
the prototype set was built.   And so on.

I'm working with ideas and bits of circuits from other designs that 
were incompletely tested (or at least, incompletely reported) and 
parts of somewhat unknown properties.   (1415 kcs IFTs retuned to 
3180 kcs, for example, for reasons of availablity and ease of 
modification.)   Some of the preliminary testing that I perhaps should 
do, I don't have the right instruments for.  Some I'm just too impatient 
to do -- I don't know what I *really* need until I see the problems and I 
fear spending a week studying something that turns out to be 
unimportant in the larger picture.   Probably this reluctance to do 
side-studies has hurt me, overall, but of course it's harder to see that 
one step at a time.

And then there are the 'just plain problems.'    When I began this 
note yesterday I was optimizing the 2nd mixer.  I had tried a couple 
of different circuits, minor variations on one of them, and was down to 
tweaking component values.   The trend was to better performance 
(IFDR) at higher mixer plate currents so (since I've got more overall 
gain than I need) I decided to raise the cathode resistor of the 2nd IF 
stage and save some ma there.    Huh?   All of a sudden the noise 
floor and IFDR took about a 2 db hit.   There's no logical reason they 
should be the least affected by the 2nd IF -- it shouldn't be 
contributing any noise, and indeed when everything up to there is 
disabled, the receiver is completely dead, even at full volume.   AGC 
is disabled for these tests so that couldn't be it.   It had to be a 
subtle feedback effect, but bedtime intervened.

In the early AM, sudden wakefulness.   There were a couple of 
obvious feedback paths from that last IF to early stages and I started 
by listening to the noise audio output with stages ahead of the 2nd 
mixer disabled.    Sure enough, there was a nearly inaudible signal 
that must have been coming in at around 3080.5 kcs -- so faint you'd 
never hear it with any other signal, let alone with an antenna 
connected.   I started poking around with bypasses and it was on the 
filament line, probably coming in through the power line despite a pi 
filter on the line input connector, the usual .01/2kv bypasses after the 
fuses and 0.1 bypasses at the main filament distribution point.  

If the filaments can propagate an external stray signal, they can do 
the same with an internal one -- there's a feedback path! 

I had wired all the filaments with 0.01 bypasses.  But the reactance 
of those at 3080 kcs is about 5 ohms -- not exactly a dead short, 
when filament levels of resistance (about 20 ohms per tube at 
operating temp) and tens or more of mmf filament to cathode 
capacitance are considered.   I got rid of the stray signal with 0.47 
bypasses where the filaments enter the front end compartment under 
the chassis.   And sure enough the noise level dropped a couple of 
db and the NF and IFDR are now the same with 150 ohms for a 2nd 
IF cathode resistor as with 68 ohms.   Most likely the path was from 
the cathode of the 2nd IF to the (unbypassed) cathode of the 2nd 
mixer, via the filament circuit.  The difference in feedback voltage 
when changing the cathode resistor was small, but there's a heck of 
a lot of gain (three very hot tubes) between those two cathodes and a 
2 db change in the noise floor caused by regeneration isn't much.   

In the future I'll be rewiring the filaments again and when I do, the 
bypasses will go from 0.01 to 0.1 from the 1st mixer on.    

I think you learn this stuff only by practical experience of which I 
didn't have enough for this job. Theory certainly does inform, but it 
doesn't let me avoid all such several-hour pitfalls.  I have assorted 
handbooks, but the approach is either -- "Build it like this" with a 
circuit diagram and where to get the parts (in 1965), or "Here's the 
theory of a receiver" with the parts being square blocks of specified 
gain and noise figure.   I really need a lab manual -- from about 1960 
or so.   As I write the 'notes' for the receiver -- 40 pages and counting 
-- I am including all the pitfalls I hit, so hopefully they only have to be 
hit once.

The grand plan has been to build the receiver from the power supply 
and audio, thence one stage at a time to the antenna, get it 
'basically working', then to repeat the journey from audio back to 
antenna making things as near 'really right' as I can.  I'm well along 
with the second pass; satisfied with everything from the filter driver on 
and probably close on the second mixer.  I need to finish up the 
optimization there, before moving back to the 1st mixer.   

A long story which I hope makes clear why a noise figure around 30 
db does not seem impossible right now, considering problems that 
may be lurking in the still only 'basically working' RF and 1st mixer 
stages.   I know for a fact that compared to really good receivers, this 
one is too noisy.  I strongly suspect the that 1st mixer has lower 
gain than it should because of insufficient injection.  The RF stage ... 
well, I'm pretty sure it's bad, although I can't tell you the details yet. 

I began with the belief that the set ought to be able to outperform an 
R-390 (IFDR around 80 db) because I've got better tube types and 
coils than were available in the 50's. I've seen nothing to change that 
view.  But it's still pretty far from there.

As always, thanks for the comments, Mike.   

Walt Hutchens
KJ4KV