[HBR] The long, SLOW HBR project

[Walt Hutchens]

Subtitle: This wasn't PLANNED as a learning experience ...

> I guess I'll tidy up the local oscillator area first.

Not too much to that ... just a couple of tacked on parts needing more
permanent mounting.  But the warm-up drift -- up a couple of hundred cps
(since the slightly space wound LO coil)  then downward 800 cps or so below
the starting point (max to min = 1 kcs) -- seemed just a bit high,
especially when translated to higher bands.

Changing the one remaining mica cap for an NPO ceramic and putting a
tinplate shield over the tuning cap to reflect heat from the nearby mixer
tube -- I should have put a bit more space in there when laying out the
chassis -- got the 1 kcs max-to-min down to 600 cps or so.

But ... where was THAT?   Nothing in the tank circuit area increased in temp
by over 10 degrees F or so, the coil had a negative coefficient, and the
(shielded) high quality cap didn't seem likely to contribute much.   Plus
all this testing was done at 4 Mcs, close to minimum capacitance.

I started measuring temps here and there as the set warmed up and after a
couple of days of intermittent testing and staring, it hit me:  The top
surface of the chassis is constrained on all four edges.  It's aluminum
(high temp coefficient) and where's it going to go as it expands?  In theory
it could either bulge upward or downward, but the way it was bent, I
expected it to bulge up.

The tuning cap is mounted on the 4"x5" surface of the top of a minibox,
fastened on its side on the right edge of the main chassis.  When I did
that, I thought "plenty stiff."   Well, it IS stiff enough to avoid backlash
due to chassis flexure when the dial is turned (the first thing you worry
about), but if you assume the top surface of the chassis bulges a bit with
temp. increase, then the minibox would be twisted, causing the tuning cap
frame to rotate in the direction of increasing capacitance while the rotor
was held in place by the dial.

There's a link to a picture showing the tuning cap mounting, below.

This effect is, of course, MUCH too small to be visible.   But (say) 500 cps
when half a revolution is 500 kcs, is 1/2000 of a revolution of the tuning
cap ... that wouldn't be enough to see.

When I pressed on the various chassis parts in the corresponding directions
I had no trouble reproducing the amount of frequency change.   And when --
after a full warm up -- I dropped a small lump of ice at what I judged to be
the hot spot on the chassis, I was rewarded with a 50 cps increase of
frequency over the next minute or so, even though that location is several
inches from the oscillator parts.

Good enough for me.   I spent a couple of hours making a vertical plate to
go inside the chassis under the front of the minibox, to brace it to the
side of the chassis.  Because the lower lip of the chassis is folded inward
(and the corners are welded) such a brace should greatly reduce bulging
along its top edge and thus reduce rotation of the capacitor frame.   Of
course this concentrates the bulge in the left 2/3 of the chassis surface,
but for an effect too small to be detected except by oscillator drift, that
should be okay.  

For good measure I put small triangular plates in the lower corners of the
chassis to lock the right bottom lip to the front and rear lips.

Warm up drift was reduced to ~400 cps, max to min.

There's a reason all those military sets are made from 1/8" or thicker
material AND folded and braced to be extremely stiff.    And for those
wondering why anyone would build on a steel chassis, here's a reason:  Steel
has a lower coefficient of expansion AND is considerably stiffer.

I'll stick with aluminum, but it's not a perfect material.   However steel
might be a good choice for a VFO.

The next step was reducing the heating of the chassis by drilling six 1/4"
holes around the 117N7 -- at more than 10 watts (out of about 35)  the
single greatest source of heat, plus it is almost directly lined up with the
front edge of the minibox.  This reduced warm up drift to 260 cps (max to
min) over half an hour.

Hey, if six little holes is good, one great big hole must be better, right?
So I knocked out a 2" octagon around the 117N7 and mounted a 2-1/4" square
plate with the tube socket about 1/4" below the chassis surface; air pulled
by the high temperature of the tube can come up between the plate and the
chassis, thus cooling both.

That reduced the drift to ... er, wait ... it INCREASED  it to over 500 cps.

???

Another day of unhappy head scratching clarified the issue: While the large
hole and sub-mounted plate/socket no doubt reduced the temperature rise
somewhat, the way I did it greatly increased the stiffness of that (hottest)
part of the chassis, thus INCREASING the bulging of the rest and the
resulting drift.   Stiffening oscillator area = GOOD, stiffening elsewhere =
BAD!   

Perhaps the answer was to mount the new plate and socket less rigidly.  One
trip to Lowes later, I had four 1/4" nylon bushings with nylon 6-32 screws
and nuts to match and a few minutes work with needlenose and screwdriver
replaced the steel and brass hardware I had originally installed.   THAT did
the trick: Max-to-min drift was reduced to 153 cps.

Testing at 3.5 Mcs with the final set of changes the drift is about 500
max-to-min.    

'Max-to-min' is of course a fairly stringent standard.   Drift after 20 min.
warmup might better reflect ordinary use: Those numbers are 62 cps at 4 Mcs
and 153 cps at 3.5 Mcs, to a single measurement at least two hours after
starting, with the per-hour drift thereafter expected to be <50 cps.

There are pictures of the first chassis here:

https://picasaweb.google.com/105255380483660395824/WaltS2008Project

The 5th pic from the bottom is the clearest view of the tuning cap mounting.
The rotor of the cap turns CCW (as seen from the rear) to increase
frequency.  If you imagine the chassis bulging upward and rotating the front
of the minbox on which the cap is mounted CCW, you can see how downward
drift would occur. 

I started over with a new chassis perhaps a year ago, transferring most
stages with minimum change.  All the major parts are in the same places (and
the panel was reused) but several tubes and the IFTs got relocated for a
much cleaner signal flow and shorter leads.   I'll get new pictures up
within a few days.

Doing it again I'd rethink the stiffness/heat flow issues yet again.
Whether that would lead to significant change, I don't know: There are
already many other constraints on the layout.

Examples of constraints: The minibox shields the antenna coil from the mixer
and IF stages and keeps the heat of the oscillator tube off the oscillator
coil -- one piece of standard metalwork replacing two or three needing
custom fabrication.   Easy access to all coils is required. Since even
objects 8" away affect the frequency it is impossible to have too much
clearance around the oscillator coil.   The oscillator coil and BFO are best
placed near the front since both have front  panel controls.   The two
filter chokes and the BFO coil are conveniently mounted on the left chassis
side because the wiring of those stages is simple enough that the sockets
can be mounted under these three components.

Then there are the usual rules about trying to get the stages in a straight
line.   With one thing and another, this project is not too open to further
improvements in layout -- but that doesn't mean that they're impossible.

However, the lesson that the top of a chassis can change shape during warm
up and thus contribute substantially to drift is a useful one.

The idea is that air will enter the chassis from below.  For testing this
was done by raising the left and right edges on 1/4" strips.   When the set
gets a cabinet, I'll probably cut a 1" or larger hole in the rear of the
chassis.   Forced air into the chassis using a small computer fan might make
sense, although the DC power requirement may be an issue; I'll look at using
the cathode current of the 117N7 output stage.

Walt
KJ4KV