[HBR] That General Coverage HBR Project -- 5

[Walt Hutchens]

Call this 'fun with a VXO' -- variable frequency crystal oscillator.  Or
"how I learned more than I wanted to know about this subject."  Or --
still closer to the truth -- "What happens when you suddenly think
you've made a big design mistake."

Quartz crystals are equivalent to a tuned circuit consisting of an
inductor, a capacitor, and a resistor all in series, with the lot
shunted by a small capacitor.  What's remarkable is that the inductor in
this equivalent circuit can be in the ballpark of 0.1 to a quarter of a
henry at frequencies of a few Mcs, giving typical Q's well in the tens
of thousands.  For modern crystals with plated electrodes, Q = 100,000
is common; the older pressure mounts (FT-243 ...) are not so good.  You
can often see where a pressure-mount crystal that has been driven hard
has rubbed some metal off on the corners of the plate itself showing
that friction does occur.

Crystals have two resonant frequencies.  They appear as capacitors over
most of the spectrum, however, over a narrow range of frequencies
between the series and parallel resonances, they look inductive.  Nearly
all oscillators work within this range.

The series resonant frequency is the lower of the two; at this point the
crystal appears to be a resistance that's usually in the ballpark of
10-1000 ohms.  Conceptually the oscillators that use the series
resonance are those like the Butler (crystal between cathodes of two
triodes) in which you could think of replacing the crystal with a small
resistor.

A fraction of a percent higher in frequency, the crystal appears to be a
parallel resonant circuit.  However what we usually think of as parallel
resonant crystal oscillators (Pierce -- crystal from plate to grid,
Miller -- crystal from grid to ground, many others) actually operate
more or less on the inductive side of resonance -- that is, between the
series and parallel resonant points but nearer the parallel resonance.

Typical spacings between the two resonant points for crystals I've
fooled with in the last few days are 1 to 4 kcs at 4 Mcs.  The spacing
goes up roughly proportionately with frequency.  Crystals are made and
calibrated to operate either at series resonance or at parallel
resonance with a specified parallel load capacitance -- a convenient way
of saying how close they're supposed to be to the parallel resonant
point.  Common load cap specifications for modern crystals are 20 mmf
and 32 mmf; some of the large pre-WW II ones were operated with much
more, I believe.

Either series or parallel resonant crystal circuits can be adjusted in
frequency by varying the reactance in the external circuit.  For series
resonant circuits the practical range of variation is fairly small
because of the awkward component requirements (either a variable
inductor or a capacitor that, if of reasonable size, will give only a
modest range of tuning) and accordingly these oscillators are usually
made adjustable only for trimming purposes -- series resonant 100 kcs
crystals for calibrator service are an example.

On the inductive side of parallel resonance, however, the crystal
oscillation frequency can be shifted considerably with a fairly small
shunt capacitor.  In the Pierce VXO circuit, stable oscillations can be
had over most of the range from the parallel resonance (with minimum
shunt capacitance) down to the series resonance.

A commonly used trick is to add an inductor in series with the crystal.
If you think of the inductive reactance of the crystal dropping from
infinite to zero at the frequency goes from the parallel to the series
resonance you can see that the added inductor moves the apparent series
resonance down but has very little effect on the parallel resonance.
The VXO can now be tuned over a wider range.

However, if you keep adding inductance, the inductance of the crystal
forms a smaller and smaller part of the circuit.  Since it's the
variation of the crystal's inductive reactance that allows it to control
the frequency, you may lose crystal control.  VXOs with too-large series
inductors may suddenly hop from a crystal controlled frequency to a
nearby frequency that is only slightly affected by the crystal as they
are tuned downward.  Reducing the Q of the coil with a resistor reduces
the effect but gives you an overall reduction of Q (meaning less stable
frequency) in exchange ... now why were we using a crystal?

A typical value for the series inductor at a few Mcs is 10-50 uH.  I'm
using 44 uH right now and getting a tuning range of about 4.5 kcs at
4096 kcs.  It's very stable, even at the low end, and I'll probably go a
bit higher on the coil.

Tuning of a VXO tends to be non-linear: the first few mmf in parallel
with the crystal change the frequency rapidly but as you continue to add
capacitance, each successive mmf has less and less effect.  Broadcast or
similar type non-linear capacitors help.  Typical maximum values at a
few Mcs are 100-350 mmf.

This suggests that the range can be increased by decreasing stray shunt
capacitance and it can.  However the oscillator tube sets a lower limit
and the instability of the tube capacitance becomes proportionately
greater at the same time that it has a greater effect on frequency per
mmf of change.  A push-pull oscillator (crystal between the grids,
fixed-tuned plate tank circuit) might work well.

An additional issue is that as shunt capacitance is increased the tank
reactance falls, meaning reduced voltage output.  At some point, adding
more capacitance causes the circuit to go out of oscillation.

I suspect that the common Pierce VXO is not the best possible circuit --
as in other areas, tubes went out of favor before an optimum was
developed. But it's probably not worth chewing on this just now.

There are two very interesting techniques for further increasing VXO
tuning range.  One is the use of two (or even more) nominally identical
crystals in parallel -- called the 'super VXO.'  This effectively
parallels the two inductances so the effect of the minimum (stray)
capacitance becomes smaller and by using a slightly larger tuning cap,
the tuning range is increased.  Very simple and with the better crystals
it works well.

The second technique is the use of two crystals of different frequencies
feeding a mixer which delivers the difference.  For example, a VXO at 5
mcs could be implimented with crystals of 12 and 17 Mcs.  Some builders
tune the two in opposite directions: with (17-12) = 5 you'd get ~six
times the tuning range of a 5 Mcs VXO while somewhat correcting the
tuning non-linearity and output voltage variation.  You would, however,
need a differential tuning capacitor.  I have seen one version of this
circuit in which a single bipolar transistor was used, simultaneously
running one crystal in series mode (fixed tuned), the other in shunt
mode (tuned) and providing output to a tuned circuit in the collector
that was tuned to the difference.  The same probably could be done with
a vacuum tube.

As with most things, the circuits that do the most to lift the inherent
limitations are also those that most complicate a circuit that's of
value mainly for its simplicity.

My day of wondering if I had seriously goofed came from discovering that
the clock crystals I used for the filter had series-parallel spacings of
only around 1 kc.  Nor did these crystals take well to use of a series
inductor. However, after some digging I found an FT-243 at 3990
kcs with a 4 kcs spacing and ground it to the desired range.  I had not
tried this exercise in about 20 years and I don't think I ever moved an
80-meter rock by 100+ kcs successfully, but it worked this time.  Hint:
a piece of quartz .016 x 3/8" x 3/8" is easily mislaid if a dog runs
between your legs while you're handling it.  Part of the reason for the
success was undoubtedly that I did most of it between 2-6 AM on two
nights.

It looks like I'll end up about where I had hoped, with a Pierce circuit
covering 4095.5-4099.5 kcs.  I'll probably wind an adjustable coil to
replace the two 22-uH chokes I've been testing with and then install a
socket for the crystal.

With the 'broadcast' type capacitor I'm using, the non-linearity isn't
too bad.  The tuning range is 4.7 kcs and the midpoint of the knob swing
is about 2 kcs from the lower end.  However cleaning up the wiring is
likely to add another kc or so at the high end making the linearity that
much worse.  It wouldn't be too difficult to improve that by grinding
the capacitor plates some on the high frequency end, give or take the
possiblity of trashing the cap.

There are still a couple of serious issues.  The plate waveform is
dreadful at the high frequency end -- a square wave with trash.  That
means lots of harmonics that might (will ...) be picked up by the front
end of the receiver.

Furthermore, there's a tenfold variation of output voltage across the
tuning range.  That's not going to work, driving a 6BN6 product detector
directly.  Fortunately I don't actually need much output -- just a
couple of volts p-p, I think, for the second grid on the 6BN6. I will
try a pentode ECO circuit (screen of pentode acts as plate for the
oscillator); I suspect that the tube current is a lot more consistent 
across the range and that the variation in the plate circuit will be 
correspondingly more reasonable. I had noticed that Pierce ECOs are 
common as VXO's; maybe I now know why.

A tuned circuit in the pentode plate will reduce the waveform problems 
too, although there'll still be radiation from other components that 
carry the trashy square wave.  With only a tuning range of 0.1%, this 
circuit can be fix-tuned.

A minor additional factor is that crystals that are ground (rather than
etched) tend to 'age' upward in frequency as fragments break off the
sides of the scratches caused by grinding; this effect can be hundreds
of CPS.  So I'm not going to be in a hurry to move the crystal the last
kc or so.

Doing this again I would reconsider use of separate crystals for upper
and lower sidebands.  In this design it was necessary to put the
oscillator at the rear of the chassis; I thought that using a VXO with a
long shaft for the capacitor was a good solution to that problem while
offering a measure of IF shift operation.  However the relatively low IF
is making that a challenge.

I think there's room to add another tube and a second crystal socket if 
that should be necessary.  It's always interesting, the way the various 
design choices interact.

Walt
KJ4KV