Afternoon all,

For the last few weekends I've been playing about with a VCO based around a
ceramic resonator - before that I was using an LC network, but short-term
stability was not terribly good and my PIC Huff-Puff stabiliser was not very
effective.  I realised that I had some resonators and found they are a lot
better and can easily be locked on frequency for long periods of time, but
specifying circuit values is proving more difficult than with the LC version.

I'm using a 74HCU04 in the standard Pierce topology (with both feedback
resistor and damping resistor).  I have two sets of varactors as the tuning
element (anodes to ground, cathodes to series capacitors to isolate the DC
control voltage from the resonator).  I've found that the lowest frequency
resonator that I have that can be pulled the desired 35-85KHz is 10MHz.  My
problem is not how to get the oscillator to cover my intended frequency/pull
range (I did that in a few minutes), but rather I'm struggling to calculate
part values that take into account tolerances.  It's trivial to add trimmers,
but it's a bit of a cheat.

The big problem is that the resonator has a +/- 0.5% tolerance, and +/- 0.3%
stability; at 10MHz, that's +/- 80KHz before I even start.  So to pull my
resonator down 85KHz below 10MHz I could potentially have to pull it anywhere
from 160KHz to 5KHz.  I've been plugging values into a spreadsheet to see
the effects of various capacitances, strays, tolerances, etc. and it's doable
if I use varactors with a very wide tuning range (over about 1:4), but these
are difficult to find/expensive over a 5V range.

I have some BB149a diodes (about a 1:2.75 range over 0.5-4.6V, 22pF down to
8pF).  I've used a +/- 10% tolerance for the varactor capacitances; the
datasheet doesn't give a figure for this, but I thought that it was a
reasonable error margin for reading from the characteristic curves.  These
diodes work great on the bench and give me my desired pull range.  The results
tally quite well with my calculated values, but according to my spreadsheet,
marginal parts (resonators, capacitors, varactors or inverter) will be outside
my specs.

This is a one-off application, although I will probably build several more in
future, so there's no real issue adding trimmers; this is a self-imposed design
(and learning) choice to make the construction repeatable.  The nominal
frequency of the resonator in the VCO is not terribly important either, as long
as there is a common off-the-shelf crystal with the same value - the VCO is
basically being used at an offset from a crystal filter's centre frequency and
is fed to one half of a mixer.  But obviously the higher the VCO frequency,
the more difficult it is to stabilise.

Am I flogging a dead horse with this approach ?  I'm finding that strays are
very significant with the Pierce gate oscillator even if they're tiny - would
I be better moving to a discrete Colpitts or Clapp oscillator where strays
tend to be swamped by larger values ?  The Pierce was chosen basically because
the DIP package takes less board space than TO92 transistors and their biasing
resistors, but that's not a major constraint; it just gives me more room to
play with other things.  How would the pros do this if they were doing a volume
application (without DDS... :) ?

I hope I've given sufficient detail; any questions, please feel free to ask
(although responses may be a little tardy as I have a fair amount of
travelling coming up over the next few days, so please bear with me).  Much
appreciated.

Regards,

Pete Restall
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist