The way I do it is thus:
Crystals age. If you measure them and then repeat the procedure
24 hrs. later, the results will be different.
First, use good quality crystals, usually of a name brand like
Fox, KDS, or Crystek. If one orders a batch, they are almost always
from the same manufacture lot, and will match fairly well.
Different lots may not match together very well. The manufacturers
pack them in bags of 500 or so, sometimes bags of 100. It is a
good idea to order at least a lot this size.
Lately, surplus "old new lots" are a better deal than new ones.
if you can find them. Sometimes you will get a lot with wildly disparate
results. If that happens, start over. Also, figure on at least a few bad ones
in even good lots. Throw them away.
I use a G3UUR muppet board oscillator featured a while back
on this list. Actually, I have 5 or 6 different test oscillators; that one
seems to work the best. I first mark all the crystals with a black marker,
numbering them consecutively. Then I measure them and record the results.
including the "output" of each crystal in mv. Capacitor measurements
are averaged for the lot and that figure is used. In building a filter, almost all
of the material criteria lies in closely matching frequency. The other things matter,
but not a whole lot. Both the series and parallel resonance should match
fairly well. But good results can be had with a "straight" measurement and
a "pulled" measurement with a capacitor. Usually the crystals that match
with one will be fairly close with the other. Sometimes they will not and those are
sent to other uses.
Then I can enter the results in a spreadsheet and sort them for best match.
Usually I can find five or six groups in a hundred that match to within
5-10 Hz. The rest vary, but I can usually get quite a few sets matched to within
50 Hz.
Then a day later, I do the same thing. But this time I will let each lot of
4-6 that matched, age for 1/2 hour or so. Measurements are taken and entered
into the spreadsheet as before. Sometimes a comparison of the two methods
reveals that the sets no longer match. But I can still get about the same number of
sets with a good batch; only the results will be a bit different.
A good SSB Cohn filter can be made with crystals that match to within 50 Hz or so.
However, the best results are from sets that match close; the closer the better. There
is a list in the files section of some 20.480 MHz crystals that I bought from
a surplus house and matched. They turned out pretty good. I have also had good luck
with some 11.092 MHz and 12.480 MHz ones found the same way. I have not been so lucky with
10 MHz HC49S crystals -- but they produced some good sets as well.
It is easier to match lower frequencies than higher ones.
I think there are 2 basic criteria: 1. Get good crystals from known good
manufacturers. 2. Age the crystals for a bit as they are measured.
(yeah; I know. It is a pain of sorts, but the results are better).
And, of course, always use a good oscillator, and counter, and DVM.
and capacitor meter to make measurements. It pays.
Now, building filters is another matter...
john
AD5YE
wb6ogd