[time-nuts] How far can I push a crystal?

Ed Breya eb at telight.com
Fri Jan 18 06:32:58 UTC 2013


Thanks all, for the feedback on this issue. In summary, I got these 
points out of the discussion on crystals:

1. The correct terminology is "pulling" the frequency.
2. Getting beyond about a few hundred ppm from the nominal frequency 
ranges from very difficult to pointless.
3. It's easier to pull down than up.

It looks like it would not be worth fooling around with crystals, so 
I'll just use the ceramic resonators. By the way, I just tonight managed 
to reach the correct geological layer of stuff out in the garage, and 
found the missing 10 MHz resonators, and a whole tray of other parts 
that were in reserve for completing this project from a couple of years ago.

For the curious: The 10.0594444... MHz is made by a PLL using the 
59.4444... kHz reference, which is 10.7 MHz divided by 180. The 10.7 MHz 
is a from another VCXO (which can use a standard crystal, ceramic 
resonator, or ceramic IF filter - easy) that's phase locked to a 10 or 1 
MHz reference, using two fixed dividers. The 10.0594444... MHz is used 
as the reference for a phase locked microwave brick oscillator, using 
n=120, to make 1207.1333... MHz, which is exactly one-third of 3621.4 
MHz, the low-band upconversion IF of the HP8566B spectrum analyzer. The 
1207.1333... MHz is harmonically mixed (m=3) with the first LO of the SA 
to produce the tracking signal centered in the passband of the SA. All 
of this is built into the modified carcass of an HP8443A tracking 
generator, originally built for older SA models. Using the new stuff, 
plus parts of the 8443A, the net result is a 50 kHz to 250 MHz tracking 
generator, with power up to +10 dBm, leveled within about 1 dB, and with 
130 dB step attenuator range - very nice for low RF and baseband work.

The 10.0594444... MHz is only one of many frequencies that could be 
multiplied by various n-values to give the same result, but it was 
chosen because it was very close to a standard frequency available in 
ceramic resonators, high enough that n didn't need to be too large, and 
it could be synthesized with a very simple PLL system.

I had all of this built and running, but I had made the fatal 
engineering mistake of putting way too much stuff in too small a space. 
Space was tight, so I squeezed the entire LF control system and 
synthesizers into one small can, and necessarily optimized for minimum 
IC package count. Then I found that there was too much crosstalk between 
virtually all the signals in the box, so there was too much phase noise 
to work at 300 Hz and less IFBW. The problems were irreversible - 
sharing IC packages for multiple signal processing was an especially bad 
move. After many hours of rearranging signal paths, adding shielding and 
grounds, and changing topologies, I concluded that I had to rebuild it 
the right way. So here I am. The two main frequencies will be generated 
in separate boxes, and no ICs will contain multiple signals that aren't 
being processed together.

This time I'll get it right, and finally wrap it up.

Ed





More information about the Time-nuts_lists.febo.com mailing list