[time-nuts] Best way for generating 8994.03 MHz from 2899.00042272.....MHz?

[Javier Serrano]

Dear nuts,

A colleague from a Free Electron Laser lab has the following problem:
he needs to make a frequency to use as an X-band LO that is
*exactly*8994.03 MHz (3*2998.01 MHz) and it
*must* be locked to his S-band LO which is exactly 2998.01*732/757 MHz
(2899.00042272.....MHz). He intends to multiply his S-LO by 3 and that gets
him close, about 297 MHz away. Then he can add another frequency he has(that
is locked to his S-LO) of 241.6..... MHz (2998.01*61/757 MHz to be exact)
and that brings him to about 55 MHz. To generate that 55 MHz he has several
options:
- Cascading two DDS chips to get many bits of frequency resolution and leave
the thing in open loop. I don't like the absence of feedback in this option,
and I have never cascaded DDS chips to achieve an increase in frequency
resolution, although on paper there seems to be no problem. The increase in
accuracy would be such that even in open loop it would take a very very long
time to go out of spec (his spec is 0.01 degrees at X-band, during a whole
year). This time would be long enough for him not to care about it.
- Generating it using a standard PLL with some kind of good quality VCO. He
is concerned by the required VCO quality in this case.
- Combining the two above, i.e. using the DDS as a VCO by controlling its
Frequency Tuning Word, but in a closed loop configuration.

Has anybody out there been confronted to a similar problem? Many thanks in
advance,

Javier