[time-nuts] NLTL comb generators going away...

John Miles jmiles at pop.net
Fri Dec 28 13:26:45 EST 2007


> I have a nice 3585B, 11729C (with all filters) and a nice working 8662A. I
> assume the PN of the 640 output of the 8662A is still more than adequate.

Right; it doesn't make sense to swap out the 11729's SRD multiplier unless
you are also upgrading the 640-MHz drive source at the same time.

There is some room for improvement in that 640-MHz multiplier chain, though,
and the SRD multiplier appears to be the next weak link after the 640-MHz
chain is cleaned up.

> You are suggesting replacing the multiplier within the 11729C
> with the NLTL
> comb generator instead of the currently used SRD. I do not see how the SRD
> itself would significantly degrade the source's PN beyond the 20 log N
> factor. I will be definitely very interested in your results. 73

The 8662A's 640 MHz output comes from multiplying a 10811A-class oscillator
by 64 (with a 36-dB noise penalty) and shaving off the noise sidebands with
a couple of monolithic crystal filters.   At 1 kHz from the carrier, the PN
spec from the 640-MHz output is -121 dBc/Hz; at 10 kHz it's -145 dBc/Hz.

So, say you're measuring a 10 GHz carrier with your 11729C.  You use the
15th harmonic of 640 MHz for that (24 dB penalty).  Your measurement floor
at 9600 MHz is -97 dBc/Hz (1 kHz) to -121 dBc/Hz (10 kHz).  These levels are
comfortably above the 11729C's residual noise spec at 10 GHz (-132 dBc/Hz
quoted on page 42 of AN 11729B-1).

However, the target specs for my upgraded 8662A reference chain's 640 MHz
output are circa -142 dBc/Hz (1 kHz) to 158 dBc/Hz (10 kHz).  At 9600 MHz
this would be -118 dBc/Hz to -134 dBc/Hz.  My thinking is that the NLTL
multiplier adds some headroom insurance at that point.  According to
Picosecond, the NLTL floor is at or below the thermal limit of their test
setup (-140 dBc/Hz or better).  Their marketing spiel is centered around the
idea that their multiplier will not degrade the PN of the quietest crystal
oscillators available.  That's not true of SRDs, in my understanding.

An exercise in lily-gilding, certainly, but that seems to be the point
around here. :)

-- john, KE5FX




More information about the time-nuts mailing list