[time-nuts] HP 5372A vs. 5370A
SAIDJACK at aol.com
SAIDJACK at aol.com
Wed Feb 9 21:31:33 UTC 2011
Hi Ed,
excellent email. You raise the SNR of this forum!
I have tried to measure the CW-12 output myself with phase noise and Allan
Deviation equipment, and the constant cycle jumps prevent these from giving
sensible results. Both PN and ADEV plots look awful, many orders of
magnitude noisier than even an average oscillator. Glad to know that your 5372A
could get the job done.
btw: jumping cycles 200 times a second is not disciplining an oscillator,
it's numerically controlling a frequency. Thus the CW-12 cannot be
considered a GPSDO due to the massive 8ns phase jumps. But for many apps such as
driving a microprocessor clocks this may be good enough, even though one has
to be careful that their internal PLL's can lock to such a noisy (jumpy)
source. Of course the number of corrections per second should have a very
strong correlation to temperature as the TCXO used on the CW-12 drifts over
temperature.
bye,
Said
In a message dated 2/9/2011 13:14:16 Pacific Standard Time,
ed_palmer at sasktel.net writes:
The CW-12 also has a 10 MHz output. I used the Histogram Time Interval
function to measure the periods of 100 Million cycles with a resolution
of 200ps. It took less than 30 seconds to measure, process, and display
the results. The results showed that there was a normal distribution
around 100 ns and a second normal distribution around 92 ns. The
difference is approximately the period of the internal clock (120 MHz).
This told me that the 10 MHz is kept on frequency by occasionally
shortening the period by one cycle of the internal clock. For my unit,
this happens about 200 times a second. I know this by the ratio of the
number of long periods to short periods. This behaviour explains why
Navsync warns that this output needs to be cleaned up before using it as
a frequency reference.
More information about the Time-nuts_lists.febo.com
mailing list