[time-nuts] HP 5328A Divider / Timebase Output performance

Bruce Griffiths bruce.griffiths at xtra.co.nz
Wed Mar 17 20:14:42 UTC 2010


Bert, VE2ZAZ wrote:
> OK, I am doing a 24-hour ADEV analysis of the HP 5328A Timebase Output when set to divide-by-10,000,000. I use the 10MHz output from an HP 8644A sig.gen. OCXO. I split that 10MHz signal into two; one end goes to the the input of the HP 5328A counter. The other 10MHz end goes to the Stop input of the 5370A. The Start input gets the 5328A Timebase Output PPS. The 5370A is free running. I have included the plot of what I get so far.
>
> I would tend to say that the divider is pretty lousy for short term, but it is all fine for longer runs, right?
>
> Is this what I should expect from a TTL/ECL divider chain designed in the '70s-'80s? How would this compare to a modern divider chain, like the PIC divider or David Partridge's divider board?
>
> Since I get a straight line pretty much up to 500s or so, do I conclude that the divider dominates the system noise up to there?
>
> Of course, the 5370A timebase drift has to be taken into account but is not subtracted on that plot.
>
> Thanks,
>
> Bert.
>
>    
Bert

Trigger jitter in the 5328 input?
Trigger jitter at the 5370A input?
What are signal levels at the 5328 and 5370A inputs?
What were the input attenuator settings on the 5370A and the 5328A?
What were the input impedance settings for the 5328A and 5370A?

A well designed divider with no input trigger jitter issues will have an 
output jitter well below the 5370A noise floor when dividing down a low 
noise source.
Typical cycle to cycle jitter of  ACMOs logic is ~1ps rms.
Typical cycle to cycle jitter of  HCMOs logic is ~4ps rms.
Typical cycle to cycle jitter of  ECL logic is <1ps rms.
However using a clean well filtered power supply for the critical parts 
of the divider is advisable.

Even the noise level exhibited by the 5328 divider is well below that of 
your GPS receiver.

When comparing one low noise frequency standard against another you 
using the divider and the 5370A you can significantly reduce the effect 
of the divider noise by

1) Connect the divider output to the EXT arming input

2) Connect the divider input signal to the START input

3) Connect the frequency to be compared to the STOP input

4) Select EXT ARM operation.

The divider output then merely selects the threshold crossing of its 
input signal to start the 5370A time interval measurement.
As long as the jitter is small compared with the input signal period 
this should work well.

Bruce




More information about the time-nuts mailing list