[time-nuts] HP53132A vs SR625

SAIDJACK at aol.com SAIDJACK at aol.com
Sun Mar 17 17:33:34 EDT 2013

Hi Volker,
there are some issues here, first the worst case  frequency systematic 
uncertainty is 100ps for the 53132A, not 350ps as  on the SRS unit or 500ps as 
you stated. So they are not the same, they are 3.5x  different.
>From the Agilent manual:
Systematic Uncertainty:
Agilent 53131A Agilent 53132A
tacc  tacc
typical 100 ps 10 ps
worst case 300 ps 100 ps 
Notice the 10ps typical error, and 100ps worst case error. That  compares 
to a 100ps typical error for the SR-620 or 10x worse typically than the  
So we get 10x worse typically, and 3.5x less for the worst case -  in my 
opinion these units are not even in the same class.
Now for practical matters, I just measured the SR-620 we have with a  
randomly selected 53132A. Both connected to their own reference input.  2-second 
samples on both, and here are the results:
The SR-620 shows a frequency error of -0.00203Hz consistently.That's  
2.03E-010. Within its specifications but making the unit useless to  me.
The 53132A showed an error of only 2E-012 to 8E-012. So about 25x better  
accuracy! And the 53132A is showing 12 digits on the front panel as well for 
2  second gate times at 10MHz. Nor does it require time-consuming and error 
prone  and annoying internal adjustments to achieve this.
What's even more damming for the SRS unit: as I increased the sample size  
(1s gate time is the max front panel selection, so I had to increase sample 
size  instead of gate time) the error stayed persistent independent of 
sample size or  thus measurement length.
On the HP unit however, increasing the gate time made the error get smaller 
 and smaller, and at 10+ seconds gate time I got 13 digits of resolution 
out of  the unit, and an error of only 1E-012 at that point.

So in summary, the SR-620 requires careful user adjustment of internal  
adjustment points. I don't have time to do that, so sent it in and paid the  
$600+ or so (if I remember correctly) for the standard calibration fee they  
charge. I got a unit back with the error unchanged, which was the original  
reason I sent it in to them in the first place. An error of 2E-010 makes the  
unit useless as we are in need of measuring xE-011 accurately. If I had 
time to  learn how to calibrate the unit myself, I may do so, but even then you 
showed a  2E-011 error on your carefully adjusted unit, whereas I measured 
a 2 to 8E-012  error on a random non-adjusted 53132A unit here. Still about 
3x to 10x  difference in performance.
If someone is interested in a swap of a working 53132A with input-c option  
for our SR-620 I would like to talk to you offline. I would even throw-in  
an FEI Rubidium reference in that swap, even though the SRS' sell for about  
$2300, and the 53132A'a go for about $1400.
In a message dated 3/17/2013 13:02:46 Pacific Daylight Time,  
ailer2 at t-online.de writes:

I  just powered on my SR and looked for the offset, when the 10 MHz 
reference  is connected to the input (at a gate time of 1s without 
further  averaging). It shows an offset of 0 to 400uHz which should 
represent a  mean error of 2E-11, while the manual predicts an error of 
about 1E-10 (as  Said told us, and as my manual tells me). That's within 
the  spec.

Unfortunately I don't have a 53132, but the manual of the HP  predicts an 
error of E-10 - just the same value as with the  SR.

More information about the time-nuts mailing list