[time-nuts] My old new 53132A just arrived

Magnus Danielson magnus at rubidium.dyndns.org
Mon Aug 9 22:18:42 UTC 2010

On 08/09/2010 11:54 PM, Dr. Frank Stellmach wrote:
> Bonsoir Loïc,
> and congratulations for your new, nice device.
> You should order the T.Bolt from fluke.l on e...y, standard easy kit
> with power supply and antenna, just working fine,
> The LCD monitor is also quite handy. You may also order an LPRO 101 Rb
> standard from him, so you have always a very stable reference for a low
> price.
> Alternatively, you may get a receiver for the German DCF77 signal (I can
> see the transmitter every day from my work place), but its a real pain
> for such high precision TICs.
> I've got the predecessor of the 53132A, i.e. the 5335A - uses a similar
> time interpolation circuitry with 1ns LSD / 2ns resolution (jitter),
> giving 9 digits/sec.
> If you examine your TIC/counter, could you please verify the '132A
> counter specs, saying 12digits /sec, although LSD is 150ps, RMS
> resolution 300ps, which should give 10 digits/sec only.
> (My 5370B resolves 11digits/sec , but has 20ps LSD and 35ps RMS
> resolution.)
> I think the agilent people are cheating a little bit (compared to former
> HP).

Uhm... well, sort off...

For the older counters, the performance mapping is fairly 
straight-forward, but for the 53131A/53132A they follow the track laid 
down by J.J. Snyder in his 1980/1981 articles showing how averaging 
several measurements affects noise and this can be used to significantly 
improve the frequency/period readings, but it useless for TI readings 
for Allan dev style of measurements.

So, comparing measurements like you just did id not a representative 
means of comparision. The hardware noise and resolution is aided by a 
software algorithm. It's not cheating for their intended use, but it can 
fool you if you don't know what you are looking at.

This has been discussed extensively in history several times, mostly in 
relationship to Allan deviation measurements, where the approach gives 
incorrect values due to the pre-filtering it does.


More information about the time-nuts mailing list