[time-nuts] noise of a 5370

[Mike S]

At 10:45 AM 3/12/2011, Jean-Louis Noel wrote...
>Yes, I saw that!
>But, where is the logical when your display starts at 100fS?

A mean (average) measurement improves things by the square root of the 
number of measurements, if I'm not mistaken. So, if you measure 100,000 
times, then a 35 ps jitter is reduced to ~ 110 fs ( 35 ps / 
sqrt(100000) = 35 ps / ~316 = ~110 ps). Using HPIB, you can do 
measurements up to 2^24-1 = 16777215 times, so 35/4096 = ~9 fs.

There are other error sources which make the actual accuracy worse than 
that, though. For instance, HP gives a +- 1 ns "Sytematic error" for TI 
measurement, regardless of the number of measurements. They say it's 
typical 300 ps and can be reduced to 30 ps. I'm not sure if that is 
random or what consistency it has between different measurements. It 
may only effect absolute accuracy, not relative accuracy when comparing 
measurements.

Someone will tell me if I'm wrong.