[time-nuts] Comparing PPS from 2 GPS units

Tom Van Baak tvb at LeapSecond.com
Fri Dec 14 06:55:02 UTC 2012


Hi Said,

> the 53132A only does that for a couple of nanoseconds. Then it jumps to a  
> stupid value such as:
> 
> 0.004 us
> -0.002 us
> 0.999,999,993 s

It looks like you were measuring a 1PPS pulse that was very close to the 1PPS reference. As Hal noted when he started this thread, this is tricky.

But the results you're seeing from the 53132A are fine. The RS232 time interval output of these counters is either "s" for seconds or "us" for microseconds. These counters also use thousands separators. It's easy on the eye; trivial to parse too. Or you can use GPIB and get the data in other formats. I would not call it stupid.

What you're seeing in this case is phase wrap. Your small TI numbers just went over the boundary and now your tiny delta is 1.0 seconds minus a tiny delta; hence the 0.999999993 number. It's not the counter's fault. It's also an indication that you have missed one sample. The counter is doing the right thing. See previous posting on how to properly deal with this.

> It get's even better when the counter decides it doesn't have enough  
> resolution in frequency mode:
> 
> 2,3** u

This is actually one of my favorite features of the counter. It shows how experienced the hp engineers who designed it were. The "*" are correct; they are digit precision place holders. This counter, unlike most others, keeps track of resolution and precision and does not report digits that aren't valid.

> Absolutely horrible to parse, these guys should have heard of scientific  
> notation. Not sure who programmed that unit, or if there is a firmware  
> upgrade that gives proper numbers.

They are more proper than you think. Do you remember one of the first lessons in high-school science class: scientific measurements have both value and precision. Thus 2.3 is not the same as 2.30 which is not the same as 2.300. Precision is important. When the 53132A adds "*" it conveys to the user that precision is missing.

/tvb




More information about the time-nuts mailing list