Why is HP 5370/5345 Time Interval Measuring less accurate trhan Period Measuring

Jim,

Thanks for response.

For unknown reason I am unable to respond directly to this group's postings the last few days. Sorry for my failed attempt ( rubbish in subject line ) couple a minutes ago. I try in this way again now.....

Basically you're right. However, my problem is that, even when I average the TI mode ( x 100 , x 1k , x 10 k etc ) the stability resolution ( though considerably improving compared to single shot )lags remarkably behind the Frequency and Period Modes ( which modes, as said, perform outstanding with better than e-11 ).

E.g. : while averaging a 100 Khz signal ( e-11 locked to frequency standard ) in the TI mode 10k ( = gate time = 4 sec ) the observed

5370A jitter is +/- 160 pS, which is not really too bad but for one reason or another this is far worse than a 1 Sec gated Period measurement ( differs > 4 magnitudes but, since the absolute 5370A jitter is > 20 pS, this magnitude comparison doesn't make sense ).

BTW : Measuring 10 Mhz and above, the 5370A jitters only ( even in single shot mode ) only 20-40 pS which is perfectly within the specs.

So, I still wonder what makes the TI measurement mode differently from the Period measurent mode ? I was so naive to presume that both ( TI and PERIOD modes ) were counting during gate open time in the same way, only from/to different point on the waveform.

Is my guess correct that the ( 5370A ) TI and Periode counting systems are NOT based on the very same mechanism ?

Thanks for any further comments.

Peter van Daalen. PE1E ex PE1ECM

I'm really not familiar with your instrument, so I don't quite know how to respond to this. I THOUGHT that a TI measurement was specifically a single event measurement, so that things like gate time and the like would be irrelevant.

Assuming that the 5370 actually DOES allow the option to concatenate TI measurements, wouldn't 10k of a 100kHz signal actually amount to

0.1 sec? Or is there a typo there?

My assumption has always been that a TI measurement was an inherently single event measurement, meant for non-repeating events, while a period measurement was for repeating events so one could take advantage of averages over large numbers.

I guess it's possible that the 5370 allows you to make a "series" of individual TI measurements and then take the average of these measurements. This would give you a better result than a single measurement, but still not as good as a period measurement (where there is the implicit assumption that each period is the same.)

If I try to think about this too hard, I think it will start to be painful, but my guess is that the precision of the former will go as

1/sqrt(N) while that of the latter will go as 1/N.

If they give different results, that must be the case.

I've found that the HP manuals are generally pretty good about explaining these differences. Do you have the manual? If not, you could try to get it. You could also see if HP (now Agilent) still has copies of their application notes #200 (a, b, c, & d.) These explain a LOT of the precision limitations in frequency measurements and make for facinating reading for any one interested in time and frequency measurements.

----------------------------------------------- Jim Adney snipped-for-privacy@vwtype3.org Madison, WI 53711 USA

-----------------------------------------------