RE: [SI-LIST] : Jitter measurement

About this list Date view Thread view Subject view Author view

From: Larry Miller (ldmiller@rhapsodynetworks.com)
Date: Mon Mar 26 2001 - 06:10:57 PST


Tom,

That is not being a contrarian.

That is precisely how a counter-based Time Interval Analyzer (TIA) works.
Basically, it is comprised of a comparator which, when the threshold is
crossed, stores the count ("the time") of a very precise counter in a large
memory. Zillions of these samples are collected (in accordance with the
Zillions of Samples Parameter Dialog) and then can be analyzed for
statistical properties. The Good Old Central Limit Theorem clicks in, and
the measurement uncertainty of any single zero crossing gets swamped out by
the large number of samples averaged.

This is great for measuring long-term (low-frequency) effects like
frequency, drift, and so on, and is the standard tool for synchronous SONET
systems.

However, when you are looking at cycle-to-cycle effects one of two things
happens:

1) you have to decrease the sample population down to the point where
statistics fail you and the sampling uncertainty of individual measurements
comes back into play.

2) In a TIA you get cycle-to-cycle data by essentially taking the derivative
(OK, finite differences) of the time interval data and this greatly
increases the mathematical "noise" in the calculations. Also, the time bases
in counters do not necessarily have great short-term jitter (that's why the
right hand 4 or 5 display digits are a blur); they are mostly designed for
longer interval (0.1s, 1s, 10s) stability and long term frequency accuracy.

The oscilloscope-plus-software packages described heretofore address these
problems and provide much greater accuracy for cycle-to-cycle measurements,
which was the subject of the question. "If the oscilloscope has a lot of
jitter" as you say disqualifies it for measuring, but the ones noted like
the TEK 694 have most excellent time bases (and price tags to match).

Your counter notion is fine (and is a standard product) for longer time
frame effects. One thing a TIA can find is anomalies in PLL-based
oscillators, for example. Many of them have quite a bit of frequency
deviation and limit-cycling down in the audio frequency range.

Larry Miller

-----Original Message-----
From: Thomas Jackson [mailto:tjackson@fmi.fujitsu.com]
Sent: Monday, March 19, 2001 10:39 AM
To: 'Sunil Kumar'; si-list
Subject: RE: [SI-LIST] : Jitter measurement

Let me be a contrarian.

I think the best way to measure oscillator jitter is with a good counter,
ideally a time interval measurement system. The accuracy of most counters
is much, much better than the timebase in oscilloscopes, real-time or
sampling. Make multiple measurements of the period and calculate some
statistics: max, min, mean, sigma, etc.

If you do use an oscilloscope, over the long-term, both should provide
equivalent results, assuming they both have the same timebase accuracy and
stability. If I understand sampling oscilloscopes correctly, they make one
(1) sample per trigger and build up a trace from a number of widely spaced
samples. Real-time oscilloscopes take a number of consecutive samples when
triggered. If the oscillator has a lot of jitter, it seems to me that a
real-time measurement will converge on the measurement faster.

Tom

Thomas L. Jackson, P.E.
Staff VLSI Design Engineer
Network Access Development
Systems Solutions Group
FUJITSU MICROELECTRONICS, INC.
3545 North First Street
San Jose, CA  95134-1804
telephone: (408) 922-9574
facsimile: (408) 922-9618
http://www.fujitsumicro.com

-----Original Message-----
From: Sunil Kumar [mailto:sunilb@cdotb.ernet.in]
Sent: Monday, March 19, 2001 7:05 AM
To: si-list
Subject: [SI-LIST] : Jitter measurement

Hello everybody..

I want to measure cycle-to-cycle jitter generated by a crystal
oscillator. I have two options:
        
        1) Real time oscilloscope
        2) Sampling oscilloscope

The bandwidth of the real time oscilloscope is enough for my
measurements. Both the oscilloscopes are equipped with jitter
measurement packages. Can anybody suggest which method is better?
And why?

Thanks a lot..

Sunil Kumar

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Jun 21 2001 - 10:11:20 PDT