Measuring jitter of pulse generator

[tjkubo]

I have pulse generator that produces negative Gaussian pulses with fwhm~0.3ns. It is driven by the falling edge of a regular square wave from a function generator. I want to make some kind of measurement on an oscilloscope of the jitter of the pulse by comparing the output from the function generator (ch1) and the output from the pulse generator (ch2). The oscilloscope model I'm using is Lecroy WR620Zi and it has a lot of measurement capabilities that I don't really understand. I've tried the delta delay measurement, which measures the "time betw/ the 50% crossing of first transition of two waveforms," so it seems to give the time delay from the falling edge of the square wave to the pulse. It also gives the standard deviation for the measurements. Is this standard deviation the jitter I'm looking for? Am I going about this the wrong way? Any guidance would be appreciated.

 

[yungman]

I never measure jitter before, so I don't have a suggestion. But I don't think what you described will work. The weakest link is the jitter of the scope when it is triggered on the falling edge of the ordinary signal generator. A good pulse generator with very short rise time will help, but you still cannot get rid of the jitter of the scope.

I don't know what is fwhm of the "fwhm~0.3ns" means.

 

[berkeman]

I have pulse generator that produces negative Gaussian pulses with fwhm~0.3ns. It is driven by the falling edge of a regular square wave from a function generator. I want to make some kind of measurement on an oscilloscope of the jitter of the pulse by comparing the output from the function generator (ch1) and the output from the pulse generator (ch2). The oscilloscope model I'm using is Lecroy WR620Zi and it has a lot of measurement capabilities that I don't really understand. I've tried the delta delay measurement, which measures the "time betw/ the 50% crossing of first transition of two waveforms," so it seems to give the time delay from the falling edge of the square wave to the pulse. It also gives the standard deviation for the measurements. Is this standard deviation the jitter I'm looking for? Am I going about this the wrong way? Any guidance would be appreciated.

Usually for measuring jitter or checking eye diagrams, you can turn on "infinite persistance", let it accumulate for a while, and then use time cursors to measure the variation of the jittering signal...

 

[tjkubo]

Thanks for the replies, guys.

yungman, fwhm = full width at half maximum
Also, is it possible to measure the scope's jitter and somehow get the real jitter of the device under test by "subtracting" out the scope's jitter from the total measured jitter.

berkeman, if you don't mimd, can you elaborate on what infinite persistence does?

 

[berkeman]

Thanks for the replies, guys.

yungman, fwhm = full width at half maximum
Also, is it possible to measure the scope's jitter and somehow get the real jitter of the device under test by "subtracting" out the scope's jitter from the total measured jitter.

berkeman, if you don't mimd, can you elaborate on what infinite persistence does?

Infinite persistance means that each new trace does not erase the previous trace. In normal operation, each new 'scope trace erases the previous image. If you do several single-shot traces, for example, each one is separate.

But if you turn on persistance = 4, for example, then the 4 most recent traces are displayed together as you gather traces. If you turn on infinite persistance, then you see all of the traces overlaid on the same screen. You have to manually clear the screen when you want to start over gathering traces.

So, a jittering signal viewed with infinite persistance appears as a number of very similar waveforms, but there will be a spread that corresponds to the jitter. You can measure that spread on the display using the time cursors.

http://www.tek.com/document/competitive/persistence-display

.

 

[yungman]

Problem is you don't know the amount of jitter of the scope.

About the persistance, is a mode on the scope that will remember the trace of the last digitized signal for a set amount of time. The scope trace out ( digitize) the shape of the pulse every time the trigger comes, so if the scope displays and hold the trace on the screen every time the it is trigger, you have a lot of traces displayed at the same time. The variation of each trace show the little difference between each time the pulse is digitized. So if you look at the horizontal and measure the thickness horizontally, that is the time UNCERTAINTY of the signal, which is the jitter.

Say the thickness measuring horizontally at the half way point is 0.1nS, then the jitter is 0.1nS.

BUT, the uncertainty is the sum of the jitter of the scope and your circuit, there is no way to tell unless you have a control pulse generator that has NO jitter or one that is much better than the scope and your circuit. This is not easy to find if you are already looking at a pulse that is only 0.3 nS wide.

Using an eye diagram just make it easier to see, it does not over come the problem of the scope. I heard an engineer in communication talking about a "jitter meter". I have no idea what it is, but look into it a little bit.

 

[berkeman]

At the Tektronix website, they have a product category (look on the Products tab) called Jitter Measurement...

 

[f95toli]

The main problem here is that you need a good timebase. Most good instruments will have a 10MHz input.
Hence, one solution would be to borrow/buy a 10 MHz reference. Alternatively, if you have access to an instrument with very good timebase (say a microwave synth with ovenized crystal) you can just the 10 MHz output of that.

Also, I would personally try to meausure the jitter using a counter (but again ,use an external timebase).

 

[yungman]

The main problem here is that you need a good timebase. Most good instruments will have a 10MHz input.
Hence, one solution would be to borrow/buy a 10 MHz reference. Alternatively, if you have access to an instrument with very good timebase (say a microwave synth with ovenized crystal) you can just the 10 MHz output of that.

Also, I would personally try to meausure the jitter using a counter (but again ,use an external timebase).

I think he is trying to measure sub nano second pulses.

 

[f95toli]

I think he is trying to measure sub nano second pulses.

That should be possible with a good counter (1ns is only 1 GHz, not very fast in modern electronics). The counters I use can typically resolve time intervalls (between ch1 to ch2) down to about 20ps and they are just typical bench-top instruments.
You probably don't even need the external timebase if you are comparing two signals (as opposed to measuring the jitter in a single channel).

 

[yungman]

What is the fastest scope today. I have not been in the field for 7 years. Those day, we had scope that can look at eye pattern of SONET OC48 about 2.5GHz. OP is talking about the pulse width of 0.3nS, for that, rise time has to be not more than 100pS. To measure jitter, you need a scope a lot better than 100pS jitter, 20pS might not do it.