How do I measure jitter in a 50Hz square wave

[Hgspine]

I am wondering how can I measure the edge jitter of a 50Hz square wave? What equipment would I need (with cost containment in mind)? Thanks.

[fourfathom]

I am wondering how can I measure the edge jitter of a 50Hz square wave? What equipment would I need (with cost containment in mind)? Thanks.

Simple peak-peak jitter can be measured with any modern 'scope in the "infinite hold" setting.  What level of precision are you looking for?  It can be a deep subject.

[Hgspine]

Thanks for your reply. The 50Hz square wave that I am interested in is generated by a (FG) winding in a PLL motor.  There are motor drive parameters adjustable within the motor controller to optimize motor speed stability but what I have to work with to assess speed stability (for the purpose of optimizing speed stability adjustments) is the 50Hz square wave.  So the square wave is a feedback wave to the PLL circuitry.  The more stable the speed stability, the less jitter I will see in the 50Hz square wave.

I need to measure jitter (I assume but correct me if I am wrong - EDGE jitter). I need to be able to see the jitter measurement change as I make adjustments within the motor controller.

I have a Tektronix TB2102 DSO. I see in the User Manual how I can adjust persistence from 1 to 10 seconds or infinite. What am I looking for on the waveform on the screen to help me see what the edge jitter is?

Thanks

[fourfathom]

Thanks for your reply. The 50Hz square wave that I am interested in is generated by a (FG) winding in a PLL motor.  There are motor drive parameters adjustable within the motor controller to optimize motor speed stability but what I have to work with to assess speed stability (for the purpose of optimizing speed stability adjustments) is the 50Hz square wave.  So the square wave is a feedback wave to the PLL circuitry.  The more stable the speed stability, the less jitter I will see in the 50Hz square wave.

I need to measure jitter (I assume but correct me if I am wrong - EDGE jitter). I need to be able to see the jitter measurement change as I make adjustments within the motor controller.

I have a Tektronix TB2102 DSO. I see in the User Manual how I can adjust persistence from 1 to 10 seconds or infinite. What am I looking for on the waveform on the screen to help me see what the edge jitter is?

Thanks

The simple way is set trigger to rising-edge, set the timebase to show one complete cycle, then go to infinite persistence.  The jitter will show as a wide smear on the following edges.  This will show cycle-to-cycle jitter.  If you want to see longer-term jitter you can slow down the sweep and show more cycles.  To study lower-frequency characteristics such as "wander" then you need to do more.  Let us know...

[TurboTom]

Get yourself one of the venerable HP 53310A Modulation Domain Analyzers (MDA). These are great, vastly underrated pieces of gear and once in a while quite inexpensively to source. If you also like to cover the frequency range > 200MHz, option 31 (down converter) is favorable over option 30 (prescaler, which is included as well in option 31). If your requirement is only the 50Hz job, even the most basic configuration will be good enough though.

Cheers,
Thomas

[DaJMasta]

Any old frequency counter should do the job just fine, though one that can measure a single pulse is best.  You just measure the signal and you store the result, keep that up for long enough and you have a nice plot you can extrapolate timing noise from.  A scope can visualize it well, but you probably want to use the measurement functions and get statistics based on them if you want to actually quantify it, and while many will do it, it takes a somewhat more capable scope to give you max/min/mean/sdev of a period measurement (though again if you capture the measurement data on a computer, you could do the analysis and plotting stuff as post processing on the PC instead).



Since 50Hz is such low frequency for these sorts of devices, if you're going for the visualization route on the scope, I'd try a slightly different approach from what has been mentioned: trigger off the rising edge (or falling, same concept either way), zoom in quite far on it, then delay from the trigger to visualize the falling edge, with infinite persistence on to get your smear.  Basically, if you can fit an entire period of your 50Hz signal on the screen, the jitter is likely to be to small that it won't appear as a smear unless you're pretty far zoomed in on your timebase.  I assume a ms of jitter, which would basically be required to be significantly visible on a scope with like a 2ms/div timebase, may be too large for your PLL to even lock, though this may not be the case.

[fourfathom]

Yes, delayed sweep, and smart scope or counter measurement functions are a good way to enhance your measurement capabilities.

It would help if you told us the likely range of jitter you expect, and perhaps the PLL loop-filter bandwidth characteristics.  Do you have jitter characteristics that you are trying to meet?

[Hgspine]

Any old frequency counter should do the job just fine, though one that can measure a single pulse is best.  You just measure the signal and you store the result, keep that up for long enough and you have a nice plot you can extrapolate timing noise from.  A scope can visualize it well, but you probably want to use the measurement functions and get statistics based on them if you want to actually quantify it, and while many will do it, it takes a somewhat more capable scope to give you max/min/mean/sdev of a period measurement (though again if you capture the measurement data on a computer, you could do the analysis and plotting stuff as post processing on the PC instead).
--I have a Siglent Technologies SDM3055 5.5 Digit Digital Multimeter, which can measure max/min/sdev of frequency. I can try that.


Since 50Hz is such low frequency for these sorts of devices, if you're going for the visualization route on the scope, I'd try a slightly different approach from what has been mentioned: trigger off the rising edge (or falling, same concept either way), zoom in quite far on it, then delay from the trigger to visualize the falling edge, with infinite persistence on to get your smear.  Basically, if you can fit an entire period of your 50Hz signal on the screen, the jitter is likely to be to small that it won't appear as a smear unless you're pretty far zoomed in on your timebase.  I assume a ms of jitter, which would basically be required to be significantly visible on a scope with like a 2ms/div timebase, may be too large for your PLL to even lock, though this may not be the case.
-- I have a Tektronix TBS 2102 DSO.  I think it has a zoom function.  When you say smear do you mean the width of the leading or falling edge of the square wave?

Maybe I should use both the DMM and the DSO at the same time to get the best indication of frequency stability?

Thanks

[Hgspine]

Sounds like a good piece of equipment to try for this purpose. I am waiting for a good one for a good price to show up on eBay.

[DaJMasta]

-- I have a Tektronix TBS 2102 DSO.  I think it has a zoom function.  When you say smear do you mean the width of the leading or falling edge of the square wave?

The one you're not triggering it on.  The idea is that the trigger will be lower jitter than your device, so you trigger on one (let's say rising) then you time delay over to the falling edge and zoom in to see the jitter (which manifests as different pulse width).  If for some reason your architecture has a fixed pulse width output, you should be looking at the width of the trough between pulses, so trigger on the falling edge and look at the next rising edge to see any movement.  If you're using the scope's measurement functionality you can probably just do period measurements, but again in a zoomed out timebase you may not see a lot of detail.


Not sure if the DMM would be a great choice, but since the signal is slow it could be worth a shot.  Depends on the resolution it can give you on frequency - so 50.0Hz wouldn't be useful but 50.000Hz or better resolution could maybe show some jitter.  Most dedicated frequency counters can manage 5-6 digits on a single pulse measurement or <1s gate time measurement, though measuring multiple pulses at once will hide some jitter, since you'd be averaging multiple periods together in the measurement.  There's also a chance that the DMM's counter can only operate in the averaged mode, which would give you an idea of overall stability or deterministic jitter, but not random jitter.


EDIT: it occurred to me that the option that makes the most sense would be measuring from rising edge to rising edge or falling to falling, since jitter is the difference between cycles, but using rising to falling or falling to rising may allow you to zoom in farther on your scope since the timebase delay has a limit that's locked to the timebase and memory depth.  Any of those will let you measure the performance of the PLL, but if you were comparing your findings with a spec or something, it would almost always be measured over full cycles.