Rise time measurement on a scope.

[Dan Moos]

When using my scope's (stock Rigol 1052e) rise time measurement feature, how do I set up my scope so the number is as accurate as it can be?

It seems the faster the time-base setting, the more accurate it gets, but that leads me to a question. Does it just use the portion of the wave that I can see on the screen, or is it using more? In other words, if I'm zoomed in so close in the time domain that the peak of the wave is just off screen, does the scope know this, or does it assume the highest visible point on screen is it? I notice I lose the software frequency measurement when I'm in that close, but I can see why that would be if there isn't a full cycle in memory.

Part of my curiosity comes from discrepancies between having the scope do the measurement, and me doing it manually with cursors.  My manual method (as per W2aew's video) is to alter the vertical scale so that the wave takes up 8 divisions, setting one cursor to the base of the rise, and the other to where the wave crosses the fifth horizontal division (~63%). I then read the difference on the deltaX read out.  ( **EDIT** Ok, I now realize my 63% thing was wrong. Confused the method for determining a time constant with rise time measurements using the 10% to 90% method)

Using this manual method on a 5 kHz, 4v P.P. wave, a signal from a simple oscillator I built, I get about 60 nS rise time. Using the scopes rise time feature, it says closer to 80 nS. Which method is more accurate?

Then, if I zoom in a little more, such that the peak of the wave is just off screen, the scope's measurement gets more inline with my manual measurement. Ok, thats all good, but if I zoom in more, it gets even lower!

So whats the correct procedure?

[T3sl4co1l]

The scope will say (if not on screen, then in the manual) what the measurement is defined as.

Traditionally it's the time taken to go from 10-90% of base to top.  20-80% is also common.

For certain waveforms, a 63% measurement, or any other definition, can give similar, even identical results, but there is no general case where such a measurement can still apply to any other waveform.

In short, if you aren't measuring it the same, don't expect the same results!

Consider this analogy: suppose you need measure the arc length of a circular segment.  Consider three methods:
1. You use a straight ruler to measure the distance from 'start' to 'finish'.  Obviously, this works okay for very small angle arcs, but fails dramatically for large angles.  (If the total angle is 359.99.. degrees, the measurement is ~zero!)
2. You use a straight ruler to measure many very small segments.  You probably want to mark off which points you're measuring between, so you can measure between pairs of points in a consistent manner.  But you'll get a more accurate measurement -- if still a piecewise approximation.  (You also get a ton of error from lining up each set of marks, but nevermind that for now.)
3. You use a tape measure and stretch it along the arc.

Now consider if, instead of a circular path, you need to measure an arbitrary squiggle path.  Method #1 is right out (what if the curve pinches in on itself?  It could be off by 10x or more!).  Method #2 gives different measurements for different numbers of segments -- until you use so many segments that the curve-per-segment is usefully small, anyway.  But that could take a great many points!  #3 is the only reasonable one (but, you can only measure one side of the curve -- if it's a piece of metal that's been bent into a shape, you have to measure one side or the other, or both, and neither measurement may be quite the length of the original piece of metal!).

The scope uses method #2, measuring rise time with thresholds placed at particular levels.  If you use a different threshold, expect different results, depending on how the waveform goes between those points.

To generalize the concept, you might even ask yourself: under what conditions is -- what we would call a smooth rising edge -- even reasonable?  How messy a waveform can you imagine?  What if it has a squiggly or ringing "edge", or it's some lumpy thing that doesn't really rise at all?  Does a sine wave have a rise time -- should it?  It is even useful to speak of?  (No, not usually... but maybe.)

Tim

[tggzzz]

When using my scope's (stock Rigol 1052e) rise time measurement feature, how do I set up my scope so the number is as accurate as it can be?

You will have to define what is important in the measurement you are making. Quite apart from the 10%-90% or 20%-80% question,  a scope is unlikely to get the "right" answer to a typical waveform found in a real circuit using a "normal" probe.

Most waveforms will have some degree of overshoot, ringing, reflections, and other suboptimal characteristics. What "is" the risetime of a signal with 100% overshoot? What "is" the risetime of a signal with a step at 2.5V? (Hint: the answer depends on the receiver).

And then, of course, the probe will change the signal being measured - particularly true if using a simple *10 "high" impedance probe on a digital signal.

I get about 60 nS rise time. Using the scopes rise time feature, it says closer to 80 nS. Which method is more accurate?

Any measurement in nanoSiemens (nS) is, by definition, completely wrong. A measurement in nanoseconds (ns) is more likely to be useful.

[Benta]

Don't forget to calibrate your probes first.

[tggzzz]

Don't forget to calibrate your probes first.

Always do that, every time you connect a probe to a scope!

If you think about it, the standard low frequency probe calibration isn't particularly critical for risetimes less than, say, a microsecond. It is for the amplitude, of course.

[David Hess]

You will have to define what is important in the measurement you are making. Quite apart from the 10%-90% or 20%-80% question,  a scope is unlikely to get the "right" answer to a typical waveform found in a real circuit using a "normal" probe.

I disagree with this.  Rigol DSOs may have problems but older DSOs which use a histogram for transition time measurements and use the acquisition record instead of the display record are accurate to the limits and sometimes beyond of the sampling system.  As you point out however, this is of questionable utility since it is the circuit which matters.

Most waveforms will have some degree of overshoot, ringing, reflections, and other suboptimal characteristics. What "is" the risetime of a signal with 100% overshoot? What "is" the risetime of a signal with a step at 2.5V? (Hint: the answer depends on the receiver).

These imperfections are why we use an oscilloscope where it matters instead of a frequency counter that also has the capability of making rise and fall time measurements.

And then, of course, the probe will change the signal being measured - particularly true if using a simple *10 "high" impedance probe on a digital signal.

There is another confounding characteristics of probes.  Tektronix and HP calibrate their probes differently such that one, and I forget which is which, calibrates their probes to show the signal taking into account the probe loading while the other calibrates to show the result of probe loading.  The difference is subtle enough though that I have never noticed it but people using low-z probes on ECL certainly would.

Probes are almost always characterized with a 25 ohm load which is the result of terminating a 50 ohm test source at the probe tip.  This is of course completely unrealistic for most circuits although not bad for fast logic.  This brings into light why even "low" frequency active probes and sometimes x100 low-z probes are useful down to 100 MHz.  Tektronix even made a "low-z" high impedance passive probe (the 200 MHz P6048) at one time for use where it matters.

I get about 60 nS rise time. Using the scopes rise time feature, it says closer to 80 nS. Which method is more accurate?

Given the known unreliability of the automatic measurements that Rigol DSOs are known for, I suspect your manual measurement is more accurate.  Rigol seems to be in the habit of making measurements on the display record instead of the acquisition record which leads to all kinds of bizarre results.

The same rules however apply to getting the best results.  For automatic measurements, use the fastest time/div and most sensitive volt/div settings which are feasible without allowing the waveform to clip or in the case of Rigol DSOs, exceed the boundaries of the screen.  My DSOs will optionally place markers to show exactly what is being measured which is helpful but I do not know if Rigol supports that.

Does it just use the portion of the wave that I can see on the screen, or is it using more? In other words, if I'm zoomed in so close in the time domain that the peak of the wave is just off screen, does the scope know this, or does it assume the highest visible point on screen is it? I notice I lose the software frequency measurement when I'm in that close, but I can see why that would be if there isn't a full cycle in memory.

The Rigol DS1000Z series of DSOs *only* use the displayed part of the waveform; they make measurements on the display record.  So I assume the DS1000E series does the same.

This has some interesting consequences including averaging and high resolution mode reducing noise but not improving resolution and measurements being made with effectively an 800 point record length or whatever the graticule display width is in pixels.

[tautech]

So whats the correct procedure?

Correct or not this is what works for me: 

Adjust the scope to a reasonably fast timebase, fast enough to at least see some of the risetime. Then adjust the amplitude to have any ringing and/or the start of the rise OUTSIDE (above and below) the risetime graticules (whatever % they are) and if you have a Risetime OSD let the scope do the rest.

Depending on the waveform amplitude you may have to use the Var (DSO Fine) amplitude adjustment if the V/div "steps" are too great.
I've often checked new instruments this way as sometimes they'll better their factory spec by a ns or two.

In order to build trust in your scope, once it's setup as above, adjust the Timebase some more to see how it might affect the measurements.

[danadak]

You might find these useful -

file:///C:/Users/danadak/Downloads/55W_19248_2.pdf


file:///C:/Users/danadak/Downloads/55W-18024-3.pdf


Regards, Dana.

[David Hess]

In order to build trust in your scope, once it's setup as above, adjust the Timebase some more to see how it might affect the measurements.

This is a good sanity check.  On my one DSO which makes automatic measurements, there is no practical difference in the volts/div and time/div over several steps when making these types of measurements; I was shocked to discover that Rigol DSOs had an issue with this.  My analog oscilloscopes which make the same measurements including rise and fall times of course do not care at all about the time/div setting (unless a gated measurement is being made) but that is because their built in hardware frequency counter is really making the measurement in conjunction with the oscilloscope trigger circuits.

I never really thought about this until now but analog oscilloscopes with dual delta delay timebases can make easy but still manual rise and fall time measurements which are only limited by the sweep calibration.  I will have to try that and see what kind of accuracy is really achievable.

[neil t]

My understanding is the standard 10 - 90 % measurement of signal contained within 10 graticules with correct probing technique ie proper grounding  ,also correctly terminated, this then becomes relative to current calibration minus uncertainty somewhere between 0 - 3 %.
Out of three scopes ie siglent dso , tek 2465b , and kikisui6100m, I would guarantee you three differing results at any given point in time. hope this makes sense and helps.