Dialing in oscilloscope to measure rise time of a signal

[newtekuser]

I'm trying to measure the rise time of a square wave of 1kHz coming from the built-in probe compensation terminal of my TDS5054B scope using a P5050 probe and channel set for 1M termination.
I enabled vertical cursors and adjusted my timebase to 200uS so I can fit in the full 1kHz period with the cursors. However, if I set the sampling rate below my scope's full capabilities of 5GS/s, the rise time measurement displays an exclamation mark as can be seen from the first screenshot. It also reads a rise time of 307.5ns, whereas if I bump the sampling rate to max, the exclamation mark disappears and the reading is 4.2ns.

I'd like to verify if I'm doing this right and that the correct reading is of the value done at the max sampling rate. (without the exclamation mark)

[tautech]

Typically Rise/Fall time measurements are accessed from the Measure menu where either can be selected and it is not necessary to display the full period to get measurements of one or the other.

In the old days when we did it by eye we expanded the waveform to fill the display vertically with the edge surpassing the 10 and 90% graticules and used a fast enough timebase to allow edge measurement against a vertical graticule.
Some scopes have graticules specifically designed for edge measurements where some use a different and identifiable graticule at 20 and 80% levels.

Where you are using Cursors in your screenshots, with a fast timebase setting, just the edge could be measured in much the same way.

Many varied examples of how it's done are here:
https://www.eevblog.com/forum/testgear/show-us-your-square-wave/

[David Hess]

I'd like to verify if I'm doing this right and that the correct reading is of the value done at the max sampling rate. (without the exclamation mark)

That is right.  At the lower sample rate there is not enough resolution for the rise time measurement.  There is no need to capture the entire cycle.  Only the edge is required.

[tggzzz]

There is no need to capture the entire cycle.  Only the edge is required.

And if you only capture the edge then you can eyeball it to verify that the waveshape is decent. Typical problem: poor probing technique showing a distorted edge.

[CaptDon]

On my TDS644B it says on the front panel "Adjust probe compensation at 1ms sweep". This will put either the rising or falling edge in the middle of the screen depending on trigger slope. Using this very same technique with the sweep trigger point still centered you could then increase the sweep speed to a couple of microseconds and start to observe the rise time (+ slope trigger) or fall time (- slope trigger). As mentioned in a previous post we observe rise time start as the signal passes through 10% of its total amplitude and we end measurement as it passes through 90%. Hope that makes it more clear for you. Older scopes had the 10% and 90% markings in the graticle whether that was a plastic overlay or printed within the face of the C.R.T. itself.

[David Hess]

Older scopes had the 10% and 90% markings in the graticle whether that was a plastic overlay or printed within the face of the C.R.T. itself.

The examples below clearly show the 10% and 90% horizontal graticule lines used for rise and fall time measurement.  The first example is from my Tektronix 2232 and shows a typical "good" transient response with a rise time of 3.5 nanoseconds which is exactly what the 100 MHz 2232 should have.  The aberrations shown are well within the specifications of the 2232.  The second example is from my 465 which needs some adjustment.  In both cases, the PG506 pulse source is essentially perfect on this scale.

[newtekuser]

Thanks folks! 

What still eludes me is that at different time-bases I get different readings and I don't know which one I should take, is this because the waveform is not stable?
For instance, I increased the timebase to 1ms to capture more pulses and 500MS/s sample rate I get fluctuating rise times between high 4ns up to 9ns. There is sometimes an exclamation mark next to the reading which I assume means the values are bogus.

If I decrease the timebase to 400uS and increase the sample rate to 1.25GS/s, the rise time values show up with the exclamation mark (between 3.6ns up to 7ns sometimes).
If I further increase the sample rate to 2.5GS/s, rise time fluctuates between 3.8ns up to 4.9ns.

If I set a one time edge trigger, I get different rise time values for each reading and they can be as high as 6.9ns or as low as 3.4ns (at 400uS at 2.5GS/s).

I do understand that these snapshots are taken for the signal at that point in time and that they can vary, but if I need to add them into some calculations, what values do I use, do I average them?

[MarkT]

Turn timebase to fastest and measure it on the screen - then you don't have to worry about what (broken?) algorithm the 'scope uses - it may be sensitive to under-shoot or ringing - once you can see the waveform you can verify its a clean rise/fall and consistent too.

[David Hess]

Unless they changed it, the algorithm that Tektronix uses for rise and fall time measurements is very good.  They included descriptions of the various measurement algorithms in the documentation for their old DSOs.

For instance, I increased the timebase to 1ms to capture more pulses and 500MS/s sample rate I get fluctuating rise times between high 4ns up to 9ns. There is sometimes an exclamation mark next to the reading which I assume means the values are bogus.

If I decrease the timebase to 400uS and increase the sample rate to 1.25GS/s, the rise time values show up with the exclamation mark (between 3.6ns up to 7ns sometimes).
If I further increase the sample rate to 2.5GS/s, rise time fluctuates between 3.8ns up to 4.9ns.

If I set a one time edge trigger, I get different rise time values for each reading and they can be as high as 6.9ns or as low as 3.4ns (at 400uS at 2.5GS/s).

At 500 MS/s data points are only every 2 nanoseconds, so a 4 nanosecond rise time will have considerable uncertainty.  Older DSOs which have equivalent time sampling will easily generate 20 GS/s or a point every 50 picoseconds or 80 points over 4 nanoseconds and deliver more consistent and accurate results.

Try using averaging to get more consistent results.  Rise time is not usually something which varies a lot.

[jonpaul]

the scope time/div and signal freq interact according to the sampling theorem.

Read the excellent book from TEK on digital and analog oscilloscopes.

https://download.tek.com/manual/070869001.pdf

Normally several cycles without aliasing need to be on screen for a good auto RT meas.

We just display the edge and use cursors to measure the 10% to 90% RT.

j