High pulse width measurement inaccuracy

[ben@willcockses.com]

I have a circuit which generates an accurate pulse 500ms wide. I use the PicoScope 6 measurement function to measure the high pulse width between rulers. When the timebase is set to 100ms/div, I get sensible results:

Min=500.1ms; Max=500.1ms; Average=500.1ms; sigma=24.46ns

If I change the timebase to 200ms/div (and move the rulers accordingly), I get rubbish results:

Min=483.4ms; Max=498.6ms; Average=497ms; sigma=4.245ms

Am I doing something wrong, or is this a serious bug? Screen shots attached.

[Martyn]

Can you go to Tools->Preferences->Sampling and change the Slow Sampling Transition time to 500ms/div, or slower, and repeat the test.

[bennog]

There is an enourmus difference in sampling speed.

2500 kS / sec in the first image.
12.27 kS in the second image.

this is a factor of 200 so the result will be a factor 200 worse.

Benno

[ben@willcockses.com]

Bennog,

Thanks for pointing this out, but I was already aware of it. The thing is, I'm seeing good accuracy at the lower sampling rate, and lousy accuracy at the higher sampling rate.

At 12.27KS/sec you'd expect to measure a 500ms interval to about 0.02% accuracy. At 2.5MS/sec, you'd obviously expect better accuracy, not worse.

Rgds,

Ben

[Martyn]

If you make the change I suggested does the accuracy improve ?

[bennog]

In the following topic is explained how it is calculated,

topic28921.html

Maybe this explains the differences you see.

[ben@willcockses.com]

Martyn,

Yes, if I change the slow sampling transition to 500ms/div, then the measurements are good at 200ms/div. So it appears there is a bug which only affects slow sampling mode - do you agree?

Thanks,

Ben

[ben@willcockses.com]

Bennog,

Thank you for suggesting that topic, but no, it quite obviously does not explain the differences I am seeing.

Ben

[Martyn]

Thank you, that was the information I needed.

It does appear that when using the slow sampling mode the downsampled data that is used by the measurements engine is introducing inaccuracies, compared with the downsampled data retrieved directly from the scope when using faster timebases. I will pass the information on to the development team.

Can you send the text from Help->About, with the scope attached, so that we know the versions of S/W, F/W and H/W you have.

[ben@willcockses.com]

Martyn,

PicoScope® 6 - PC Oscilloscope software version: 6.12.9.2917
Copyright © 1995-2017, Pico Technology Ltd.

Model: PicoScope 2205 MSO
Serial Number: CZ114/032
USB Version: 2.0
Calibration Date: 30 May 2014
Hardware Version: 1
Driver Version: 1.3.0.14
Firmware Version: 1.3.3.0 / 0.2.48.0

Thanks,

Ben

[Martyn]

Thank you for the information. I tried with an identical setup and am seeing stable high pulse width values, with the same values in both modes.

When replicating the same sampling intervals as you, I did notice that the digital channels are enabled but not showing in your screenshots. Does the issue disappear if they are turned off.

Would it be possible for you to post the two psdata files, corresponding to the two pictures.

[ben@willcockses.com]

Martyn,

I re-ran it at 100ms/div (m_100ms_div.psdata) and 200ms/div (m_200ms_div.psdata). As before, the results were correct for 100ms/div but wrong for 200ms/div.

I turned off the digital channels and tried again at 200ms/div (m_200ms_div_digital_off.psdata). This does seem to fix the problem.

I've also attached settings.pssettings which will hopefully allow you to recreate the problem.

Thanks,

Ben

[ben@willcockses.com]

I wasn't able to attach the files as they were rejected as too large (sigh). So I have put them on Google drive:

https://drive.google.com/open?id=1Jleod ... taFL-hPHPx

Rgds,

Ben

[ben@willcockses.com]

Oops, that last link probably wasn't public... try this one:

https://drive.google.com/drive/folders/ ... sp=sharing

[bennog]

In the psdata with digital on you can see the timing difference in the visual representation also.
If you switch between sample 2 and 3.

I suppose the software is missing samples without detecting it is missing samples.

Normally the software will notice you is the PC can't keep up with the scope sending data.

have you tried running the test on a different (faster) pc ?

Benno