Accurate Measurements on pulsed system's harder than you think.

[TinselKoala]

Yes, the scope is awesome. I've often said that the Lathe is the King of Tools, and the Oscilloscope is the King of Test Equipment.

However.... the low-end scopes are not especially good voltmeters. Most of them have 8-bit ADC front-ends, which means you only have 256 voltage levels across the full height of the screen in the best of circumstances and usually even fewer than that. So the precision is not that good. And as far as accuracy is concerned, they are subject to drifts due to ambient temperature, component ageing, and who knows what else.

I've just spent some time checking my scope, comparing manual calculations from the screen data with what the scope comes up with. The three screenshots below show what I've been doing. I used one data capture from the "Woopy" circuit, and used cursors to read off the heights of the CH1, CH2 and Math traces to get values for the manual calculations. And I told the scope to provide Measurements of the Duty Cycle, the CH1 and CH2 RMS, the Math and CH2 Averages.

Here are the results.

[TinselKoala]

So as you can see the cursors in the CH1 test are showing 12.30 V for the peak values. Multiplying the Peak by the square root of the 19.44 percent Duty Cycle gives
12.30 x sqrt(.1944) = 5.42 V_rms, and the scope is reporting 5.40 V_rms. Not too bad, an error of less than 1 percent.

In the second test with the CH2 readings... the cursors are giving me a value of 128.0 mA for the peak values. Multiplying this by the square root of the Duty Cycle gives 0.1280 x sqrt(.1944) = 0.0564 or 56.4 mA _rms, and the scope is reporting 63.1 mA _rms. WTF? This is a large error. And for the Mean value, the 0.1280 x .1944 = 24.9 mA, whereas the scope says 30.8 mA. Again, a large error. WTF? Is this due to an inaccurate channel, or bad calibration, bad cursor positioning, or what? Perhaps the channel is just not accurate at very sensitive settings?

In the third test with the Math readings, the cursors give me 1.540 W for the peak value, and factoring in the duty cycle gives 1.540 x 0.1944 = 299.4 mW as the average. And the scope is reporting 305 mW as the average. Again not too bad, about 2 percent error.

These three shots are using the same sample set, I have stopped the scope and am just moving cursors around between the screensaves.

The poor result on CH2 has made me start the Self-Calibrate routine; maybe it will get better, or maybe not.

[TinselKoala]

Well, I ran the self-cal routine and the CH2 discrepancy did go down, from about 11 percent to around 8 percent. The discrepancies on the other two readings went up slightly to about 2 1/2 percent difference each. I've noticed that the cursor position values seem a little off from what I think they should be, so maybe the error is in the reported cursor positions rather than in the scope's computations themselves.

Yes, I just checked the cursor positions carefully. There is definitely an error in the reported cursor positions.

In the screenshot below, the CH1 marker is exactly on the screen center horizontal marker (offset 0.000). The Cursor B is positioned exactly on this line and reads -100.0 mV. The Cursor A is exactly on the second horizontal line below center and should read -10.00 V but actually reads -10.20. There is a similar error on other channels.

Grrr. I'm reporting this as a bug to Rigol.

[EMJunkie]

TK - A nice example of how we can see some inaccurate measurements on the Oscilloscope. Thanks for sharing this.

What's the Channel like at higher voltage levels? Does this error margin come down some?

   Chris Sykes
       hyiq.org

[TinselKoala]

TK - A nice example of how we can see some inaccurate measurements on the Oscilloscope. Thanks for sharing this.

What's the Channel like at higher voltage levels? Does this error margin come down some?

   Chris Sykes
       hyiq.org

The error in the cursor position reporting seems to be 1 or 2 ADC counts or maybe vertical pixels at most, regardless of the v/div settings, but depending on where the cursor is positioned vertically on the screen.  It's annoying but like many errors in measuring instruments (they all have them) it can be worked around once it is known that it exists and how it behaves.

As far as I can tell the Measurements from the measure menus aren't subject to this kind of error (which is why my computations using the Cursor data didn't agree with the Measurements), but of course they too are limited by the resolution of the 8-bit ADC front end of this and all other low-end DSOs. You could think of this ADC resolution limitation as "rounding error".