Rigol DS1052E - Inaccurate voltage readings

[Cigarsnob]

Hi,

I don't know if it's me or the scope, so any advise?

The problem is is that the scope will not always measure the correct voltage value when I use the measure-voltage function. I have the probes set to X10 and the scope set to X10 and I calibrated the scope. However, I am currently building a bench power supply and when I test the circuit's voltages, say 5 volts at a particular node, the measure-voltage function will display a value much lower then that and the signal on the screen is not correct either (if you were to measure the voltage by reading the screen). This only happens when I have the volts/div set to say...10 volts/div (really anything above the measured voltage value). But when I change it to a lower volts/div setting (5v/div or below for this example), then the measurement comes out correctly. I have also connected my old analog scope with the same volts/div (10v/div) and the measurement on the display is correct. I can even put it to a much higher volts/div setting and still count off the correct value off the screen.

This isn't a major issue, but it's very annoying as I will not always know what voltage I am measuring and may measure a value and think that it's one particular value when it's really much different than then. So what do you think? Any advise? I bought this scope back in July and probably should have posted this question then.

Oh and just in case, the firmware is 00.02.04
Thanks in advance,
Felipe

[Psi]

What measure voltage mode are you using?
Vmax / Vmin / Vpp / Vtop / Vamp / Vavg / Vrms etc..?

With such a high volts per div (10) and trying to measure 5V with the probes reducing the signal by 10 maybe your expecting a little much accuracy wise?


I tried on mine with a power supply voltage of 9.99V
Probes were on 10x

With dso set to measure Vrms I tried v/div of 2V, 5V 10V and 20V.
The range of measured voltages was between 9.92V and 10.2V.

[cybergibbons]

10V/div, 10 divs - 100V full deflection. 8 bit resolution therefore an LSB of resolution is 100/256 which is approx 0.4V. Probably double that for the error you'll get - so make it 0.8V. Is it out by more than that?

[Psi]

I tried it again on my scope (Measuring Vavg)
This time input voltage was 5.00V (according to cheap DMM)
I tried it with both probe on 1x and 10x

This is what i got

1x probe 50v/div = X 20v/div = X 10v/div = 5.4V 5v/div = 4.99V- 4.98V 2v/div = 5.08V 1vdiv = 5.05V

10x probe 50v/div = 6.63V - 6.93V 20v/div = 6.00V - 5.99V 10v/div = 5.73V - 4.71V 5v/div = 5.02V - 5.03V 2v/div = 5.01V 1vdiv = 5.03V

Ya have to remember that most scopes aren't designed to be accurate voltmeters, as cybergibbons said, it's only 8bit resolution.
Can't expect it to be more accurate than 256 steps from bottommost div to topmost div.

Most scopes are only 8 bit vertical, its not like the rigol 1052e's 8bit is low.

[Cigarsnob]

Thank you both Psi and Cybergibbons...My measurements using Vavg, Vrms, and Vmax were all within the 0.8v range when switching to different volts/div. I currently tested a 6 volt supply (measured with DMM) and by using the above voltage modes, the measurements came out to be 6-6.4 volts at 5 volts/div. when I put it at 20 volts/div (just for shits and giggles), I was getting approx. 5.8 volts for Vavg and Vrms and 7.2 volts for Vmax.

So overall, it was my bad and I just need to not set it to some ridiculous volt/div range when measuring a much lower voltage. Thanks again. 

[saturation]

To increase the accuracy to its best capacity, rerun its internal self-calibration now and then.

[Simon]

the problem is that the scope only has 8 bit resolution and will make at least a +/-1bit error, the larger the V/div the bigger the steps in measurement and the less accurate the scope is, it all works on percentage, your looking at a .4% of full scale resolution and an error of 0.4% or more (depends on how good the ADC's are), so the bigger scale you use the more error you introduce, if your full scale is 100V less than half a volt error is acceptable on say 80+V, but if your measuring 5V thats a much higher percentage error of your measurement

[Cigarsnob]

Thanks you guys...I will be sure to run the calibration every now and then and I'll be sure to keep the volts/div within an appropriate range for all measurements to keep the error down.

[Simon]

you will find the same thing on the time scale, if you are taking duty cycle and frequency measurements aim to get 1 or 2 full waveforms on the screen, any more and you start to loose accuracy and at worse the scope refuses to take a measurement

[drZoidberg]

One thing to keep in mind is that the rigol calculates the voltage based on the display, and not the measured voltage... So the resolution is really not 8 bits, it's whatever resolution the display is.

edit:
The same goes for the time scale btw ;p

[Simon]

holy crap that is downright unfortunate, with a 320x 288 resolution that's not a lot to go by. Really Rigol should be ashamed !!!

[Bored@Work]

holy crap that is downright unfortunate, with a 320x 288 resolution that's not a lot to go by. Really Rigol should be ashamed !!!

Guess why they go with a 288 vertical resolution? An 8 bit DAC has 256 bit resolution. 256 pixels, plus a little bit of decoration on the top and bottom, and you have 288 pixels. Enough resolution to display the full data from an 8 bit DAC. Nothing to be ashamed off.

There is also nothing special about not getting every value sampled (and displayed) with the maximum possible resolution. You have fixed vertical measurement ranges (just like with a multimeter). If a value doesn't fit into a particular range, you have to go to the next range, losing resolution.

[Simon]

oh yes of course your right and thinking about it I worked this all out for myself when I first read the specs 

[vl400]

There is also nothing special about not getting every value sampled (and displayed) with the maximum possible resolution. You have fixed vertical measurement ranges (just like with a multimeter). If a value doesn't fit into a particular range, you have to go to the next range, losing resolution.

What about pressing the vertical adjustment knob and use the vernier function? Does this give a perceived increase in resolution or an actual one? It seemed to improve the accuracy when I set a DC voltage that was just higher than one of the 'normal' range settings. The measurement would have normally meant having to switch from 2V/div to 5V/div - instead used 2.5V/div.

[saturation]

Also to add, specified in the user's manual there is a gain error caused by the ADC chip up to 4%.

[Simon]

There is also nothing special about not getting every value sampled (and displayed) with the maximum possible resolution. You have fixed vertical measurement ranges (just like with a multimeter). If a value doesn't fit into a particular range, you have to go to the next range, losing resolution.

What about pressing the vertical adjustment knob and use the vernier function? Does this give a perceived increase in resolution or an actual one? It seemed to improve the accuracy when I set a DC voltage that was just higher than one of the 'normal' range settings. The measurement would have normally meant having to switch from 2V/div to 5V/div - instead used 2.5V/div.

sounds like the voltage ref for the ADC's is adjusted along with the V/Div then, or a resistive divider on the input mirrors the change in V/div

[Franki]

What I would be interested in is whether one can vertically zoom into the (averaged)waveform on the Rigol DS 1052E.

And I am also interested to know whether the waveform avering of the Rigol DS 1052E averages adjacent samples, corresponding samples of adjacent cycles or samples of consecutive measurements.

[saturation]

Its not a 'zoom' like you'd expect on an Agilent or Tek scope, but you can adjust vertical amplitude and timebase on captured or realtime waveforms.  You can use a slow time base on a modulated waveform and freeze the sampling for example, then change to a faster time base to examine the carrier, as best as the memory can represent it.

I don't know the answer your second question, whether Rigol uses one or several methods.  Do you have a reference on what difference this will cause from one method to the other? 

It think Rigol uses adjacent points on the same cycle.  To sample adjacent cycles, or more. causes artifacts on non-repetitive waveforms, one can see this switching between real time and equivalent time sampling.  The net effect of averaging on the Rigol is the waveform still appears pretty much like it is, with much of the noise on the trace turned into a much narrower line.

What I would be interested in is whether one can vertically zoom into the (averaged)waveform on the Rigol DS 1052E.

And I am also interested to know whether the waveform avering of the Rigol DS 1052E averages adjacent samples, corresponding samples of adjacent cycles or samples of consecutive measurements.