Author Topic: Are scopes expected to be highly accurate on DC measurements?  (Read 1397 times)

0 Members and 1 Guest are viewing this topic.

Offline Dundarave

  • Regular Contributor
  • *
  • Posts: 50
  • Country: ca
I'm just getting back into electronics after many years in networks & software, and have recently bought an SDS1104X-E scope, a 121GW meter, and a few other pieces of entry-level gear.  Yesterday a cheapie AD584KH voltage reference arrived, and naturally I set about comparing the accuracy of anything in sight that could possibly measure 2.5, 5.0, 7.5 or 10.0 volts.

The new 121GW meter seemed to prove that the AD584 was worth the $25 I paid for it:  2.5001, 5.0007, 7.501, 10.000 ...  I was impressed by both pieces of gear.

However, when I turned my attention to the scope, I realized things weren't so simple.  The SDS1104X-E (with the 200M key liberated & installed) couldn't give me a clear answer on what it thought of the voltage reference.  It offers up many metrics:  p-p, mean, min/max, rms, top, base, etc. , but only RMS seemed to come close to what I was feeding into it.  Max, min, top, base, amplitude, etc. were all off by significant margins.

But even then, the RMS parameter showed readings of 2.40, 5.00, 7.44, 9.98 (i.e. the "RMS" parameter, all from the "Measure" screen)  Only the 5.00 was what I considered close, with the 9.98 being "close enough".  Different channels (same probe) showed different results, essentially all around 0.5 volt or so out.

I considered that I was likely measuring some AC noise along with it, so I tried a battery that the 121GW measured at 4.4319, and the scope showed 4.40 VRMS for the channel under test.  Channel 2 showed 4.76 VRMS for the same battery source.  Ch 3 was 4.44 and Ch4 was 4.43, essentially dead on.

I checked the specs, but wasn't sure which ones were relevant (e.g. dc gain accuracy, dc offset accuracy?) so I thought I'd ask:  Is a +/-  0.5 volt RMS variance in readings to be expected on a new entry-level scope?

I'm not terribly worried about it, as measuring DC voltage is the job of my 121GW, but as we all know, "a man's got to know his <test equipment's> limitations".

Thanks -
 

Offline tautech

  • Super Contributor
  • ***
  • Posts: 14582
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #1 on: January 24, 2019, 08:36:59 am »
I'm just getting back into electronics after many years in networks & software, and have recently bought an SDS1104X-E scope, a 121GW meter, and a few other pieces of entry-level gear.  Yesterday a cheapie AD584KH voltage reference arrived, and naturally I set about comparing the accuracy of anything in sight that could possibly measure 2.5, 5.0, 7.5 or 10.0 volts.

The new 121GW meter seemed to prove that the AD584 was worth the $25 I paid for it:  2.5001, 5.0007, 7.501, 10.000 ...  I was impressed by both pieces of gear.

However, when I turned my attention to the scope, I realized things weren't so simple.  The SDS1104X-E (with the 200M key liberated & installed) couldn't give me a clear answer on what it thought of the voltage reference.  It offers up many metrics:  p-p, mean, min/max, rms, top, base, etc. , but only RMS seemed to come close to what I was feeding into it.  Max, min, top, base, amplitude, etc. were all off by significant margins.

But even then, the RMS parameter showed readings of 2.40, 5.00, 7.44, 9.98 (i.e. the "RMS" parameter, all from the "Measure" screen)  Only the 5.00 was what I considered close, with the 9.98 being "close enough".  Different channels (same probe) showed different results, essentially all around 0.5 volt or so out.

I considered that I was likely measuring some AC noise along with it, so I tried a battery that the 121GW measured at 4.4319, and the scope showed 4.40 VRMS for the channel under test.  Channel 2 showed 4.76 VRMS for the same battery source.  Ch 3 was 4.44 and Ch4 was 4.43, essentially dead on.

I checked the specs, but wasn't sure which ones were relevant (e.g. dc gain accuracy, dc offset accuracy?) so I thought I'd ask:  Is a +/-  0.5 volt RMS variance in readings to be expected on a new entry-level scope?

I'm not terribly worried about it, as measuring DC voltage is the job of my 121GW, but as we all know, "a man's got to know his <test equipment's> limitations".

Thanks -
Highlight passages need more attention paid to.
Some further good info in this thread:
https://www.eevblog.com/forum/testgear/testing-dso-auto-measurements-accuracy-across-timebases/
Avid Rabid Hobbyist
 

Online Specmaster

  • Super Contributor
  • ***
  • Posts: 6617
  • Country: gb
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #2 on: January 24, 2019, 08:49:48 am »
Personally I would always go with the results from a decent DMM over those from a digital oscilloscope because the meters only function in life is to provide the user with a high precision reading of a voltage and it is doing this even when in resistance and or current ranges etc. But an oscilloscopes main function is to present to you a visual representation of voltage over a time period and hence draws a waveform, or a pictorial view of what is happening with the voltage over time and therefore is more biased to to the pictorial element and any readouts of frequency. volts etc are never going to be as accurate as piece of kit dedicated to a particular function.
Who let Murphy in?
 

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 4155
  • Country: gb
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #3 on: January 24, 2019, 08:50:31 am »
Your SDS1104X-E, in common with most standard DSOs, has an 8 bit vertical resolution, that's 256 discrete levels - that's assuming that you can scale the input voltage to make use of the full ADC dynamic range.

Accurate voltage measurement, in voltage reference context, isn't something you're going to get.
Chris

"Victor Meldrew, the Crimson Avenger!"
 

Offline Dundarave

  • Regular Contributor
  • *
  • Posts: 50
  • Country: ca
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #4 on: January 24, 2019, 11:14:23 am »
Thanks Gyro, and the others who replied.

Gyro's point regarding the 8 bit ADC resolution makes it very clear that that DSO's aren't designed for great DC voltage precision.  Reading other threads, I noted a reference to ±3% being the usual voltage accuracy on such scopes, and indeed, the DC Gain Accuracy for my scope is given as ≤±3.0%: 5 mV/div-10 V/div and ≤±4.0%: ≤2 mV/div.  (I presume that's the relevant spec, as Tautech highlighted in his earlier reply.)

So that gives me a good grasp of the DC accuracy of the scope, thanks again.
 

Online Specmaster

  • Super Contributor
  • ***
  • Posts: 6617
  • Country: gb
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #5 on: January 24, 2019, 11:59:26 am »
Thanks Gyro, and the others who replied.

Gyro's point regarding the 8 bit ADC resolution makes it very clear that that DSO's aren't designed for great DC voltage precision.  Reading other threads, I noted a reference to ±3% being the usual voltage accuracy on such scopes, and indeed, the DC Gain Accuracy for my scope is given as ≤±3.0%: 5 mV/div-10 V/div and ≤±4.0%: ≤2 mV/div.  (I presume that's the relevant spec, as Tautech highlighted in his earlier reply.)

So that gives me a good grasp of the DC accuracy of the scope, thanks again.
Yes thats correct, and given that most of the better brand meters are around 0.1 to 0.03% accurate you can begin to see just how far out a scope can be, but its good enough for a guide, but if absolute accuracy is required, use a meter. 
Who let Murphy in?
 

Online pigrew

  • Frequent Contributor
  • **
  • Posts: 445
  • Country: us
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #6 on: January 24, 2019, 12:26:41 pm »
No, oscilloscopes are not expected to be highly accurate for DC measurements, but they should be fairly stable over time (if the temperature doesn't change much).

Usually I try to avoid oscilloscopes for high-accuracy measurements, but they are quite useful when used within their limits. Their time-bases are usually quite accurate, but the voltage amplifiers usually have significant offset (even before thinking about AC effects). Also note that with a 10X probe, it's making a voltage divider with the probe's series resistance and the scope's input resistance, so there are also probe to probe variations. The input impedance will also vary channel to channel, yielding even more accuracy errors to account for.

As a rule of thumb, I only use an oscilloscope for numerical measurements after calibrating it in that particular range, at the particular channel, with a particular probe, and only use that for a short period of time while the temperature has been stable. Even then, I'll want to do the calibration at the exact frequency I intend on measuring (be it DC or AC). I then correct the measurement data based on the calibration.

Some oscilloscopes have a higher number of bits in their ADC yielding a larger dynamic range. This greatly simplifies calibration since fewer ranges will need to be measured. I've recently worked on a project using the PicoScope 4262, with a 16-bit ADC and 0.25% DC accuracy in the upper ranges. With it, I'm relying on the linearity of the ADC.

Your SDS1104X-E has a DC accuracy of "±(|offset|*1%+3 mV)", so 3 mV when the offset is set to zero. The input resistance "(1 MΩ±2%) || (18 pF ±2 pF)", so that adds a few more percent error between channels if you are using 10X probes.

Noise has already been mentioned in this thread. Most oscilloscopes have an 8-bit ADC with a bunch of noise added to get about 6.5 ENOB (effective bits). The noise lets the mean measured value be more accurate than a single sample (this is called dithering). You'd expect each individual sample to be wrong by about 1% of the full-scale scale because of this 6.5 ENOB.

I believe the above accuracy specs are assuming you have already done a bunch of averaging (getting rid of the high-frequency noise). You should use the "mean" function for measuring DC values, not the RMS function (which will skew your result a bit when the input voltage is non-zero). Also, the high-res mode could be a good thing to use as it'll be able to average more samples in a given time period.
 

Offline Rerouter

  • Super Contributor
  • ***
  • Posts: 3969
  • Country: au
  • Question Everything... Except This Statement
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #7 on: January 24, 2019, 12:35:12 pm »
For the sds1204x-e the way to get the most accuracy for a DC voltageis to use the vertical offset on the lowest v/div that covers your signal.  The linearity of the offset DAC used for each channel is quite accurate. And this lets you work around the ADC only being 8 bit
 
The following users thanked this post: Richard Crowley

Offline Performa01

  • Frequent Contributor
  • **
  • Posts: 710
  • Country: at
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #8 on: January 24, 2019, 03:09:40 pm »
However, when I turned my attention to the scope, I realized things weren't so simple.  The SDS1104X-E (with the 200M key liberated & installed) couldn't give me a clear answer on what it thought of the voltage reference.  It offers up many metrics:  p-p, mean, min/max, rms, top, base, etc. , but only RMS seemed to come close to what I was feeding into it.  Max, min, top, base, amplitude, etc. were all off by significant margins.
For DC, there shouldn't be a significant difference between these metrics, except for minor noise effects.

However you need to keep in mind that errors can be expected to be an order of magnitude higher in a scope than an average DMM.

But even then, the RMS parameter showed readings of 2.40, 5.00, 7.44, 9.98 (i.e. the "RMS" parameter, all from the "Measure" screen)  Only the 5.00 was what I considered close, with the 9.98 being "close enough".  Different channels (same probe) showed different results, essentially all around 0.5 volt or so out.

I considered that I was likely measuring some AC noise along with it, so I tried a battery that the 121GW measured at 4.4319, and the scope showed 4.40 VRMS for the channel under test.  Channel 2 showed 4.76 VRMS for the same battery source.  Ch 3 was 4.44 and Ch4 was 4.43, essentially dead on.

I checked the specs, but wasn't sure which ones were relevant (e.g. dc gain accuracy, dc offset accuracy?) so I thought I'd ask:  Is a +/-  0.5 volt RMS variance in readings to be expected on a new entry-level scope?
No, an error of 0.5V (equivalent to >10%) is not to be expected. The actual accuracy of an SDS1000X-E, correctly tested under lab conditions, is usually significantly better than specified, e.g. <1% gain error.

All in all, I would expect measurements to be somewhere between 4.4 and 4.47V in your scenario, which appears to be the case except for channel 2.

Try the following:
1. Make sure the scope has been running for at least 30 minutes.
2. Disconnect everything from the channel inputs.
3. Run a self calibration.
4. Repeat DC accuracy tests with direct coax connections (or at least probe in 1:1 mode).
5. For correct and stable results, turn the measurement statistics on and use the mean values over at least 60 acquisitions for the rms or mean measurements.
6. To measure the DC component of noisy signals, Average acquisition mode is highly recommended. The effect is huge, even with just 4 or 16 averages. It is much better suited than Eres, because Average also kills the low frequency noise (which usually is the true problem for DC measurements).

All channels should be well within specifications, i.e. the errors should be far less than 3%. At voltages that high, the offset error usually isn't a significant contributor anyway.

If the result is within spec, then you can repeat the test with the 10:1 probe. If Channel 2 is off again, then there might be something wrong with its input impedance.

Finally I happen to have a screenshot (see attachment) that demonstrates the point made by Rerouter, i.e. the fairly accurate measurement of 205V DC utilizing the vertical offset control. Measurement error is less than 0.1% here.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9043
  • Country: us
  • DavidH
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #9 on: January 27, 2019, 06:42:09 am »
Oscilloscopes have limited vertical accuracy for several reasons:

1. Limited resolution.
2. Limited accuracy of the input attenuators.
3. Limited accuracy of any attenuating probe if used.  Some probes have attenuation adjustments.

The exception to this is oscilloscopes which can be configured for slideback measurements where the oscilloscope operates as a null detector.  When this is done, oscilloscope accuracy becomes irrelevant and measurements can be made to high precision using a voltage reference.  The old Tektronix 7A13 differential comparator works this way.  A modern implementation is available in the various LeCroy differential amplifiers although it appears they will shortly be discontinued.  Some DSOs have a limited ability to do this with their vertical offset feature.
 

Online ArthurDent

  • Frequent Contributor
  • **
  • Posts: 663
  • Country: us
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #10 on: January 27, 2019, 08:57:26 am »
Performa01 is correct that on D.C. the metric used isn't important because the peak=average=mean, and so on. The link below shows how the different methods compare for D.C. and various other waveforms and shows the wave shape they describe. No one method of representing voltage is better or worse than the others but which you choose is determined by what you're interested in.

With an incandescent lamp powered by A.C., the R.M.S. value of the A.C. gives the same heating effect as the identical D.C. value powering the same lamp. With a narrow pulse you might be interested in the peak value since a high peak like a lightning spike on your power line might cause component failure.

https://meettechniek.info/compendium/average-effective.html   
 

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 5240
  • Country: us
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #11 on: January 27, 2019, 10:13:01 am »
Because you are looking at a DC signal (super slow) and the scope will sample many times faster, if you have a way to program the scope, you could possible gain some resolution and accuracy.  I went through a simple exorcise like this.    Even with my efforts, I would never suggest the scope would outperform my meter.   Two different tools.

How electrically robust is your meter?? https://www.youtube.com/channel/UCsK99WXk9VhcghnAauTBsbg
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9043
  • Country: us
  • DavidH
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #12 on: January 27, 2019, 02:10:44 pm »
Because you are looking at a DC signal (super slow) and the scope will sample many times faster, if you have a way to program the scope, you could possible gain some resolution and accuracy.

You can get higher resolution this way but not accuracy.  Even after calibration, the vertical signal path including the ADC which is optimized for bandwidth and transient response has poor linearity.
 

Offline Rerouter

  • Super Contributor
  • ***
  • Posts: 3969
  • Country: au
  • Question Everything... Except This Statement
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #13 on: January 27, 2019, 02:30:16 pm »
The ADC is not always the best, but the channel offset DAC atleast in these scopes are the root of trust as far as the user  self calibration is concerned. the calibration routine is:

set the amplification / attenuation to a certain amount
shift the offset DAC until a reading of 0 is given,
shift the offset dac higher / lower to get another reading

Save zero for that amplification, and gain for that attenuation range.

This is why the offset can be trusted for accuracy to a reasonable level, the rest of the scope is user calibrated to that. 
 

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 5240
  • Country: us
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #14 on: January 27, 2019, 04:55:53 pm »
Because you are looking at a DC signal (super slow) and the scope will sample many times faster, if you have a way to program the scope, you could possible gain some resolution and accuracy.

You can get higher resolution this way but not accuracy.  Even after calibration, the vertical signal path including the ADC which is optimized for bandwidth and transient response has poor linearity.

I assume you watched the opening bit of the video and jumped to your own conclusions without ever seeing where I talk about the non-linearity and how modeled them.    I stand by my statement.
How electrically robust is your meter?? https://www.youtube.com/channel/UCsK99WXk9VhcghnAauTBsbg
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 28376
  • Country: au
    • EEVblog
Re: Are scopes expected to be highly accurate on DC measurements?
« Reply #15 on: January 27, 2019, 05:21:30 pm »
No, look at the specs of any scope, the guaranteed accuracy is rarely better than even the cheapest $2 multimeter.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf