Why are oscilloscopes so inaccurate?

[coppercone2]

scopes are good at telling you what is actually happening instead of giving too useful a measurement etc

its like being able to look at something to see if its doing the right thing then trying to break out and measuring tape

ANd I think high end unfriendly lecroys do a bit more then 'suck' (well they do suck balls but they are good)

its like, being able to look at something and then decide where you put the ruler to measure it. The cabinet looks crooked (not lets measure every part of the cabinet to see if its good.. carpenters would go insane).

[David Hess]

You use an oscilloscope that includes a differential comparator.

Sounds like a use case for a vintage analog scope...

http://w140.com/tekwiki/wiki/1A5

Or something a couple years newer:

http://w140.com/tekwiki/wiki/7A13

Those differential comparators are ancestors to the ones now made by Preamble/LeCroy:

https://teledynelecroy.com/probes/differential-amplifiers

An alternative method is to make an integrated measurement at a specific point in the waveform which gives the speed advantages of sampling with the low noise of integration:

http://w140.com/tekwiki/wiki/7D12

And that sort of thing went on to become gated integrated measurements in automated analog oscilloscopes like the Tektronix 4 channel 22xx series and maybe 24xx series.

A DSO's averaging mode is effectively the same thing but is still limited by the precision of their signal conditioning and digitizer.  In practice, very few applications require better accuracy so the alternatives are no longer economical.  They were used in the past because there was no better alternative.

[dom0]

As an owner and not-so-infrequent user of a 7A13 I can say that it is a surprisingly useful tool to have in the bag, just like the 7A22.

Measurement functions of analog 'scopes also tend to be decent. For example, the DC VM in the 2246 has an accuracy spec of +- 0.5 % of reading. Peak measurements are implemented using precision rectifier, so relatively accurate as well. Only if you use manual cursors you are actually getting the full deflection and aberration errors.

Edit: On second thought I am not so sure about the manual cursors. My guess would be that those are implemented using S&H circuits, so deflection and aberration errors are zero since cursor and measurement position are identical. So even those might actually be quite good. Have to check.

[David Hess]

As an owner and not-so-infrequent user of a 7A13 I can say that it is a surprisingly useful tool to have in the bag, just like the 7A22.

I really like the 7A13's versatility but it comes at the cost of high noise.  I am not sure there is anyway around that when a bootstrapped differential amplifier is involved.

I do not know about Tektronix but HP makes low voltage probes which include the voltage comparison function.  The offset controls of a DSO do the same thing but usually operate over only a very limited range unlike a 7A13 which is a very useful +/-10 volts.

Measurement functions of analog 'scopes also tend to be decent. For example, the DC VM in the 2246 has an accuracy spec of +- 0.5 % of reading. Peak measurements are implemented using precision rectifier, so relatively accurate as well. Only if you use manual cursors you are actually getting the full deflection and aberration errors.

Peak voltage measurements on the 22xx series are actually made using the B trigger which is how these oscilloscopes achieve full bandwidth peak voltage measurements.  DC voltage measurements are made by inserting a very low pass trigger coupling filter which explains why they cannot be gated which is too bad.

Signals are measured by using the B trigger level comparator as a successive approximation ADC to determine the peaks or DC level of the applied signal.

Edit: On second thought I am not so sure about the manual cursors. My guess would be that those are implemented using S&H circuits, so deflection and aberration errors are zero since cursor and measurement position are identical. So even those might actually be quite good. Have to check.

I think that is correct; cursor measurements should be more accurate than graticule measurements for that reason.

[HendriXML]

I had some fun using a DSO as a decent precision voltage meter.

By doing some calibration steps, using the vertical offset (using multiple iterations narrowing down the measurement window) and a lot of averaging it is possible to get pretty descent precision.

https://www.eevblog.com/forum/testgear/using-a-awg-and-a-scope-do-a-low-voltage-level-characterization-of-a-1n4005/msg2476305/#msg2476305

[drummerdimitri]

I've recently noticed that the voltage measurement of an AC signal on my scope change a lot with the time division setting.

Is there a norm to set the time division to X seconds or show X amount of cycles for the most accurate readings?

[David Hess]

I've recently noticed that the voltage measurement of an AC signal on my scope change a lot with the time division setting.

Is there a norm to set the time division to X seconds or show X amount of cycles for the most accurate readings?

That should not happen unless the AC measurement is made over the entire record and only a fraction or small number of cycles are captured.  Usually the default is to make the measurement over 1 or a whole number of cycles if they can be detected.

[nctnico]

I've recently noticed that the voltage measurement of an AC signal on my scope change a lot with the time division setting.

Is there a norm to set the time division to X seconds or show X amount of cycles for the most accurate readings?

No. Some oscilloscopes do measurements on a decimated (downsampled) set of data to speed up measurements. If you have such an oscilloscope: As a result the accuracy goes down if you have many cycles on the display. As a rule of thumb you should be able to see the shape of the signals for the measurement to be accurate. Ofcourse you should always have at least one full cycle on the display.

[Bassman59]

1. Unless your signal is just the right amplitude to fill the entire input range of the ADC (ie, just before clipping) then it won't be giving you the full 8 bits anyway.

What? No.

[TheUnnamedNewbie]

Another thing to keep in mind: Even if you have the most accurate scope in the world, have you stopped to consider the accuracy impact of your probes, connectors, coax, etc?

If you want to measure to a fraction of a percent, a small amount of impedance mismatch here, a bit of parasitic inductance and capacitance there, and boom, you lost your fraction of a percent. Think about the fact that you have to manually go in and tweak that 10x probe. How on earth are you ever supposed to get accurate measurements out of that?

This is a similar reason why things like spectrum analyzers can't measure to extreme accuracy.

[magic]

You will sooner reach limitations of the scope than the probe.

Particularly at DC. Scope probes are actually usable with DMMs in a pinch if you calibrate them beforehand. I did it once to measure something that needed not to be disturbed by extra capacitance and noise, though I was only after 1% or so accuracy.

A scope I wouldn't trust to get even 1% right.

[tggzzz]

You will sooner reach limitations of the scope than the probe.

DC voltage maybe, maybe not.

Otherwise poor probing technique frequently dominates measurement validity.

[SilverSolder]

Playing around with a ratio transformer the other day, I decided to check a 54622D scope, just for fun.  Fed an accurately sourced 1V rms signal at 50Hz via a 10:1 probe, the Agilent was able to achieve ~0.7% accuracy on the calculated RMS value it displayed on the screen. 

This level of accuracy (better than the spec of 2%) was only achieved right after running the self calibration function.  The accuracy was a lot worse before running user cal - several percent out!

Overall better than expected performance, and a good demonstration of the value of the Auto Cal function.