Rigol DC accuracy?

[AlexI]

I'm looking at using a Rigol DS1054Z oscilloscope in an automated test&measurement setup.  It works okay controlled over USB from python code (long story for another time on all the steps needed to get there).  Now I'm looking at the measurement accuracy.

Here is what the spec sheet says:

DC Gain Accuracy
<10 mV: ±4% full scale
≥10 mV: ±3% full scale

DC Offset Accuracy
±0.1 div±2 mV±1% offset value

0.1 div?!?   If I'm reading this correctly, there could be a 1 VOLT offset on the 10 V/div range and that is within spec.  Again at 10V/div, for a 30V signal the worst case error would be 4*10V/div*3% + 10V/div*0.1 = +-2.2V (the 2mV is insignificant here; the offset is zeroed out for the measurement). For a 0V signal it could be  +-1V.  So, as a DVM, the scope is basically a 2-digit model.

Could someone check my math on this?  Does this seem a bit (okay, a lot) inaccurate?  How does it compare to other scopes?

[Kleinstein]

DC accuracy of scopes is generally not good. Quite often one has to do an DC level adjustment / Zero before use. A scope is mainly an AC instrument.

The specs do not specify a time frame or temperature change till last zero. So the specs would be rather wide to cover worst case samples.

The higher ranges are often done with a divider up front. So especially for the higher ranges the specs can be rather poor compared to a DMM. After a zero check / adjust the accuracy can be quite a bit better.

[Berni]

Yep scopes in general are terrible in terms of accuracy on the larger ranges. Hence why they still make digital multimeters.

Here have a look at the spec for a Keysight MSOS804A that costs about $50 000:

[Fungus]

0.1 div?!?   If I'm reading this correctly, there could be a 1 VOLT offset on the 10 V/div range and that is within spec.

Yep.

How does it compare to other scopes?

It's about average.

Remember that most 'scopes only have an 8 bit ADC so one ADC step is two pixels on a 400 pixel tall screen.

It doesn't get better by spending more money, either. High end ADCs are still 8 bits.

[AlexI]

Remember that most 'scopes only have an 8 bit ADC so one ADC step is two pixels on a 400 pixel tall screen.

True!  The 8 bits doesn't quite explain the full offset though.  If the offset is one bit or one LSB, then it would be 1/256th of full scale, say 0.3V at 10V/div. 

[jjoonathan]

High bandwidth 1MOhm amplifiers all seem to have loads of DC drift (thermal and otherwise), even on expensive oscilloscopes, so I tend to suspect it's a fundamental amplifier tradeoff.

Which isn't to say that the 50 ohm path makes a better DMM 

[2N3055]

Well, PicoScope 4262 has 0.25% at more than 10mV/DIV..
It also has 5 MHz bandwidth, and 16 Bit resolution...
And is 3x more expensive than DS1000Z

[tautech]

Some insights into DSO measurement accuracy/inaccuracy can be found in this thread:
https://www.eevblog.com/forum/testgear/testing-dso-auto-measurements-accuracy-across-timebases/

In short, the users actions can influence DSO measurement accuracy.

[David Hess]

For a 0V signal it could be  +-1V.  So, as a DVM, the scope is basically a 2-digit model.

That is about right for a typical oscilloscope.  But keep in mind that the digit specification for a digital voltmeter excludes the sign while the oscilloscope's specification does not.  So an 8 bit oscilloscope is 256 counts or +/-128 counts which is almost exactly 2 digits or +/-99 counts.  2.5 digits would be +/-199 counts and of course digital voltmeters start at 3.5 digits or +/-1999 counts or basically 12 bits.

Better instruments can be 2% or even 1%.

[Fungus]

True!  The 8 bits doesn't quite explain the full offset though.  If the offset is one bit or one LSB, then it would be 1/256th of full scale, say 0.3V at 10V/div.

The 8 bits of the ADC isn't the offset. The offset will be generated by some sort of DAC that feeds into the opamps.

In the case of the Rigol the front end was reverse engineered:

https://www.eevblog.com/forum/blog/eevblog-675-how-to-reverse-engineer-a-rigol-ds1054z/

[Berni]

Yep its not as simple as the ADC.

There are DACs that let you look at a tiny signal on top of a large DC bias and the amplifiers in the analog front end all have a tiny offset and even the ADC itself can have a offset larger than a LSB. All of these errors together can pile up or move around as the scope warms up.

In general you are not supposed to trust more than 2 digits on a vertical axis measurement on scopes. Its only the horizontal axis that is truly accurate since in electronics its much easier to measure time accurately and crystal oscillators provide a very stable reference to measure from (Or you can bring your own precise 10MHz into the back of a scope if needed).

[BravoV]

In general you are not supposed to trust more than 2 digits on a vertical axis measurement on scopes.

+1 , and if the OP has a signal generator that can generate precise consistent signal level through out the sweep, from say 1 MHz up to the scope's official bandwidth at 50 MHz, he will definitely see how crappy it is for the scope's vertical signal accuracy. 

[Berni]

Oh and if you do need better vertical accuracy from a scope this is how to do it:
https://tmi.yokogawa.com/solutions/products/data-acquisition-equipment/scopecorders/dl850edl850ev-scopecorder/


You can populate it with one of these:
https://tmi.yokogawa.com/solutions/products/data-acquisition-equipment/high-speed-data-acquisition/701251-analog-voltage/


This gets you to +/-0.25% of range accuracy. And as a bonus its also a fully floating input but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

[BravoV]

..... but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

Reading above points, its really an appetite killer isn't it ? 

[hgjdwx]

The measured vertical accuracy error of MICSIG STO1104C is less than 1% in most cases and less than 1.5% in some cases, So the third digit is useful

[frogg]

Yep its not as simple as the ADC.

There are DACs that let you look at a tiny signal on top of a large DC bias and the amplifiers in the analog front end all have a tiny offset and even the ADC itself can have a offset larger than a LSB. All of these errors together can pile up or move around as the scope warms up.

In general you are not supposed to trust more than 2 digits on a vertical axis measurement on scopes. Its only the horizontal axis that is truly accurate since in electronics its much easier to measure time accurately and crystal oscillators provide a very stable reference to measure from (Or you can bring your own precise 10MHz into the back of a scope if needed).

100% this

[Berni]

The measured vertical accuracy error of MICSIG STO1104C is less than 1% in most cases and less than 1.5% in some cases, So the third digit is useful

Yes it most certainly can. But a spec in the datasheet doesn't cover "some cases" or "most cases" but it covers "all cases". There is always some breathing room on top of the spec to make sure they can guarantee that every scope leaving the factory meets that spec.

I'm not saying that the 3rd digit is useless, but what i mean to say is anything past 2 digits on vertical can't just be blindly trusted to be accurate. If you do want to use digits past that , then you have to make sure the scope is in the best possible configuration and also repeat the measurement from a few slightly different configurations to make sure you get the same result. With a bit of manual null calibration and averaging i'm sure a decent scope can actually be more accurate than cheap multimeters.

Scopes can take a significan't amount of knowledge to use correctly. And its not just about knowing how to set a scope up correctly, its also a lot about how to correctly probe things, especially for high speed stuff above 10 MHz where all sorts of things come into play like probe loading, correct grounding, bandwidth limitations, time delay over cables, impedance matching, reflections etc. A lot of the time a fast square wave looks crappy on a scope just because the scope is showing it like that rather than the circuit actually generating a crappy rounded wobbly overshooting square wave.

[Fungus]

I'm not saying that the 3rd digit is useless, but what i mean to say is anything past 2 digits on vertical can't just be blindly trusted to be accurate. If you do want to use digits past that , then you have to make sure the scope is in the best possible configuration and also repeat the measurement from a few slightly different configurations to make sure you get the same result.

But before you do any of that, ask yourself what would it prove? What difference would it make?

Is it even a good idea to try to do it, given the overall uncertainty in the readings?

[David Hess]

Oh and if you do need better vertical accuracy from a scope this is how to do it:

You can populate it with one of these:

No, populate it with one of these:

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

This gets you to +/-0.25% of range accuracy. And as a bonus its also a fully floating input but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

Only 0.25% to 300kHz?  A 7A13 is 0.1% to 100MHz.

In all seriousness, LeCroy make differential comparators which can be used with any oscilloscope but they are being discontinued:

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

[vk6zgo]

Oh and if you do need better vertical accuracy from a scope this is how to do it:

You can populate it with one of these:

No, populate it with one of these:

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

This gets you to +/-0.25% of range accuracy. And as a bonus its also a fully floating input but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

Only 0.25% to 300kHz?  A 7A13 is 0.1% to 100MHz.

But it only works with real oscilloscopes! 

In all seriousness, LeCroy make differential comparators which can be used with any oscilloscope but they are being discontinued:

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

The "old school" way to set "any old" analog  'scope for best accuracy is to find a standard voltage source (a 9v battery will do in a pinch, but you need to check it with a DMM, & record the real voltage), turn the volt/div on the 'scope to "uncal", measure the source with the 'scope then adjust the display to the exact number of divisions it should be for that source.

Then, without changing the v/div from "uncal, look at the signal of interest.(if you use a DC reference, the 'scope must be left in "DC coupled" mode.)

This is how we checked 1volt analog video signals back in the day, before 'scopes &, particularly, waveform monitors (the Marconi combined pix & waveform monitors were particularly dire) became  as stable as they were in later years.
The reference standard voltage in this case was usually a high quality video source.

[Berni]

Oh and if you do need better vertical accuracy from a scope this is how to do it:
You can populate it with one of these:

No, populate it with one of these:
http://w140.com/tekwiki/wiki/7A13

This gets you to +/-0.25% of range accuracy. And as a bonus its also a fully floating input but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

Only 0.25% to 300kHz?  A 7A13 is 0.1% to 100MHz.

Yeah that's also one way to go about it.

But the scope will still have the typical tolerance of the displayed waveform, with this it can just be made insignificant by zooming into the area of interest. The 0.1% is just the accuracy of the zero offset. The Yokogawa gives its claimed accuracy for every spot on the waveform, not just for points at zero.

Two different products for different use cases. But i know id prefer the user interface of the Tek with its immediate switches rather than menus upon menus. Clear front panel UI is a lost art these days.

[David Hess]

Oh and if you do need better vertical accuracy from a scope this is how to do it:
You can populate it with one of these:

No, populate it with one of these:
http://w140.com/tekwiki/wiki/7A13

This gets you to +/-0.25% of range accuracy. And as a bonus its also a fully floating input but the bandwidth is really low(300KHz) to make all of this happen. Tho this setup likely costs a few times what your Rigol cost.

Only 0.25% to 300kHz?  A 7A13 is 0.1% to 100MHz.

Yeah that's also one way to go about it.

But the scope will still have the typical tolerance of the displayed waveform, with this it can just be made insignificant by zooming into the area of interest. The 0.1% is just the accuracy of the zero offset. The Yokogawa gives its claimed accuracy for every spot on the waveform, not just for points at zero.

Two different products for different use cases. But i know id prefer the user interface of the Tek with its immediate switches rather than menus upon menus. Clear front panel UI is a lost art these days.

That is not how a different comparator is used for precision voltage measurement; they are for "slide-back" measurements.

The first point to be measured is aligned with the graticule.  Then the differential comparator is used to "slide back" the second point to be measured to the same spot and the voltage difference is displayed.  Since the oscilloscope is being used as a null meter, calibration and linearity are irrelevant. (1)

Time measurements made with delta delayed sweep are "slide back" measurements in the horizontal direction.

This technique is basically obsolete with digital storage oscilloscopes which can make measurements between any two points however it is still more precise.

(1) This ignores the issue with thermal balance but a differential comparator clamps the signal level to minimize errors from this.

[Berni]

Yes that is what i meant with "zooming into the area of interest". When the signal is shifted down close to zero you can use the more sensitive vertical scales of the scope and make the feature you are looking for fill the screen as much as possible around zero, helping pinpoint exactly where that part of the waveform is.

And yeah this is much bigger of a deal with analog scopes that usually have a completely uncalibrated pot for the offset control so once you run your offset off the screen you have no idea what offset it is until you hook up a DC reference voltage to it and manually tweak it to match it. Digital scopes know the offset voltage they are applying so the can figure it out on there own.

Because i still like to count graticules on a scope i find it very convenient when digital scopes mark Volts next to each graticule. It lets you still graticule count even if your offset is off screen. Tho when i want a bit more accuracy rather than at a glance, i still use cursors and automated measurements. But if like me you often use "average volts" measurement as a quick handy multimeter for checking supply rails you soon see how way off the result can be if the scope is not set up correctly.