Fluke 87V - "hi-res" (19,999-count) mode - accuracy?

[caspencer]

I was reading over the detailed specifications in the 87V user manual and found something interesting:

For all detailed specifications:

Accuracy is given as ±([% of reading] + [number of least significant digits]) at 18° C to 28° C, with relative humidity up to
90 %, for a period of one year after calibration.

For Model 87 in the 4 ½-digit mode, multiply the number of least significant digits (counts) by 10.

So I take it that means means that since the basic DC voltage accuracy is 0.05% + 1 count (in regular, 6000-count mode), in "hi-res" the accuracy would then be 0.05% + 10 counts (hence the multiply by 10).

So my question is why bother "hi-res" mode at all? Doesn't that mean the extra digit you get in hi-res mode is sort of useless and almost guaranteed to be wrong. For those that have an 87V do you use "hi-res" mode? If so, when?

I've seen EEVBlog #26 - Multimeter Counts, Accuracy, Resolution & Calibration (http://www.eevblog.com/2009/08/21/eevblog-26-multimeter-counts-accuracy-resolution-calibration/) so I get that more counts doesn't mean more accuracy. I guess I was just surprised that with the 87V more counts in fact means less accuracy.

[Kiriakos-GR]

I will offer you few practical examples , about who cares for total accuracy.

a ) Food industry   ( so all the potato chips to baked by the same way and taste the same )
b ) chemicals industry   ( keep reading the other categories )
c ) medical equipment  ( the simple measurement of sugar in the blood , it must be confirmed by all the devices that monitors that )
d ) The high quality scales  ( so all the amount of ingredients for mixing, to be always at the same amount)
e ) The army  ( All the targeting systems , they have to be checked too , from time to time.) 
f ) The electronics industry it self  (  That builds tools that measures distance - Air speed - Pressure - Temperature )
g ) Navy and Air force  (  GPS and navigation systems  + depth measurements - altitude - humidity - radars )

I can keep writing for hours , but it will not be an example , any more .. 


 

   

[EEVblog]

More resolution can be very important for many things.
Comparison/adjustment of two measurements for example.
Because (on a good meter) the internal reference essentially has no drift if you keep the same temperature, those extra digits aren't useless at all.
For example, if you are measuring two signals that are say 1.0000V and 1.0001V, the absolute displayed value could in theory be 0.05%+/-10 counts out. But that's not a random thing, it's consistent between different readings. So you won't get say 1.0010V and 0.9990V, the two measurements will measure only 0.0001V different as you'd expect, and you may not care about the absolute value.
(not including LSD error)
So you could have a 1% absolute accurate 50,000 count meter, and that resolution can still be very useful.

Dave.

[alm]

The accuracy doesn't get worse, it just doesn't get any better.

I agree with Dave, extra resolution is useful. As an extreme example, the Agilent 3458A has 8.5 digits of resolution, but only has 0.5ppm accuracy best case over a 24 hour period, which means the last two digits are 'garbage'. And even just 8ppm (about 5.5 digits of accuracy) with a yearly calibration interval. 10 minute transfer accuracy is specified as 0.05ppm though, so that's 7.5 digits of accuracy over short time spans.

I don't believe linearity and transfer accuracy is specified for the 87V, so you are basically treading into unspecified territory if you assume anything better than 24h accuracy specs (do they even specify those? or only 1 year?). I would also expect the short interval to have more effect on gain errors than offset errors. I agree with Dave that it's likely to be better over say a 10min time span, but I wouldn't assume it to be accurate to the LSD either. There's alway some inherent noise and drift in a voltage reference, ADC and divider. It may be significantly less than 1 LSD, or it may not be.

[caspencer]

Because (on a good meter) the internal reference essentially has no drift if you keep the same temperature, those extra digits aren't useless at all.
For example, if you are measuring two signals that are say 1.0000V and 1.0001V, the absolute displayed value could in theory be 0.05%+/-10 counts out. But that's not a random thing, it's consistent between different readings.

So basically the value of more resolution is more for comparing hi-res readings than just taking single readings.
Whatever potential error there is should be consistent between readings... at least over a short time frame and same test environment conditions (temp, humidity, etc).

 

[Kiriakos-GR]

So basically the value of more resolution is more for comparing hi-res readings than just taking single readings.
Whatever potential error there is should be consistent between readings... at least over a short time frame and same test environment conditions (temp, humidity, etc).

I use the Hi-res only so to check if the meter is spot on , with reference voltage sources.
And this is how I use it mostly ..
I had also use the hi-res when I was build my resistors decade box.
It is also good about comparing resistors , if you need to make an Stereo preamp,
and you need both sound channels , to be identically build. ( like a face on the mirror )     

[EEVblog]

So basically the value of more resolution is more for comparing hi-res readings than just taking single readings.

Usually, yes, but not always. Some people simply do need the utmost absolute accuracy for a single reading.
To them, all the stuff alm spoke about matters a lot.

Whatever potential error there is should be consistent between readings... at least over a short time frame and same test environment conditions (temp, humidity, etc).

Yes, you'd expect that of any good quality meter. And that's why good quality meters like the Fluke use a precision thick film hybrib resistor network that are very stable over temp and time, instead of just using off-the-shelf 0.01% resistors or whatever.
Maybe not to the last LSD over the FULL temp range like alm said, but for any practical temp range in a lab, you's expect it to be pretty darn stable between measurements.

Dave.

[gobblegobble]

And that's why good quality meters like the Fluke use a precision thick film hybrib resistor network that are very stable over temp and time, instead of just using off-the-shelf 0.01% resistors or whatever.

I'm interested to know if anyone can shine on light how that resistor network relates to the Gossen's choice, as both of them are obviously head and shoulders above the through-hole resistors ubiquitously present in cheap crappy multimeters.

[saturation]

Networks are matched, not only are they precise they are often balanced so that one positive tempco is added to another with a negative one, canceling each other out.  The net results is a resistor that is fairly stable of a wide range of temperatures, usually in the 0-100F range.

And that's why good quality meters like the Fluke use a precision thick film hybrib resistor network that are very stable over temp and time, instead of just using off-the-shelf 0.01% resistors or whatever.

I'm interested to know if anyone can shine on light how that resistor network relates to the Gossen's choice, as both of them are obviously head and shoulders above the through-hole resistors ubiquitously present in cheap crappy multimeters.