EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: jpb on May 14, 2014, 07:25:18 pm
-
Sorry for the slightly obscure title - I've been comparing DMM specs (Gossen MetraHIT Ultra and the MetraHIT Energy).
The specs for the MetraHIT for the lowest VDC range are given as 0.02% of reading + 15 counts where each count corresponds to 1 uV
this is all quite normal.
BUT the MetraHIT Ultra gives its specs as 0.02% of reading + 0.005% of max range + 10 counts with each count corresponding to 1uV as well.
The max range is 300,000 counts so 0.005% is 15 counts so why not just state the accuracy as 25 counts ? (Or as a bigger % of maximum range.)
Is there any reason for splitting into two different error terms like this (other than just to hide the total)?
My reading of this is that the Energy is more accurate though it has a lower count.
-
I think that there is some reason to show MetraHIT Ultra accuracy as "+-(...% rdg + % MR +... d)" 0.02 + 0.005 + 10, because this meter possibly has few resolution modes (300k and 30k). I am not sure of this, you have to read the manual. If this is true, then
- for 300k mode, 0.02 + 0.005 + 10 is actually +-0.02%+15+10 = +-0.02%+25
- for 30k mode, 0.02 + 0.005 + 10 is actually +-0.02%+1.5+10 = +-0.02%+11.5
3 accuracy numbers allows more flexibility in defining accuracy of particular meter.
In the end accuracy specification type doesn't meter at all, you have to take into account all errors and calculate uncertainty for a particular measurement.
-
I think that there is some reason to show MetraHIT Ultra accuracy as "+-(...% rdg + % MR +... d)" 0.02 + 0.005 + 10, because this meter possibly has few resolution modes (300k and 30k). I am not sure of this, you have to read the manual. If this is true, then- for 300k mode, 0.02 + 0.005 + 10 is actually +-0.02%+15+10 = +-0.02%+25
- for 30k mode, 0.02 + 0.005 + 10 is actually +-0.02%+1.5+10 = +-0.02%+11.5
3 accuracy numbers allows more flexibility in defining accuracy of particular meter.
In the end accuracy specification type doesn't meter at all, you have to take into account all errors and calculate uncertainty for a particular measurement.
Thanks for the idea. What you say has some basis as the extra %MR term is only given for those functions which use the full 309,999 the other functions just have %rd + count.
Unfortunately I have gone through the manual which I downloaded and I can see no option to reduce the range and count. If you are measuring Vdc then the count is always 300k and when you are measuring AC or AC+DC the count is always 30k.
I've tried sending the question to Gossen but as they only allow 4 very short lines in their message box I'm not sure I've explained it well enough to make sense.
-
In the end accuracy specification type doesn't meter at all, you have to take into account all errors and calculate uncertainty for a particular measurement.
This is right, but the total error in the lower ranges seems quite large and I'm hoping it is not really as bad as it seems (in comparison to other meters in the range).
-
As discussed, specification type used is not very important (there is no strict standard, just general concensus what should be reported).
I have another idea. MetraHIT Ultra is 309999 count meter, MetraHIT Energy is 60000 count meter. The higher the total meter count, the higher "error" count. However that doesn't look good on specs, does it? This reasoning leads to few alternative possibilities:- MetraHIT doesn't want to say directly that Ultra has more "error" counts than Energy (Ultra: 25, Energy: 15). Ultra is advertised as both bigger count (309999) and smaller (5) count error (vs 60000 and 15). But if you compare error for fullrange, 25/309999~0.008%+0.02%, 15/60000~0.025%+0.02%, Ultra is still slighly better than Energy.
- That %MR reading may be over pessimistic, worst case. Maybe it is long term only, for short term it could be 0.000
- As Dave said, "These Germans, they like to confuse you"
Overall, I think that Ultra uses different and better ADC than Energy (giving better overrall accuracy), so they incorporated additional adjustments for that specific ADC. Tripple digit accuracy specs are for current ranges as well.
This is right, but the total error in the lower ranges seems quite large and I'm hoping it is not really as bad as it seems (in comparison to other meters in the range).
Specs for Ultra and Energy are pretty decent. What other meters are you comparing?
-
You mix voltages and counts and compare different ranges.
The total error is almost always decomposed in a proportional error part ("linear") and a residual covers-all fixed error (includes everything from an explainable offset to 'random' effects).
For the best meter the 'reading' error (the first proportional part of the total error) is attributed to the ADC, the (also proportional) 'range' error is associated with the front end precision resistor divider, and the rest is undefined 'counts' error.
For the other meter they have found that they cannot with enough certainty separate proportional errors originating in ADC and divider, so they throw in a 'large' residual error to be on the sure side. That way they can use a small good-looking percentage error. Also - that tells you that if you do a lot of measurements and average them, the instrument is quite good.
Comparing two spec sheets directly is difficult and often leads to confusion and meaningless statements. Find the relevant tables in
http://www.produktinfo.conrad.com/datenblaetter/125000-149999/126482-da-01-de-GMC_M249A_METRAHIT_ENERGY.pdf (http://www.produktinfo.conrad.com/datenblaetter/125000-149999/126482-da-01-de-GMC_M249A_METRAHIT_ENERGY.pdf) (p.4) and http://www.gossenmetrawattusa.com/products/measurement/PDF/MetraHIT%20ULTRA%20BT%20Operating%20Instructions.pdf (http://www.gossenmetrawattusa.com/products/measurement/PDF/MetraHIT%20ULTRA%20BT%20Operating%20Instructions.pdf) (p. 58), and calculate the max voltage error for 100 mv, 400 mv, 700 mv, 1 v, 4v, 7v, 10 v ... or whatever (the meters change range on .3 / .6 and so on), and it will be obvious that the 'best' meter is best.
-
The 60,000 count Energy meter is only more accurate than the 310,000 count Ultra when on its low 60mV range. On any other range, the Ultra has more accuracy.
Accuracy can also be defined in terms of the deviation from an ideal transfer function as follows:
y = mx + b
where x is the input
m is the ideal gain of a system
b is the offset
Applying this example to a DMM signal measurement, y is the reading obtained from the DMM with x as the input, and b is an offset error that you may be able to null before the measurement is performed. If m is 1, the output measurement is equal to the input. If m is 1.000001, then the error from the ideal is 1 ppm or 0.0001 percent.
High-resolution, high-accuracy DMMs describe accuracy in units of ppm and are specified as ±(ppm of reading + ppm of range). The ppm of reading is the deviation from the ideal m; ppm of range is the deviation from the ideal b, which is zero. The b errors are most commonly referred to as offset errors.
-
Thanks for your answers everyone.
Both the Ultra and the Energy have good specs, the Ultra is quite a bit cheaper than the Energy with a higher count but fewer functions.
I'm trying to decide whether or not a high end portable DMM is close enough in accuracy to say a 5.5 digit bench DMM that I can go for the extra flexibility - they cost a similar price but in a way I'm trying to compare apples and oranges.
I did produce a little table for myself giving the error at various measurement points from 10uV to 10V and it is quite illuminating - for instance though the headline error rate of the Ultra is 0.02% which is similar to 0.014% on my bench Adventest 7451A when I created my table I found the overall error on the Ultra is around 4 times worse for much of the range.
Being excessively analytic I have decided to produce an average error as a single figure of merit by integrating from 10mV to 10V (somewhat arbitrary range) and assuming a uniform distribution of measurements in that range! If the results look interesting I'll report back.
-
I have had a very prompt response from Gossen.
There are two reasons they give, one is it is historical but the other perhaps makes more sense in that the digits error affects the display whilst the MR error is a fixed offset.