EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: jpb on May 17, 2014, 04:12:29 pm
-
I'm excessively analytical (friends and relatives would probably use a ruder term for it!) and I've been trying to compare different DMMs both handheld and bench.
Edit: NB the following is purely designed to compare the accuracy of meters and not such things as features or build quality, warranty etc.
I've come up with the following figure(s) of merits to try and compare accuracy between different meters which may have very different ranges/counts etc and to give a realistic error figure rather than the headline dc figure that manufacturers like to highlight (and to be fair it probably is useful if you are measuring voltage changes rather than absolute values).
The main figure of merit is the expected (in the statistical sense) percentage error if your measurements are randomly selected from a uniform range from 60mV to 6V. This is not as arbitrary as it seems and is easy to calculate from the spec sheets.
The other figure of merit is the maximum error in the same range.
The lower end of the range (60mV) I selected after looking at quite a few specs as being high enough that the error there is no worse than the maximum error higher up the ranges. I started trying 10mV but this was too low in the sense that for many meters the maximum error was at 10mV and much bigger than else where.
The upper end is fairly arbitrary but 6V is 100 times 60mV and if you extend it further you are (in my view) unfairly weighting the higher/easier measurements - of course anyone can choose their own range.
I chose a uniform distribution of measurements as I've no idea what other distribution you might use (a Normal distribution about 3V perhaps?) and a uniform distribution simplifies the maths.
I also plotted the results of E[error] vs meter price and came up with the following bizarre formula for Fair Price ! (to be taken with a pinch of salt):
Price (inc VAT in UK £) = -357 x LOG10(E[error in %]/1.126)
Here is how this formula works out for the meters I looked at:
| DMM | Price from Formula | Actual Price |
| Gossen M30 | 860 | 840 |
| Agilent 34461A | 828 | 846 |
| Hameg 8012 | 590 | 648 |
| Hioki DT4282 | 538 | 599 |
| Gossen Ultra | 477 | 479 |
| Gossen X-tra | 313 | 359 |
| Agilent U1273A | 385 | 350 |
| Hioki DT4252 | 124 | 119 |
I attach a pdf of my results in detail.
-
I'm excessively analytical
From what you have done here, I think so too :)
I also plotted the results of E[error] vs meter price and came up with the following bizarre formula for Fair Price ! (to be taken with a pinch of salt):
Price (inc VAT in UK £) = -357 x LOG10(E[error in %]/1.126)
Here is how this formula works out for the meters I looked at - from this the Fluke 87V looks a little pricey (the formula takes no account of build quality/features etc
but all meters compared are from good quality manufacturers):
You can justify one thing over another anytime if you rationalize your own bias. In this case you have picked on certain specifications and made them more important than others. There are many who would end up putting the 87V at the top solely for the merit of the Fluke warranty and confidence in quality. Here you have picked an arbitrary specification of a certain arbitrary range of voltage measurement specs and then use this to extrapolate the value of the meter on a whole. This is totally misleading.
You need to look at what your REAL needs are and then find the best meter to meet these needs at the price you can or want to pay. There is no magic formula.
-
You can justify one thing over another anytime if you rationalize your own bias. In this case you have picked on certain specifications and made them more important than others. There are many who would end up putting the 87V at the top solely for the merit of the Fluke warranty and confidence in quality. Here you have picked an arbitrary specification of a certain arbitrary range of voltage measurement specs and then use this to extrapolate the value of the meter on a whole. This is totally misleading.
You need to look at what your REAL needs are and then find the best meter to meet these needs at the price you can or want to pay. There is no magic formula.
Actually I think the Fluke should be higher, I certainly don't have a bias against Fluke and I did say the results needed to be taken with a pinch of salt.
I was just seeking a way of comparing accuracies with one or two numbers otherwise it is hard to compare different ranges.
I think I'll remove the Fluke results as they seem to suffer from having a single wide range with a +- 1 digit error of 1mV which I guess was a design choice.
-
There would seem to me to be a rather large problem with just using their published specifications in trying to force rank a population of DMMs. Does build quality, manufactures reputation and track record, resale value, warranty length, etc not have a place in any ranking effort?
-
There would seem to me to be a rather large problem with just using their published specifications in trying to force rank a population of DMMs. Does build quality, manufactures reputation and track record, resale value, warranty length, etc not have a place in any ranking effort?
I realise that I rather badly expressed what I was trying to do. I was comparing DMMs all from good manufacturers and often within the range of the same manufacturer so that quality, warranty and so on were all very similar but I wanted to see how accuracy compared.
The main purpose from my point of view was to see how close a handheld DMM might compare with a mid-range bench DMMs purely in terms of accuracy of typical measurements without getting lost in comparing different ranges and so on.
-
Someone has already done something very useful for visualizing the differences in multimeters based on their specifications. Have a look at this:
http://www.johannes-bauer.com/electronics/mmcomparison/app/index.html (http://www.johannes-bauer.com/electronics/mmcomparison/app/index.html)
I have contributed a couple of items to his web page data, but you can enter your own locally and view the comparisons. Just follow the examples of the data from an exported data of one of the existing meters.
-
Someone has already done something very useful for visualizing the differences in multimeters based on their specifications. Have a look at this:
http://www.johannes-bauer.com/electronics/mmcomparison/app/index.html (http://www.johannes-bauer.com/electronics/mmcomparison/app/index.html)
I have contributed a couple of items to his web page data, but you can enter your own locally and view the comparisons. Just follow the examples of the data from an exported data of one of the existing meters.
Thanks for that link. The plots are good, I will try adding some of the meters I'm interested in though it is getting a bit late to do so tonight.
What might be a useful addition would be a fixed percentage error base line or series of base lines - this would show the useful measurement range. For example, if the minimum absolute error is 100 microVolts and you want to have no worse than 10% error then decent measurements won't start below 1mV. So it would be nice to have a range of contours for 10%, 5%, 1% and so on.
-
You can't compare accuracy between meter brands from quoted specifications, you need to actually test them. The reason being that accuracy specs are worst case guarantees, not indications of expected performance. Some meters have an actual accuracy on test that is 10x better than their specification would imply. You will not discover that unless you physically measure them.
-
Yes, what Ian said, plus one additional factor you have not arbitraged. Namely the aged references inside typical used 5 to 6.5 digit bench meters (almost universally an LM399 family member) will be more stable than the new handhelds. Stability has to figured into accuracy as most home labs will not be calibrating their meters very often if at all.