(article continued from previous page)
Robot Room tested a dozen digital multimeters for measurement accuracy and features. The meter tests focused on activities that the average hobbyist robot building would perform, as opposed to residential electricians.
The goal is to determine:
Each result image in this article is from a spreadsheet that includes the measured value followed by a column score from 0 to 3. The score for the category is the average of the individual column scores. Results are sorted by largest score to smallest score. Tied scores are subsorted by smallest price to largest price, but you should consider them to be equal.
Displayed values that were rounded up or towards the correct value are credited in favor of the meter.
For voltage measurement, the tolerance of the Maxim parts was 0.002%. That means column G should be from 4.999 V to 5.001 V to be perfect. Column I should be from 2.4995 V to 2.5005 V to be perfect. The tolerance of the Texas Instruments shunt is 1%. So, column K should be from 1.22265 to 1.24735.
Results of multimeter voltage measurement tests.
What a shocker!
All of the meters provided satisfactory results, with almost all of them being very good or better. Let’s face it; you wouldn’t even be bothered by the 4.98 V reading of a 5 V source. All of these measurements should be colored a shade of green, except then it would be difficult to discern the subtle differences.
The Jameco Electronics BP-1562 was the only meter to score perfectly on all four voltage measurements. It only costs $10.95.
The resistors are 0.05% tolerance. Read that again. Not 5% tolerance (like standard resistors), but 100 times more accurate. So, a perfect score for a 1,000,000 Ω (1 MΩ) resistor would be from 999500 Ω to 1000500 Ω
Results of multimeter resistance measurement tests.
Not quite as good as voltage measurements, but still very good considering the wide range of component values. The only time you might be bothered is when you touch the test probes together and receive a resistance that isn’t zero. This is amazing.
Current was tested by passing the high-precision 5 V voltage reference (0.02%) through a high-precision resistor (0.05%) for a total tolerance of 0.07%. That means a perfect measurement would be between 4.9965 and 5.0035.
A little more leeway was provided in the scoring, such that only two significant digits needed to be accurate to score a very good (2 points). Note that the UNI-T does not include a current measurement mode, and the Innova doesn’t have a microamps measurement mode.
Results of multimeter current measurement tests.
Again, all of the meters had very accurate measurements.
This test would have been better if I could have extended the range above 5 mA, but the voltage reference components cannot supply enough current. Also, I could have taken the time to test 0.5 mA and 0.05 mA.
The biggest problem with trying to create capacitance tests is that capacitors have relatively poor tolerances. Even good capacitors can be off by 20% at room temperature; worse over the entire operating range.
The best capacitors I could find for a reasonable price are 5% tolerance. A value between 95 and 105 is considered perfect. A value between 90 and 100 is very good and is worth 1 point. Any other value is 0 points. Unlike previous tests, the points are summed rather than averaged.
Only half the meters provide a capacitance measurement mode, which is a shame.
Results of multimeter capacitance measurement tests.
For 100 nF (0.1 µF) and 10 nF, the results are generally very good. For 1 nF, only half of the meters provided a satisfactory result.
It would have been nice to have tested values between 100 and 1 microfarad. But, it wasn’t possible for me to locate accurate reasonably priced capacitors in that range. I considered using the high precision voltage source and resistors to charge and discharge capacitors; measuring their actual values based on timing. However, other factors such as ESR, leakage, threshold timing detection, and charge-slope math would have made me unsure of the tolerance and reliability of my values.
Only five meters support digital frequency measurement. (The UNI-T apparently requires a sin wave.)
The frequency tolerance of the clock source is 0.008%.
Results of multimeter frequency measurement tests.
Very good results for all of the meters. (The RSR 01MS8268 doesn’t claim the ability to accurately measure beyond 200 kHz)
Only three of the test meters have duty-cycle measurements. Bummer! This is a useful measurement for robot builders because motor and LEDs are often controlled through pulse-width modulation.
A tolerance for circuit used to generate the duty cycle should be something like:
(low-to-high-transition × (high-to-low + low-to-high + high-to-low) ) ÷ (1/frequency)
(14 ns × 3) ÷ (1/4800 Hz)
42 ns ÷ 208333 ns
A 4800 Hz square wave with 25% duty cycle should have a duty cycle measured value of between 25.00504% and 24.99496%.
Results of multimeter duty cycle measurement tests.
One or two inaccurate results here, but otherwise very good. Beggars can’t be choosers.
This is the talent portion of our multimeter competition.
There is a lot of subjectivity to these next scores. Given that the feature score is half of the total overall score, one could argue that I’m weighting feature set over accuracy. However, since all of the meters did well in accuracy, the richness of the feature set is a reasonable differentiator.
Most features are given 1 point or 0 points, based on whether or not the meter includes the functionality. The number of digits displayed, auto-range dial, and continuity buzzer are considered important enough that they are given 2 points. Note that the lack of a capacitance, frequency, or duty cycle mode has already been taken into account, as meters lacking those capabilities received no points in that accuracy test.
Lastly, a bonus of 0, 1, or 2 points is given for the totality of extra features, and -1 or -2 are taken away for the totality of flaws. In some cases, such as the UNI-T, the bonus and flaw scores are offsetting. That is, it is a calculator size meter (+1) but as such has a coin cell and hard-wired leads (-1). That seems like a fair way of accounting for form-factor decisions.
Results of multimeter feature tests.
Points have been added based on the cost of the meter (lower cost = more points). Shipping is added to the cost of the meter if it comes from eBay, since the assumption is that you can’t spread the shipping cost by ordering multiple items as you could if you bought a bunch of stuff from Jameco Electronics or SparkFun Electronics.
After adding together all of the points, here are the final total results of the digital multimeter tests...
Final results of digital multimeter tests.
The Sinometer VA38 and Protek 506 are great multimeters. I use them every day. But, all of the meters over $100 were used primarily as baselines in these tests, as opposed to true competitors.
The RSR 01MS8268 (Electronix Express #01DMMS8268 $31.95) and the eBay VC97 (comparison shop based on total with shipping) are the best overall winners for accuracy and feature richness of low-priced meters. However, all of the listed meters, including the free and $6.50 meter were surprisingly accurate in these tests.
The only meter I would avoid is the DT-830B, due to poor fitting test probe sockets (I may have received a bad meter). Also, as discussed earlier, do not purchase the kit meter or electrician’s ammeter for standard robot building.
We’ve learned that even the most humble digital multimeter is good for hobbyist work. And, for a modest amount (around $30), you can obtain a feature-rich autoranging multimeter. A wide variety of multimeters from different manufacturers roughly agreed on component values, so you can sanity test your cheap multimeter against your friends’ multimeters if you are unable to justify the purchase of a few high-precision components to act as your test standards.