Battery life of multimeters

[theHWcave]

Not sure if there is already something on this. I did search but the forum is huge. If this is an old hat, apologies. 

This was prompted by a couple of recent reviews I did on KAIWEETS multimeters and also by Joe's current review of the Keysight U1282a in his multimeter robustness testing series https://www.eevblog.com/forum/testgear/hear-kitty-kitty-kitty-nope-not-that-kind-of-cat/ in particular about the automated way of testing battery life by reducing voltage until a drop in current shows that the meter has shutdown.

My points are:

• One should not assume that a multimeter only draws DC current. I found a couple that have DC-2-DC converters inside without much input filtering that superimpose a significant AC current on top of the DC.
• If you do check for an AC current component, use a meter with high bandwidth because the switching frequency can easily exceed  the 1 kHz AC limit of many standard multimeters.
• When the “low-bat” indicator comes on, the batteries should be considered finished. Relying on the sudden current drop when a meter finally shuts down doesn’t mean it can still produce accurate measurements in the range from when low-bat comes on to when it shuts down. For low battery tests, the meter should be set to measure some known constant value and the test must check that the measurement is still accurate.
  

BTW, I do have a lot of data on KAIWEETS multimeters because they keep sending me meters for review on my YouTube channel. That is very brave from them because I do check them quite thoroughly and exposed weaknesses in each of them. But they say they use it to improve their products which is a good thing.

To show what I mean:
Case1: The KAIWEETS ST120 draws nearly 3 times more AC current than DC from its battery

The ST120 battery current tested with a Brymen 869s for example is around 8 mA DC from 2 AA cells. Moderately high for such a meter but understandable because it has a permanently on backlight.  But there is an additional AC current of more than 20 mA on top. Together that is more than 22mA, eating batteries a lot quicker than what a simple DC measurement lead you to expect.

That current is at a frequency of about 2.3 kHz which means that many cheaper multimeters like the OWON XDM1041 can’t measure it properly. In the picture, the OWON shows only about 14mA AC

Case2: KAIWEETS KM601 loses accuracy when run at below low-bat voltages
That meter shows the other problem. It used 3 AAA so its nominal voltage is 4.5V and low bat comes on at about 3.56V. In this test it displays a constant 5V voltage. The first picture shows it just slightly below low-bat with the correct (for this class of meter) readout.

The second picture shows it at 2.85V battery voltage just before it dies. The readout is has dropped by 10mV. 

For this meter the low-bat indicator really means you should stop using it and replace the batteries.

[Brumby]

The low battery accuracy issue is something I have understood for a long time - but I haven't gone into the nitty-gritty of meter power supply.  I know a DC-DC converter can be used to power the backlight but it is news to me if it is being used to power the metering circuitry.

Thank you for an interesting detail!

Today I learned something. 

[theHWcave]

I promised @joeqsmith to test the Fluke 101 and Brymen BM869s battery current using my method

On the Fluke101

There is no AC component just a nearly steady and amazingly low 0.74mA DC (with the Fluke 101 DC volts mode) and when low-bat came on at 2.23V the meter continued to display 5V DC correctly until it shut down at 2V

On the BM869s
There is no AC component, but when testing the behaviour below low-bat, the BM869s behaved somewhat unexpectedly. I recorded the BM869s main (DC V) and secondary (AC V) using its USB interface.
For this test the BM869s was displaying 5V which it shows more accurately as 5.0009 in DC V mode but in dual DC and AC V mode which I used here, the main display showed 5.0023V DC and the secondary display showed 0 V AC.

In the graph, BM869s main display is the blue trace and the secondary the green one. The battery current is the red trace.

The low-bat indicator came on at about 6.43V.

The meter kept going showing the same voltages on its main and secondary displays until  5.78V when “1nerr” was displayed with loud and continuous beeping. This beeping caused the current to increase. As the battery voltage continued to drop the “1nerr” display caused a gap on the blue and green trace.

At 4.76V battery voltage the beeping and “1nerr” display stopped and the meter showed a readout but only 4.8174V DC.

As the voltage dropped further the V DC on the main display started to increase and so did the AC voltage on the secondary display. With falling battery voltage, the readouts on the main display went crazy shown 831V at one time and -103 at another. The secondary AC readout had lots of OL but at one time was reading close to 5000V! 

At 2.29V battery voltage the IR USB interface stopped working. Unfortunately that blocked my logging program and so the rest of the test was not recorded.

I am not sure when the automated test terminated the BM869s test in this case. Joe?

[joeqsmith]

Hello Mr Cave.   

To calculate the nominal battery life, I use the ACV mode with nothing attached to the meter.   I use the ACV as I felt it would provide a more realistic idea of what to expect.   

To calculate the minimum battery life, I use what ever is the highest current when selecting the various functions (less beeper), again without anything attached to the meter.

To calculate the cuttoff, I use the current that I measured with the meter in ACV mode and lower the voltage until I reach half that value.  In some cases, depending on the meter's power supply design we may not see that knee in the current.  The BM869s is like that and will ride the battery till it is dead flat.  The low battery indicator may be on and the meter may be throwing up bad data but the user gets to decide if they can continue.   

Because I do not use the cutoff voltage and apply no other signals to the meter when determining the battery life, I can get some idea how the meters all compare.   

The data for the meters I have tested is available on-line.   I believe all of the meter listed with their battery life are still functional.  If you feel one of these meters differs from the measurements I made, let me know.  I can certainly repeat the test using my source meter along with another method and we can compare the results.    I dare say, I pretty much open to anyone repeating any of the tests I have ran and comparing results.  If we are not getting roughly the same results, I think we need to understand why.   

***
Should add that the Backlight current was with the meters set to their ACV modes as well, nothing attached and the backlight active. 

[Kleinstein]

A DCDC converter may draw some pulsed current and this can be decomposed to an DC part and AC part. However the AC part is not loading the battery very much. So the battery life would still mainly be determined by the DC part ( = average current drawn). The pulsed current however makes it more sensitive to the batteries output resistance. With modern alkaline cells and the still relatively moderate current this should not be a big deal. It would be with old style dry cells, but these are useless in essentially any application, except those who really need at least 1.5 V per cell.

Another point with SMPS is that the current drawn likely will go up when the voltage goes down. So the average current drawn may be some 10% higher than measured with full battereis.

For the meters with a higher current consumption it could interesting if the low bat warning come before 1.2 V per cell. For alkaline cells there is not much left between 1.25 V and 1.15 V, but for rechargible cells this can make a big difference. For a low power meter recharbible cells are less relevant.

[theHWcave]

To calculate the nominal battery life, I use the ACV mode with nothing attached to the meter.   I use the ACV as I felt it would provide a more realistic idea of what to expect.   
To calculate the minimum battery life, I use what ever is the highest current when selecting the various functions (less beeper), again without anything attached to the meter.

I had a closer look at the spreadsheet (sorry should have done that earlier) and I agree with your method. I looked at a datasheets of batteries and the minimum voltage used varies somewhat but in general at the end of the rated capacity there is still some useful voltage left (not zero). One hopes that meter designers took that into account when determining the low-bat indicator threshold but in any case I agree with your statement that the user should decide if they continue beyond low bat.
As a small wrinkle, it is interesting to calculate the (nominal) power consumption of the meters and that shifts some of the ranking of your table. The most efficient is the Brymen 27s, followed by the Fluke 101 and the Harbour Freight.  The difference in table ranking shows the impact of choosing the battery type for a meter.

I would suggest however, that meter reviews should at least spot-check the accuracy of a meter at low-bat as a standard (I certainly do)

A DCDC converter may draw some pulsed current and this can be decomposed to an DC part and AC part. However the AC part is not loading the battery very much. So the battery life would still mainly be determined by the DC part ( = average current drawn).

I don't agree with that. The ST120 for example draws 8mA on DC but 22mA on combined AC+DC. Yes, the AC part are pulses of extra DC current. The peaks are probably much higher (I never measured that) but the True RMS value of those pulses add up to 20mA. It is my suggestion that each meter should be checked for an AC component on the battery current or as a matter of fact always use a meter that can measure AC+DC (and has a decent bandwidth) for measuring battery current

I am with you on the thorny issue being able to use rechargeable vs alkaline cells in a meter. All my meters, except the Fluke 101 run on rechargeable batteries. For the BM869s I use a NiMh 9V 200mAh battery and overall battery life is of course much shorter than with Alkaline but unless I do extended data logging, I think its acceptable. But once the low-bat indicator comes on you have just a few minutes left before the "inerr" warning with its continuous beeping ruins your day (and data logging session).

[joeqsmith]

I have been using Lithium-ion in my BM869s.

If I don't have a meter that you feel will give wrong results, you could come up with a simple circuit which simulates what you are seeing when using your BM869s high burden, internal shunt in AC+DC mode.  Basically a meter simulator.   Maybe just a simple CMOS oscillator using the old 4000 series running at your 2.3kHz and driving enough load to get similar ripple.   I'm sure I could replicate what you come up with.   It would also allow others to play along (SPICE and what not) without having to buy another cheap meter.   

For the SPICE group, you could easily supply the BM869s characteristics.    May make for an interesting discussion and we may all learn something from it.   Let me know. 

[joeqsmith]

Maybe just buffer your function generator to get what ever freq/duty cycle you want and come up with your network to simulate your meter.   Again very simple to simulate with SPICE and in hardware.   

[Fungus]

A DCDC converter may draw some pulsed current and this can be decomposed to an DC part and AC part. However the AC part is not loading the battery very much.

Surely they would put in a capacitor to smooth it out. 

[Kleinstein]

With a relatively low impedance battery a filtering capacitor is not very effective. In a handheld device they tend to not put a lot of extra capacitance.

For a battery the main part that matter is the averaged out current or charge flowing. The RMS current is misleading - this is the current relevant for the power lost due to the internal resistance, but not the dischraging of the battery. Unless there is actually significant current fed back to the battery that it can not handle, the relevant current is the avearge, like essentially all meter measure in the DC rangs.

[theHWcave]

If I don't have a meter that you feel will give wrong results, you could come up with a simple circuit ...

I think apart from the BM869S and Fluke 101, the only meter you and I have in common is the UNI-T UT210E. Therefore I gave that one a quick test while in AC-Volts with nothing attached. It does show the problem but not very significantly. The DC current is 1.44mA and adding the AC component just pushes it to 1.58 mA combined. For such low AC currents, there is a noise problem, hence I am feeding it from batteries. To get a clearer picture on the scope I had to apply the 20MHz bandwidth limit and noise also adds a few uA to the measured AC current but not significantly. By turning the UT210E off, I confirmed the primary source is the UT210E. 
You will also find that any beep emitted by the UT210E significantly adds to the AC part but that is understandable and not the problem. 

To see the problem more pronounced you could get yourself a (used) HT118A (and run it through your tests which would be highly interesting and entertaining). Better still the KAIWEETS ST120 (cheaper too). I would mail you my ST120 but I think the postage UK to US would be far more than the meter costs new.