EEVblog #774 - Low Battery Discharge Testing Part 1

[EEVblog]

Dave shows how to do discharge testing on AAA and AA alkaline batteries, for the specific purpose of investigating how much energy is left under the industry standard 0.8V cutout voltage.
This is an explanation of the test setup, verification, and a sample plot of some data before the long term testing.
The setup consists of the BK Precision 8500 electronic load, the Keysight 34470A 7.5 digit meter, and the Rigol DP832 for testing.
Negative feedback T-Shirt: http://teespring.com/NegativeFeedbackBlack
Energizer AAA battery datasheet: http://data.energizer.com/PDFs/E92.pdf
Duracell AAA battery datasheet:
http://ww2.duracell.com/media/en-US/pdf/gtcl/Product_Data_Sheet/NA_DATASHEETS/MN2400_US_CT.pdf

[BravoV]

All of these are just for the ... half bee dick ... 

[rs20]

Dave! You briefly said "look at the area under that curve, it's tiny" -- on a constant power discharge curve? That only makes (dimensionally consistent) sense on a constant current discharge curve. To measure energy on a constant power curve, just measure the distance on the x-axis: after all, every second that passes at 100mW, the battery loses 100mJ of energy. No need for areas, integration, or that third "battery energy remaining" axis (well, you could indeed have a third energy remaining axis, but it'd be a second X-axis, not a Y-axis).

Yours,
Your local pedant.

PS/ Near the end of the video, it was cool to see that the voltage measured by the meter was counting up for a little while!

[EEVblog]

PS/ Near the end of the video, it was cool to see that the voltage measured by the meter was counting up for a little while!

It's aliens.

[Muttley Snickers]

GDay Dave,

As it seems you are going where nobody has been before, not publicly anyway and as these tests will be conducted over a long duration, is both battery and ambient temperature worth monitoring in conjunction with the other tests, just to cover all bases.

Very exciting stuff....

Muttley

[Someone]

Time to borrow an SMU and do some 4 wire measurements? Looks like the B2962A/B2961A can do it all within one instrument, though costing more than the collection of instruments you're using.

The question for the electrochemists out there is can we discharge the battery at a higher rate down to 1V or 0.9V, then let it rest and continue at a very low rate of discharge to fill in the extremely low power curves? Or does discharging at a higher rate change the distribution of charge in the cells?

[rs20]

Pedant away.

OK then: "pedant" isn't a verb

I sort of knew about the CC curve but I didn't appreciate the subtle difference until you pointed it out. Thanks.
Do you think the CP curve fell to 0V  so sharply because once the battery was empty it's game over? Would a CC curve not have dropped to zero quite as sharply?

A CP curve will drop faster due to the snowball effect that Dave has mentioned -- as the voltage drops current has to increase, conflate that with the increasing ESR of the battery (which then drops the voltage under load, which increases the current further, etc etc), and the closing stages of the battery discharge unfold pretty quickly. Of course, using a CP curve isn't unfair at all since a certain boost converter is the center of this discussion, and boost converters are roughly CP.

Having said all that, even a CC curve is still going to dive very quickly indeed.

[lapm]

Lol, just couple days ago i counted 8 different brands/chemistry AA-batteries in my local grocery store.. 2x varta, 2xEnergizer, 2xDuracell and couple shops own brands...

And thats just big supermarket, culd add one more brand if i go LIDL store, they have their own cheap brand too..

That would be a lot of batteries to test. But then i would need to build active load first.. Maybe that's good excuse for it.

[99tito99]

Hi Dave: How about putting a cap (+/- a resistor in series) to damp the oscillation at the end.  Initial thought is that it should not effect the curve other than to smooth the end.  Yes/No?  Cheers, Mark (same post @ YT)

[rs20]

While we're making requests to Dave, I'd like to see the voltage waveform of those oscillations on the scope when the battery is completely dead -- does it look like deliberate, programmed hiccup mode type thing, or is it more sinusoidal/unstable-control-loop appearance? Does it happen precisely once a second, or is it an evidently random number? See below for why I think it may be deliberate.

Hi Dave: How about putting a cap (+/- a resistor in series) to damp the oscillation at the end.  Initial thought is that it should not effect the curve other than to smooth the end.  Yes/No?  Cheers, Mark (same post @ YT)

Adding capacitance into a control loop doesn't always reduce oscillation; often it'll make it worse.

In particular, what a dummy load is supposed to do when the runaway drop in voltage occurs is interesting to ponder -- a CP load will legitimately keep on increasing its current draw as the voltage drops, all the way down to the point where the dummy load becomes practically a short circuit.  (And for the record, a capacitor wouldn't help with this at all). So, does the dummy load just stay a short-circuit forever? It's an inherently divergent system, is my point -- even with a fresh battery, there are two ways to dissipate 100mW in your load: draw 1.5V at 66 mA, or draw 0.033V at 3A (or something along those lines, I = battery's short circuit current and V = 100mW / I). So "draw 100mW" is actually an inherently ambiguous thing to request of a dummy load, and you can only expect it to find a local solution, not the local solution you were expecting.

With all this in mind, I wouldn't be surprised if the "oscillation" that we see is actually deliberate programming that notices that it has become a short circuit which no power supply can ever expect to deliver any significant power into, so it drops current to zero for a bit to let the DUT recover, and then re-starts regulation in the hope that it finds the non-short-circuit local solution. Otherwise, the load would just continue presenting a short circuit forever, even if you swapped in a brand new battery, which would be a very confusing and unintended behaviour (even if it was legitimately drawing 100mW via the short circuit).

And obviously in the case of a dead battery, it keeps on arriving back at the short circuit solution, so it keeps on pulling back and trying to find the other solution in an endless loop. That's my theory, anyway.

[LabSpokane]

Dave, I think if you keep running that test in various configurations you'll eventually find Jimmy Hoffa. 

[pickle9000]

So what is the lowest voltage boost conv that can be rigged today (actual part)?

Certainly worth mentioning because anything below that point is lost. That could be a related fundamentals Friday.

[owiecc]

BK precision 8500 has voltage sense terminals on the back. Why don't you use them for the last seconds of the test? They should not matter much but if we want to be super anal about the setup...

[Dr. Frank]

Dave,

just  for that video!

It's very instructive on the measurement techniques, especially data logging (with these nice TrueVolt DMMs) and the effect of 2W vs. 4W current / power measurement.

Maybe you'll receive again a feedback from this 'Batteriser' enterprise, what you are doing wrong here..

You just busted their business case very simply, within 60 seconds of sampling time, I think.

And there's a another 'scientific' conclusion from your experiment:

These 0.8V are not only "industry standard" for characterization of battery discharge, it's  the demonstration of the underlying mechanism of the chemical reaction for chemical cells.

The voltage drops off so quickly below 0.8V, simply because the reactants are fully consumed at that point.

Frank

[rs20]

BK precision 8500 has voltage sense terminals on the back. Why don't you use them for the last seconds of the test? They should not matter much but if we want to be super anal about the setup...

Dave is measuring the voltage on the battery terminals, whereas the BK8500 would be seeing the voltage after the drop in the cabling. They're both going to be virtually identical, but if only one of those is right, it'd be the way Dave is doing it. Just because something has "X sense terminals" doesn't mean that that is the best place to measure X under all circumstances.

Edit: Sorry, completely misinterpreted. I thought you meant the BK8500 had voltage output, and you were suggesting to connect the multimeter to that. Using a kelvin connection to the battery would be negligibly better, yes.

[owiecc]

Dave is measuring the voltage on the battery terminals, whereas the BK8500 would be seeing the voltage after the drop in the cabling. They're both going to be virtually identical, but if only one of those is right, it'd be the way Dave is doing it. Just because something has "X sense terminals" doesn't mean that that is the best place to measure X under all circumstances.

If he wans constant power load he needs to measure the power taken from the battery. Right now he has a constant power on the DC load but some additional power is dissipated in the wires. This means the battery loading is increasing with increased current (at the end of the measurement). This means the error is increasing as the voltage goes down. If he wants to see what is happening at low voltage he should have a setup that is correct at that loads.

[alter Ratz]

I think it would be more interesting how much energy is wasted, when old products with a linear regulator or no regulator are used, i.e. the area under the curve from the battery voltage down to 0.8V.

Does it make sense to add a switching regulator to your product or is the amount of saved energy so tiny that the cost and effort do not justify it. IMHO it makes sense, but I have only calculated it very roughly and never measured it. So if anybody has compared both methods please reply.

Best regards,
Bernhard

[rs20]

Does it make sense to add a switching regulator to your product or is the amount of saved energy so tiny that the cost and effort do not justify it. IMHO it makes sense, but I have only calculated it very roughly and never measured it. So if anybody has compared both methods please reply.

Surely you know the answer depends on a case-by-case basis -- if the device naturally works down close to 1.1V and consumes a tiny amount of current, then there's no point; you'll have a very tough time getting an efficient enough switching converter; and whether that's worthwhile depends on how much you and your customers care about changing batteries. On the other end of the scale, if your design uses a chip that stops working below 1.5V, then obviously a switching converter is absolutely essential.

[Fungus]

And there's a another 'scientific' conclusion from your experiment:

These 0.8V are not only "industry standard" for characterization of battery discharge, it's  the demonstration of the underlying mechanism of the chemical reaction for chemical cells.

Conclusion: People who design battery chemistry for 50-odd years and manufacture billions of batteries a year actually know what they're on about!

[EEVblog]

As it seems you are going where nobody has been before, not publicly anyway and as these tests will be conducted over a long duration, is both battery and ambient temperature worth monitoring in conjunction with the other tests, just to cover all bases.

The lab only varies by +/-2degC tops. I could get it to about +/-0.5degC but I don't fancy having the aircon running 24/7

[EEVblog]

Dave is measuring the voltage on the battery terminals, whereas the BK8500 would be seeing the voltage after the drop in the cabling. They're both going to be virtually identical, but if only one of those is right, it'd be the way Dave is doing it. Just because something has "X sense terminals" doesn't mean that that is the best place to measure X under all circumstances.

If he wans constant power load he needs to measure the power taken from the battery. Right now he has a constant power on the DC load but some additional power is dissipated in the wires. This means the battery loading is increasing with increased current (at the end of the measurement). This means the error is increasing as the voltage goes down. If he wants to see what is happening at low voltage he should have a setup that is correct at that loads.

It's not much loss in the cables at 100mW, and that is the highest I'm going, so even less worth it for the lower power measurements.

[EEVblog]

These 0.8V are not only "industry standard" for characterization of battery discharge, it's  the demonstration of the underlying mechanism of the chemical reaction for chemical cells.
The voltage drops off so quickly below 0.8V, simply because the reactants are fully consumed at that point.

Yes, and I hope to get better data on that.
But the quick'n'dirty 100mW test shows you only get a minute or less usage once it hits 0.8V. The chemistry just goes kaput.

[thewyliestcoyote]

Dave,

I had to do exactly what you have done not to long ago for work. I used 2 34470a's. One for voltage and one for current. The load was the BK8500. U ended up having to do some scripting to put it all together to make it all work. The BK8500 is a very good instrument but for some reason the software just sucks. Also the instruments is was over sized for this kind of application. For on this in a minute. I have some interesting findings about the BK8500.

Here is the batter under test.
http://www.atbatt.com/xeno-energy-xl-060f-aa-36v-lithium-battery.asp

The set up is the load set to a constant 10mA. After some time I noticed that it was over ranging the 10mA range of the DMM. This is why the current gets turned down to 9mA. The voltage and current is measured with a PLC of 10. The measurement of current and voltage are time aligned within a couple of ms. The fixture is just like yours Dave. It also was placed it a office where the temperature was not the most stable. During the day the AC in on and nights and weekends it is off. Here are some of the plots from the measurement. I would like to post better plots and the raw data. There is a limit on the file size that can be posted. So attached are some crappy png files that will fit in the rules.

Please note the noise in the current vs time plot. I think this is because of a instability of the load at low currents. It is still with in the good bounds given the specs but still a trap.

[thewyliestcoyote]

I almost forgot the table of voltage vs energy and capacity.

[idpromnut]

@EEVBlog: You should do a 250mW test to make sure your setup replicates the results of the datasheet (as a baseline for the Duracell). It would be a good control for your test setup.