Alert reader Bob Snyder sent me the UL rules for using coin cells in products .
There are a number of requirements that must be met to obtain UL approval in devices with user-replaceable batteries. For instance, the device must be marked “Replace Battery With (Battery Manufacturer's Name or End-Product Manufacturer's Name), Part No. ( ) Only. Use of another battery may present a risk of fire or explosion. See owner's manual for safety instructions.”
I have never seen this on any product that uses coin cells.
UL mandates that it’s either impossible to install a cell backwards, or that preventative safety measures of the type I explored last week be used.
In many applications coin cells are used just to maintain RAM’s contents when the mains power is down. UL is very concerned that the battery cannot be reverse-biased when the power supply is feeding memory, so requires that either two series diodes or a diode with a current-limiting resistor be placed between the battery and the rest of the circuit, as follows:
Why two diodes? UL’s ultraconservative approach assumes one may fail. The resistor is to limit current if the diode in that circuit dies.
In most cases the first circuit won’t work; even two Shottky diodes will drop about 0.8 V (as I showed last week), so RAM likely won’t get enough voltage to maintain its contents.
The second circuit requires a resistor that limits reverse current to a UL spec of 25 mA for most coin cells. But how is one to compute the proper resistance? You need to know the battery’s internal resistance (IR) when reverse-biased, and I cannot find any documentation about that. There is some crude published data on IR when forward-biased, and in this series I’ve shown lots of empirical data from my experiments.
One could assume the battery’s IR is zero, which would certainly give a nice worst-case result, but I decided to explore this a bit more.
After discharging about 100 coin cells I’ve found that, forward biased, a CR2032 has 5 to 10 ohms of IR when new, increasing to hundreds as it is discharged. If we assume the reverse-bias IR is about the same as when forward biased, then the worst case situation is with a new, fully-charged, cell, since the IR is then at its lowest point.
I applied a power supply to a CR2032 and measured the current flow when the supply’s voltage was 0.5, 0.75, and 1.0 volts above the battery’s unloaded voltage. The battery was contained in an explosion-proof container. Well, actually, an old coffee can, but it sure would have been fun to hear a boom. Alas, nothing exciting happened.
From that it was easy to compute IR, which is displayed in the lower three lines on the following graph:
I took data every minute for the first 8 minutes, and at 5 minute intervals thereafter. The dotted lines are trendlines.
Strangely, the internal resistance spikes immediately after the power supply is applied, then quickly tapers off. In a matter of minutes it falls to six to eight ohms, very much like my data for forward-biased batteries. The data is very similar for when the power supply was 0.5, 0.75, or 1 volt above the cell’s unloaded voltage; that’s not unexpected if one assumes this really is resistance and not some complexity arising from the chemistry. I have other data I’ll present soon that suggests that while modeling the cells using resistance is a good first approximation, there’s something else going on. However, for this discussion five ohms is a safe bet for the IR when computing the series resistance needed.
The top three curves are the battery’s temperature. Unsurprisingly, temperature goes up with the voltage difference. Given UL’s dire warnings about catastrophic failure I expected more heat, but the max encountered was only about 50 C, far lower than the 100 C allowed by UL rules.
This data is for a single battery so be wary, but it does conform to the IR characteristics I measured for about 100 forward-biased cells.
This leads to another question: to get more capacity, can we parallel two or more coin cells? UL is silent on the subject. I suspect that since their argument is that reverse-biasing a battery is bad, they would require diode isolation. As we’ve seen in this series of articles, diodes eat most of the effective capacity of a cell, so should be avoided.
From a non-UL, purely electronics standpoint, what would happen?
This is a debate that rages constantly in my community of ocean-sailing friends. The systems on our sailboats run off large, often lead acid, batteries. On my 32-foot ketch, the fridge sucks 50 Ah/day, the autopilot another 50 Ah, and the radar, well, at 4 amps we don’t have the power to leave it on all of the time. All of this comes from two 220 Ah six-volt golf-cart cells wired in series. After a day or so of running the systems we have to fire up the engine to recharge, which everyone hates. Can we wire two banks of golf cart cells in parallel? I have heard all sorts of arguments for and against, but many do wire their systems that way and get good results.
What about coin cells?
My experimental data shows that the maximum difference in unloaded voltage for fresh CR2032s is about 0.25 volt. This is true for a single brand and between brands and lots.
With two paralleled cells of unequal initial voltages, the lower-voltage battery’s small IR will discharge the higher-voltage cell rapidly until both batteries are at the same voltage.
LiMnO2 cells have a very flat discharge curve till they approach end of life. Discharge one by a quarter volt and you have lost around 200 mAh of capacity, or about 90% of the cell’s 220 mAh rating. So the battery with the higher voltage will quickly run down to 10% reserves. Most of its capacity is thrown away.
But it gets worse. Once heavily discharged the battery’s voltage is at a knee on the curve and falls rapidly. The one that seemed better, with a higher voltage when first installed, now acts as a load on the other!
They essentially suck each other dry. So don’t put these in parallel.
On another topic, for a long time readers have been asking me to do videos about subjects of interest to the embedded community. As that sailing community would say, “Roger, roger.” The first is Why Did My Board Crash When I Scoped a Node? . New ones will appear weekly. Enjoy!
Jack G. Ganssle is a lecturer and consultant on embedded developmentissues. He conducts seminars on embedded systems and helps companieswith their embedded challenges, and works as an expert witness onembedded issues. Contact him at . His website is.