Low temperature testing

While I probably live farther north than anyone else here (57' 08"), our climate is oceanic, not continental so we don't get extremes of temperature.

However, lots of people here do and since winter is coming in the northern hemisphere it is interesting to see what effect it has on the cells we use.

As a preliminary test I put a Romisen RC-C6 with an RCR123 in it into the freezer for a day then measured output (outside the freezer) compared to output at room temperature. The original discussion is to be found from here onwards.

The light is now at just under -22oC (-9oF) having been in the freezer for 20 hours. I've not yet figured out how to get good measurements in the freezer so I'll pull it out and do instantaneous, 30 second and 2 minute output measurements. Unfortunately I can't do these in the freezer but the light should still be pretty cold at 2 minutes. 229, 241 and 290 lux with an RCR123 which was fully charged for this test. It is not comfortable holding a light at -22 degrees.

Here's the comparison: Top row is lumens at -22oC, bottom row is lumens at room temperature.

Lumens Lumens Lumens
Temp switch-on 30 secs 2 mins
-22 41 43 52
+20 125 131 135

I'm probably going to have to build another lightbox to do more thorough testing, one small enough to fit in the freezer.

I don't have room-temperature numbers for CR123 cells in this light. Probably best to do room-temperature numbers then toss the light in the freezer for a day.

Once I've got a smaller box I'll do instrumented runtimes with different cells.

Any suggestions are welcome. I have a small cardboard box looked out and have ordered a small solar cell from DX though I'll look around locally since HK Post is running behind just now.

Then I can do the tests at -22oC which seems to be as low as the freezer wants to go just now.

Thx for the test, its usefull to know that temperature has such an impact on light output. Can you test one ni-mh light too?

Once I've got the small lightbox built I plan to try several lights and batteries. I have D NiCd, D NiMH, C NiMH, C alkaline, CR123, RCR123, Alkaline AA, AA NiMH, Lithium primary AA, AAA alkaline, AAA NiMH, 10440, 17670 and 18650 cells. I plan to try most of these in time. I thik I even have appropriate lights for most of these.

I can probably find some LR44 and SR44 cells and lights to try too.

Thank you, Don, for doing this!

I live 59'22" and try to spend as much free time as possible outdoors including longer trips north of the polar circle during wintertime. Normally I use Lithium primaries (AA/CR123A) in low temperature conditions when I am further away from any power outlet. For general use I prefer LSD NiMH and I have some 18650 lights/batteries but never bothered taking them on a longer trip due to logistic inconveniences (during summertime you don't really need a powerful light up here as the sun beats them all most of the time and when it's less bright you don't want to attract any additional mosquitos).

Therefore I am very much looking forward to your results as it may reveal some useful alternatives to the rather expensive primaries solution. Though your first results with the RCR123 look a little discouraging we don't know yet if the CR123A is really doing substantively better (but I'm afraid it does).

Although it seems that you do not need any additional cells for your test my offer about a pair of CR123A still stands. These are Panasonics from KD w/o expiry date and I am quite unsure whether their performance is as good as it should be so I am hesistating to order more of them despite the good price compared to local/european shops. My offer is, therefore, not entirely altruistic.

Thanks again for your effort,


Not a problem - this stuff is fun to do. It would probably cost more than the cells did to post them here - are they white Panasonic cells?

If they are white they are almost certainly fakes. If you look at the small print on the sides of the white ones they say something like "Elestric" instead of "Electric".

You are a long way north!

Once I get hold of a small solar cell, I'll build a small lightbox that will fit in my freezer and get proper readings over the full runtime of a cell. Since i still have the larger one I can use it for comparisons. I have ordered a small solar panel from DX but I can get an overpriced (but still cheap) one here at the weekend.

The effect on RCR cells may not be as large as the initial results suggest - it is possible that the light got into medium instead of high mode - the plastic ring that holds the switch in place in the cheaper Q3 version of the RC-C6 shrank more than the aluminium tailcap when frozen so the switch was very, very awkward to work and it may have been in medium mode instead of high. The light was too cold to handle so it is possible I had it in the wrong mode.

The difference between room temperature and -22 was much more than I expected so I'd not read very much into the results yet.

I have 20 very cheap CR123 cells that might as well get used up as well as some Duracell CR123s. They didn't cost much either. I am just as interested as you in what effect very low temperatures will have.

Pity I'm not a chemist any more or I'd have access to really low temperatures - I've been trying to get hold of some liquid nitrogen all day - none of us will ever get that cold....

I also need to get a thermometer that can handle lower temperatures - the one I used is only supposed to be "accurate" to -10.

I take it you were not quite telling the truth about your poor vision in the other thread. The battery wrapper of the white "Panasonics" says indeed "elestric"! I didn't bother to read all the small printing on the battery wrapper as most of it appeared to be japanese.

I got suspicious because I could not imagine that Panasonic would produce/sell cells without expiry date and because I thought they performed not as good as expected. When further searching the web I read somewhere (don't have the link, sorry) that Panasonic has stopped producing CR123As in Japan a couple of years ago. So I thought these were maybe old cells, sold without the original packaging to hide their true age.

Now I am glad that your vision is not so bad at all and that I test-ordered just a five-pack!

Sleepy, Huny

Don, I enjoy the tests and data that you present to us. This is a cool test. Being an outdoorsman and quite often going on treks at my cottage during our winters (which easily see -30 to -40c at times) i have wondered how these cells would handle low temperatures. Even the lights themselves!

Thanks Don!

Not my eyesight - I read it somewhere else and used a big magnifier to read the small print. :)

I bought ten of them about 5 years ago - they were not very good. One completely unused one this year actually read a negative voltage.

Is it possible to instrument temperature along with output (ie temp probe on your MM)?

It would also be interesting to see the effects on the battery seperately from the light. IE. the battery should have discharge curves that change with temp, but the light might also change efficiency. If the latter is negligible (or well understood), in theory it should be possible to derive the brightness of the light with only a voltage sweep of light (which you already kind of have with your runtime graphs, if you meter battery voltage with time), but I have no idea if this true. Really cool if it all fits together, though.


In a tangential note to the above, I was thinking with a voltage sweep (and measuring current and output) from a power supply or whatever across the light, and comprehensive data on just one battery (basically a C vs V curve all along its "power/capacity" axis, anything even close should be sufficient) it should be also possible to simulate a runtime graph for a light with very simple numerical calculus (ie possible in excel). Temp axis for battery and light optional :).

Now that would be quite impressive. I'm pretty sure I've never seen anyone even attempt this.

And Don, this doesn't mean you have to do it if it's not fun. ;)

Not simultaneously at present. QtDMM is refusing to see any meter just now. It'll take a bit of fiddling with that to get it to play nice then I have hopes of being able to record two parameters simultaneously. There just isn't room in my small kitchen for two computers, meters etc.

I think that this might be feasible with a lot of equipment and time. I don't have the equipment and doubt I could afford it. As a general rule, most circuits become more efficient with lowered temperature, most cells don't. At least in theory, the area under the curve in the runtime graphs is the usable capacity. The units however, are not readily convertible. I'd need some current sinks with known behaviour at given temperatures. I agree it'd be interesting but I suspect it'd be beyond my resources just now.

If the typical budget light were better standardised there might be some predictive value, but they aren't all that predictable - see the various comments about the new Ultrafire C3.

Interesting ideas though.

I think that this might be feasible with a lot of equipment

I may have a pleasant surprise for you, Don.

The piece that you probably don't have for a "proper" voltage sweep is a variable power supply. You've seen the graphs produced on CPF (V across the bottom, and Current/Output on Y axis). This fully characterizes the circuit fairly quickly, basically limited by how fast you can type numbers into your spreadsheet :).

However, you can do a poor man's version with a battery, and in fact, you already do it (sort of) with your runtime graphs (as long as you have current and voltage)! If you plot V vs. C's from such a graph instead of the t's, it's exactly the sweep for the voltage range of the battery! :D

The real lab supplies are somewhat expensive (>$100), though sometimes you get lucky and get cheap/free ones surplus (similar to old oscilloscopes). However, I was just browsing ebay and they makes these small ones (<2A), for various other purposes. The downside to these is that they bottom out at slightly high voltages (1.5-3v) for our purpose, but I'm pretty smallish ones applicable to this exist.



Again, just chatting about possibilities.

As long as we can produce some sane battery data (this would be very nice since it wouldn't change at all depending on the state of some random battery), the results from a simulation would be highly reproducible.

I may have to think the math through somewhat more though. It may require a bit o' diff. eq. (limit as var x goes to y, blah, blah) instead of just simple additions/deltas, but I'm reasonably confident the system is fully qualified with just the data mentioned above.

... and time.

Well, if you can simulate runtime graphs from a quick sweep with a supply, think of the time saved. A runtime graph in minutes, limited only by your number entry secretarial skills :p

Though ironically, if you get it via a runtime test, you just get to do some calculations afterward to validate the mathematical model is correct (given our normalized "battery") and the simulation reflects what you just measured! :D :D

oh, and one more thing ...

I don't see this feature on QtDMM's page, but if you get an app that supports simultaneous logging from two meters, you don't even need to enter any data--so an almost INSTANTANEOUS runtime graph! Is that an offer you can resist, or does the nostalgic romance of doing the long extended version still thrill you? :D


qtdmm does support my Metex though, but I don't have linux running on any of my boxes.

I should be able to run more than one instance of QtDMM without trouble once I get it to see the meters.

I like the pic!

Nominally QtDMM does support the Peaktech meters I have, it just doesn't want to see them right now. This will get sorted when I have time.

Any spreadsheet then will give me more tools for analysis that I can shake a stick at. The most time consuming part of the data massaging at present is firing up the loud Windows box (I don't normally do Windows, my primary use for the box is to run the not very good data capture software that came with the meter) and dealing with how badly Excel handles timestamped data that goes past midnight - it takes even longer to sort out when the data spans more than one midnight since the software timestamps, not date and timestamps the measurements.

I think there are also issues with the USB-serial chip. While it is only a 12mbit/sec device it insists on USB2 protocols - the laptop I'm using has USB 1.1 ports. I have a USB2 PC card which works but it eats PSUs for the 5V output. The meter seems to need USB power though why it does I have no idea....

The runtime test is still going to take as long as it takes - but at least it is then unattended.

Randall Munroe is wickedly awesome. His humor often parallels the best conversations I've had with colleagues.

BTW, Windows is also awesome ;).