(I typed most of this with Notepad in order to make sure it doesn't get lost in VBulletin oblivion. Not sure how it will come out like... Also, it's 2am here and my eyes are closing, so there's probably all sorts of wrongness here...)
I've been attempting to look into the kind of thing I mentioned in these posts (I won't repeat them again here):
https://www.candlepowerforums.com/posts/1455149&postcount=378
and
https://www.candlepowerforums.com/posts/1459110&postcount=450
The tests I've been doing aren't ideal, because I don't have a device to measure the temperature of batteries. Anyway, I realised there's still a lot that I could be doing, so I've been doing what tests I can along those lines. The only instrument I have in order to produce useful numeric data is an analogue multimeter, with which I could measure flash amps. I also have a finger, with which I could crudely estimate which of my own flashlights was the worst in terms of heating up the battery close to the bulb.
I finally settled on my Streamlight TL-3, which has a design which seems to apply rather a lot of heat to the first battery. Even before the head feels hot to the touch, the part of the spring that is in contact with the battery starts to feel very hot - unlike the G2, which I was initially using for the tests. The G2 actually seems much kinder in the way it distributes heat.
I'll just post the results for the TL-3.
First, the background to the tests:
Methodology
===========
Environment: TL-3 was placed on a few folded pieces of paper with some air in between them,
to provide moderate thermal insulation. I didn't go overboard with thermal insulation, because I was not trying to get the whole flashlight hot - I was more interested in the potential effects of the relative temperature difference between batteries. The ambient temperature was about 20degC / 70degF.
Flash amp measurements: First the battery terminals and multimeter probes were cleaned. Then the current from all 3 batteries was measured. Then the current from all 3 batteries was measured again. An average was taken from first and second readings. Readings from first pass and second pass were very consistent, as it happens. All flash amp measurements were taken when the batteries were cool. After a run, they would be allowed to cool for a minimum of 30 minutes. I let them cool while in the flashlight, to keep things "real".
Other: All batteries were fresh out the packet, and unused prior to initial flash amp measurement. I numbered each battery with a CD marker pen to ensure I kept track of what order they were in inside the Tl-3.
Uncertainties
=============
Ambient temperature measurement: +/- 5degC or +/- 9degF (domestic thermometer)
Timing of runs: +/- 5 seconds (used alarm)
Timing of cooling periods: + 20 minutes (watching telly and drinking tea)
Flash Amp Measurements: +/- .1 amp (parallax / timing)
Assumptions
===========
Flash amp readings are a good indication of the capacity of a cell.
Result Presentation
===================
In the result data below, battery ID 1 means closest to the bulb, ID 3 means furthest from the bulb. The four column headings should be taken to be, in this order: Battery ID, Initial FA Reading, FA Reading, FA Delta. FA Delta was calculated as (FA Reading - Initial FA Reading), so is expected to be negative in all cases.
Purpose of Tests
================
To investigate the possibly uneven depletion of cells in multi-cell, high powered lights.
It is speculated that the intense heat at one end of the light will transfer unevenly to the various batteries in the light. This would mean the batteries are operating at different temperatures. This, in turn, could mean that they deplete at different rates. The concern, then, would be that even if one placed "100%", "matching" cells into a flashlight, that, over time, they could become significantly "mis-matching" cells. This would likely lead to a reverse-charging scenario.
Therefore, when looking at the data, the area of interest is: How do the FA Deltas vary with the Battery ID?
THE TESTS
=========
Test 1
------
15 minute continuous runtime
3 x SureFire 123; Expiry Month 08 2014
DATA:
1 7.1 6.5 -0.6
2 6.8 6.6 -0.2
3 6.6 6.6 0
Comments: Exactly the kind of pattern I was looking for. Note, though, that the FA reading for the 3rd battery hasn't changed after the 15 minute runtime. This implies the flash amp methodology is not perfect (who said it was?).
Test 2
------
25 minute continuous runtime
3 x Panasonic 123; Expiry Month 05 2012 (I think - batteries simply had "512" written on them!)
Battery ID Previous FA Reading FA Reading FA Delta
1 7.2 6.4 -0.8
2 7 6.2 -0.8
3 7.1 6.3 -0.8
Comments: Exactly NOT the kind of pattern I was looking for. Or, exactly THE kind of pattern that is nice in an ideal world...
Test 3
------
15 minute continuous runtime
3 x SureFire 123; Expiry Month 05 2015
1 7.3 6.5 -0.8
2 7.3 6.6 -0.7
3 7.2 6.7 -0.5
Comments: Exactly the kind of pattern I was looking for, but not as convincing as in the first test.
Conclusions
===========
None. Not really enough data to get excited about. I do have some thoughts, but too tired to articulate them at the moment. I'll be interested in seeing what other people make of the results. Might be a few days before I get the chance to read the thread again, though.
Relevant Reading
================
Energizer 123 Technical data - interesting info regarding behaviour at different temperatures:
http://data.energizer.com/PDFs/123.pdf
That information makes it clear that a single CR123 cell will behave quite differently depending on the operating temperature. However, what exactly happens when you put multiple CR123s in series, each at different operating temperatures, is not clear. It's the kind of thing that can be speculated upon, but really I'd rather see result data for that kind of scenario.