Hello everyone!
Will wanted me to share some of my observations from testing related to runtime, moon mode, and how they are effected by cell temperature and condition. I'm also throwing in some other considerations about cell temps and safety along with this:
I've run many different bulbs (64250, HO-M3T, EO-M3T, N1, MN16, MN20, MN21, HO-M6R, IMR-M6), under many different conditions. Sometimes in continuous runs, sometimes broken up into shorter runs spread apart by all sorts of possible resting periods, sometimes right after the cells come off the charger, and sometimes after they have rested for a day or more. I'm starting to get a feel for some of the behaviors of the regulation and wanted to share a bit about those behaviors and why they exist.
Under normal circumstances, a direct drive lithium-ion powered incandecent will suffer a reduced output and runtime when cells are colder. Since the output drops, so does the load on the cells, so the runtime isn't effected as bad as it could have been. Colder cells = lower voltage output = lower load.
In our case, since we are regulating output, when the cells are colder, the regulator cranks up to a more aggressive duty cycle to offset the deeper voltage sag, which puts a heavier strain on the colder cells. With no sacrifice in output, all of the sacrifice must come in the form of runtime. What would normally be a loss of both some runtime and some output all combines to equal just a larger loss of runtime instead. The result is that swings in actual runtime can vary more dramatically than one would be accustomed to in a direct drive light.
The variations in possible temperature and cell conditions impact the behavior of moon mode as well. The point at which moon mode "kicks in" and how long it runs for is related to these variables. Since we can not control these variables, we can not control how long moon mode will last or exactly where it kicks in on a particular bulb. Will and I have spent hours on the phone discussing ways to optimize the low-battery warning system (moon mode). In fact, the default target voltage values that the pack uses to decide when to initiate moon mode have been optimized for each voltage setting based on the most likely bulbs to be used at those settings. Even with all of that effort and customization, and the huge re-working of the program required to make this happen (Thank Will for that!), there are still going to be variations in behavior that are beyond our control.
The exact changes in runtime depend heavily on the load involved. Lower power bulbs will usually experience less impact to their "normal" estimated runtime by percentage when subjected to temperature changes.
With that in mind, on the other end of the spectrum, runtime with the MN21 is the most heavily impacted by variations in cell temperature. Unfortunately, until the next generation of higher performance 17670s comes along, the MN21 will likely prove not to be a practical choice when cold conditions are expected. Room temperature or better cells are really required to get a decent run from this bulb. Then again, if you are using an M6 outside in the winter, it's most likely stored in a warm place in-between uses, considering the typical applications for the MN21 bulb.
When running high power bulbs, like the MN21, the cells will generate their own heat (actually, this occurs with all loads, but the high power bulbs cause much more dramatic heat build up in the cells). This can be used as a benefit, but is also a safety issue if taken too far. When a freshly charged set of cells starting from a cool temperature are used to drive the MN21 in a somewhat continuous fashion, say, 5 minutes straight, the cell temps will rise to a level that can help support another couple subsequent runs if those runs are performed before the cells cool off completely. However, there is a delicate balance here, because the MN21 can actually over-heat the cells if used continuously for too long. My recomendation would be to avoid running this bulb for more than 5 minutes at a time whenever possible. Resting periods between runs needed to cool off cells will vary depending on the ambient temps and such. I've found that a continuous run can be performed without over-heating the cells if the light is periodically run under cold water, but this was only done for testing purposes and is probably not ideal or practical.
The heat issues with the cells driving the MN21 is not really unlike the behavior of the MN21 when powered by CR123s. The exact same limitation comes into play with cells over-heating. CR123s are just as dangerous if not more dangerous than the Lithium Cobalt cells used in the PhD-M6. If you are accustomed to the limitations there, and use the M6 in a responsible fashion when driving the MN21, then I wouldn't expect to see any significant problems. Be vigilant and replace those rechargeable cells more often than normal if they are used to drive the MN21.
Eric