the cutoff ckt i use for the BAM is set for the battery pack i plan to use with it 3V/cell is what i use.. now i might 'fudge' it a little for some hysteresis so it doesn't blink on/off.. maybe 2.85V for example.. the resting voltage will typically be about 3.5-3.8V/cell (depending on the brightness level i had the light when it cuts out).
the hotwire driver i'm using is even 'softer' on the batteries.. it cuts out when the batteries can no longer hold more than 92% of the regulated voltage.. i will probably set the bulb voltage to about 0.1V lower than the median battery voltage (i.e. 10.7V on 3xLion).. which means.. 3.40v cutout at the bulb... meaning something close to 3.45-3.5V cutout at the battery (when they are about 95% exhausted anyhow).. the bonus other than saving the batteries.. virtually no dimming.
I have some of the BS protection packs and plan to make some protected batteries out of them.. i will run them down to see what they will shut off at as long as it's higher than about 2.5V.... they most certainly cut out lower than the 3V i use.. i think that it actually would be pretty simple to design a ckt that shuts off the battery using just an FET and a potentiometer.. an N-ch FET 'below' the load and a pot from hot to ground with the wiper on the gate of the FET... dial in the right bias so that when the Vbat drops through that 'cliff' at end of life the Vgate swings below the gate threshold.. what will undoubtedly happen is that the FET will slowly shutdown to limit the current and the Vbat will never get a chance to get below 3.0V..
Example.. 10.8V solution; 2-4V gate voltage... Vbat of 12.7 to 9.0V (3 cells)..
you want the Vgate to be at most 2V at 8.5V vbat... so you set a 100k pot from Vbat to ground such that you get 2V/8.5V or 23.5k/100... if the Vbat goes below 8.5V you will have an FET that is 'off'... when the Vbat is 12.6 you will have 22.2/100*12.6= 2.96V on the FET... well into the 'on' range.. and at 10.8v.. you have 2.54V... you might have to find a more senstive FET (gate threshold of 1V) to make sure that you have full conductance at nominal voltage.. when the battery starts to die at the end, there will be some power dissipation on the FET so it needs to be heat-sunk.
(or.. for $7 you can build the same ckt i'm making..
http://hotdriver.rouse.com ) and it has built-in shut-down at 92% of preset output voltage... you can set the voltage a little more conservatively (say 10.6V vs 10.8V) to get longer runtime.. but in that example i would reverse-engineer the most sensible choice..
A LiON cell is for all practical purposes stone-cold dead at 3.45V.. so triplify (sic).. and you get 10.35.. add 0.1V for the resistance losses.. and you get 10.45V... so right around 10.5V would be an ideal setting for 100% regulation of a mag85 (or other 9.6V bulb overdriven in a '10.8V' solution.
looking on the WA rereater:..
http://bulbrater.rouse.com (love my shortcuts).. one of the pre-fab choices is 10.56V/3.32A.. 3473 K, 1140 lumen.. but 15.9hrs!
considering that a KIU DD solution will get about 1200 lumen average but will put almost immeasurable increase of stress on the bulb.. and considering you are lucky to get 1000 lumen out of a PR based 1185 solution.. i kinda like the 1140 constant for 100% of battery life mag85 solution!
I have the most complete spreadsheet comparing all useful bulbs i can find on the WA website using each battery voltage i can make up using LiON cells.. It's not complete yet so i haven't posted it but it's amazingly detailed.
the 1185 is amazing, no wonder it's 'the weapon of choice' for so many.. but i don't like the wildly dropping initial brightness... with some reigns on that bad boy i bet the bulb life will actually be close to the calculated value and holding the voltage back to like 10.7 still nets you almost 1200 lumen, but this time for like 1/2 the runtime.
The 1154 takes about 48W to output the light that the 1185 will at about 36... but the bulb life as you have noticed.. substantially higher!.. re-rating the 1154 for 14.3V (the FET driver will drop about .1V).. it can output almost 1400 lumen.. 61 hr lifespan.. 3365K.. probably about 3.3-3.35A.
with a 12-pack of R123s in 3x4 it will have about the exact same runtime as 1185 with 9pack.. but of course you can fit the 9pack into a 2D-3 FM host. (saaaweeet version of the mag85!). the mag-85 version of the 2 1/2D is what you described above.. 3x3 17500.. no better way to go.. just over 1A/cell.
The output will drop DRAMATICALLY loooong before the cells become dangerously low.. as long as you know what you are doing you won't risk deep discharging your cells.. i also measure the individual cells to try to match them up nice in the stacks.
I'm trying to get an interest feeler on 14.4V solutions because i'd love to see some really exciting possibilities:
4x18650 in a 2 1/2 D-2 would be the same watt-hour as 6x17670s in 2 1/2 D-3 host... but with 14.4 or 7.2V options it's wide open to neat possibilities.. from 400 lumen looooong lasting solutions to regulated solutions using the 1166 from 14.4 at 900+ lumen or the 1160 running from 7.2V solution at 800+ lumen (100% regulation on either of those solutions).
interesting on the lens top.. that should keep the light from being too 'floody'.. the really big bulbs with big filaments tend to be a lot less spot and lot more flood... i hope to get some stippled reflectors because they are so the bomb and i think the combo will be very nice.. the vast majority of the light comes out the sides and bounces off the reflector.. and the 'flood' portion of the beam is usually what comes out the front, so the lens will hold back some of that spreading light and cause the center to be brighter than it would normally.. it will be very interesting to compare head-to-head the 1185 vs the 1154.. i can't wait to get them next to each other for the showdown.
-awR
ps.. some of the 'gems' i have in the works..
800-900 lumen 1160 in a 7.2V host (2D-3)
1200 lumen 1185 in this host (2 1/2D-3) (also works in the 2D-3 with 123s)
1400 lumen 1154 in this host... with 12 R123s
900+ lumen 1166 in a 1D host (don't try this at home.. over 2.5C on the batteries!)
900+ lumen in THIS host with 9x17500
We have a top-secret very sneaky design in the works.. realistic estimated output 1600 bulb lumen (taking resistances into account).. host.. 2 1/2D and also a 'pushing the envelope' version with the 2D-3 host. Hope to have that prototype running within a month for the people to see we think it'll really blow the minds of the hotwire world.
-awr