Nice work Tom, however the modes get implemented LFP has come up with a really nice driver.
You can calculate how much nanjg will produce at 0.09A : (0.09/0.16)*44=24.75lm
Now you can calculate gain of LD-1 : 34/24.75~37% more brighter at 0.09A.
I updated the OP with info on using it in a Yezl Y3 (tailcap button and side e-switch).
Nice testing of the e-switch modes!, you really made a good effort of testing the driver. I'm still away for the holidays and back next tuesday, I will think of useful tests leftover to do.
Driver mounted like that has almost no thermal transfer,I recommend you some kind of potting;if you don't like permanent potting,thermally conductive silicone sheets and "cubes" (http://www.fasttech.com/products/0/10003973/1285500-thermal-conduction-silicone-rubber-cubes-1105cm10-) will be better than nothing.
That's correct,there is a slight delay when you change modes in UI_1,because of -double click to decrease mode- ability.
Driver waits for short period of time after e-switch is pressed to see if there will be another e-switch press.
I can change the code for instant mode changes,but there will be no more double press for mode-- ability.
I'm curious about this, because this is how I typically do it, assuming the - wires connecting between the PCB's would hopefully be sufficient for thermal conductance. I was really thinking this isn't much worse than typical ways a driver is mounted, because they are usually just mounted on a thread of a ledge, and maybe 2-3 solder contact pts, if there a brass mount or copper mount available. If the pill is all aluminum, options are usually worse - I've found if you can press fit the driver in, it usually it is not electrically making contact, so even though it looks like the dirver is sitting on the shelf making contact form the copper ring on the driver to that small ledge, it's not really making contact at all, so you have to jam in some copper material to get a good electrical contact, but then, the thermal transfer is limited to just the contact surface doing the elctrical contact.
Would simply adding more wire or copper material connecting between the two driver grounds (battery neg.) be better? I know O-L seems to prefer something like that on his high amp 7135 based drivers. I believe he sees a lot of thermal sag with simple stacked 7135's, but if he spreads them out to another slave board, and adds copper heat sinking, it reduces the heat sag.
Or could I simply solder on copper material to the driver outer ring, adding mass, and adding more surface area to better dissapate the heat?
More on Modes
I'm getting quite used to the 3 groups of modes now with the e-switch control, and their are some pros/cons as any light, or any mode set would have - personal preferences vs. real issues. Only issues that might be real issues so far:
- ramping between 50% and 100%, down or up, is difficult to see going on - much more noticeable in the lower 50%. This was responsible with my initial confusion of how ramping worked -- I just didn't notice ramping was working at all when in that higher range.
- when doing the 7 sec hold to switch mode groups, if you are in a low or moonlight specially, the blink that occurs at 7 secs is hardly noticeable - every so slight flicker if you really concentrate, but very easily missed. This probably gave me a great deal of problems initially, but now that I know, it explains a lot.
- I keep triggering the temp threshold setting by doing 1/2 clicks on the power switch in sequence -- this is very frustrating... Not sure how to avoid this. It's difficult to test the functionality of the power button because every time, I trigger this temp threshold setting.
- The blinks, count, frequency, meaning, etc. gets very confusing. I see them too often and not sure what I did to get them. Certainly one reason is the temp threshold form the issue above, but not sure if there's more to it, or if that explains it all. Maybe for the proto units, the temp threshold is ok, but I'd consider disabling this user setting once in production, or implemened a different way.
led4power explained it pretty clearly how these mode groups work. The 3rd mode group is your standard 3 or 4 modes, like in group #1, 4 if you have moonlight enabled.
-ramping will be fixed easily,increasing speed in 50-100% area,and decreasing in 0-50%
Based on all tests and measurements,I think I will remove user temperature threshold and turbo,and put timer on high option instead. Since turbo is only slightly higher than 5Amps with xm-l2,it makes no sense(if someone still wants that 0.5-1Amp extra,it can be done by stacking resistor,all other modes will be also higher by same percentage-10%-20% which is not much,so no problems),and without external temperature sensor with good thermal connection with led star or flashlight body,it's hard to implement good and accurate overtemperature function(driver overtemperature-100-120Cstays).
Good example is Tom's way of mounting this driver in Y3.It practically floats in air,with no thermal connection with flashlight body,so it measures its own temperature,not flashlight.This is why user overtemp protection without external sensor is problematic.
So,with user overtemp removed,there will be no more 10 or 15 clicks to change user temp threshold,so no more problems with activating those functions with half clicks either (I thought that 10+ fast half clicks is enough to prevent that,what do you do with your flashlight Tom
).
Also with e-switch only one changeable function remains -UI change,I will increase time threshold to 10-15 sec for that.
Now different blinks problem,I have an idea:we should set some kind of standard for all custom drivers when we talk about warning/info blinks.This way it will be easier to remember what each blink means.
For example:
low_voltage_1_threshold_blink - 5 blinks,3Hz ->this is for higher (3V for ex.) voltage threshold,drivers still works but power/current is restricted
low_voltage_2_threshold_blink - 10 blinks,3Hz->empty battery,driver shuts down (2.8V for ex.)
high_temp_1_threshold_blink - 6 blinks,1Hz ->user threshold temperature(desirable max. flashlight temp) is reached,driver decreases current
high_temp_2_threshold_blink - 3 blinks,1Hz ->driver threshold temperature(set in firmware) is reached,driver decreases current
Of course,if driver doesn't have overtemp function,it wont use blinks for overtemp.
All sounds good. Again though, not sure if all my so called "issues" are "must be fixed" issues. Hoping other testers will chime in with their impressions.
For example, increasing the mode group change from 7 secs to 10-15 secs - some of these e-switch lights aren't good at holding that long.
Also, the elimination of turbo mode will negate high amp options for triples, unless the resistor is replaced like you are saying. Think you mentioned it before, maybe over 10A is not a good idea to use with this driver when in a triple build (triple - meaning 3 LED's in parallel)? This is an example Richard sells: mtnelectronics - triple S3, and he currently max's out at 5.8A, but others have gone much higher, maybe in better hosts. It's pretty insane driving 5.8A in a 18650 tube light, but some guys want this for 30 secs or so 
.
For the Y3 piggyback mount, I agree it's not great, but also not many options - many e-switch lights have to be done this way, like the Convoy L4, Warsun's, plus a lot more - basically any light that has the e-switch mounted on the driver PCB. I've talked to two Electrical Design Engineers I work with, and both thought the gray thermal pad material, if simply applied to the board, would not help if not used to bridge to a heat sink, but actually may hurt. I'm still thinking building up more contact between the two boards on the battery neg path (i.e. the grounds rings) would help, and/or soldering maybe copper to that ground ring to act as a heat sink by increasing the surface area. What do you think?
As I said before,idea of turbo was NOT to drive triple xm-l2s or xp-g2s,but to give max. current to SINGLE xm-l2 (or xp-g2,but you can't get 5Amps to xp-g2 with batteries,so it doesn't matter turbo on-off for xp-g2).
I'm not sure about 10Amps(stacked 10mohm resistor),with triple xp-g2 maybe,because of higher Vf,so power disspation is lower,but triple xm-l2, I don't think so.
It all depends on how much power is dissipated and how good is thermal transfer from driver mosfet to case,and all that depend on many other factors.
So if you really want 10Amps,you could try,but at your own risk.
Tom,I don't think that bridging ground rings would help that much.Best way to improve heat transfer is to thermally connect plastic top side of the mosfet to pill,by filling the pill with metal(alu,copper) as much as possible and then use 1-2mm silicone sheet,something like this:
http://www.ebay.com/itm/180930972501?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649
Seeing an excellent application for this current regulated driver.
I’ve read the entire thread through, trying to get an understanding as to what this driver does and doesn’t do. And for me, I see an immediate benefit in a particular light. I have a Sinner Cypreus EDC 18350 light that I just put a triple XP-L in. With the BLF17DD it pulls insane numbers, like 9.3A for 2960 lumens. Too much, too small of a light, too small a cell to push like that. Obiously the lower modes work, but as I now see are not very efficient.
This driver would bring the total output down, while regulating the lower modes to have much better efficiency. And on a light like this the lower modes are the ones that would mostly be used anyway. With the entire light being copper, thermal transfers will be optimized. Run times in low modes will be excellent and reliable. Seems to me like this is a tailor made driver for this type of EDC light.
Am I missing anything? Probably am, so please somebody tell me if there are more pros/cons to using this driver in this light.
Can’t wait for it to become available, really would like to have this in my favorite EDC.
Dale - I think you are right, it is a very good application of this driver, but somewhat depending on how the production driver turns out. If you have a 5A limit, sounds like it will be possible to tweak that limit by changing the R010 resistor, maybe making it a bit higher or even lower. Or if he keeps turbo mode available you may want to leverage it's use, but not sure how high the amps will be exactly.
The efficiency at the lower modes though is a great feature for lights that will run mostly on the lower modes, but if the highest of the low and medium modes is 1.1A, that would be 1.1A shared by 3 LED's, so 0.37A each per LED. I'm not sure of the hardware design limitations, if he's stuck with specific amp values for the low-low, low, and med modes, otherwise it owuld be nice to get choices for the low-low, low and medium modes in firmware or an advanced UI. Of course true regulation at 0.37A is much better than a PWM 0.37A that does 0.37A in spikes of 3.1A (9.3/3 parallel LED's)...
Thjis is all wrong!
An important thing to understand about this CC (Current Controlled) driver is the modes are not a percentage of high or turbo, but are fixed amps. So changing that R010 resistor has no effect whatsoever on the amps of the lower modes. But of course the big advantage of this is as the battery drains, losing voltage (increasing resistance), amps for high mode will be dropping, but, the lower modes will still be the same much, much longer.
Still seems ideal, and if the triple XP-L isn’t perfect for it, then a single de-domed XP-L under a smooth reflector will be 
Gonna be a win, either way 
No,changing sense resistor would affect all modes by same percentage.
So,if you increase high from 5 to 6Amps(that's 20% increase),all other modes would be increased by 20%.
Ahh - oops, will correct above!!
You have to say to yourself though… If I was in prison, which flashlight would I want to have under my pillows. Plunger with a big sink called bam bam
I’d rather not have.
TomE, be a sport and take one for the team, see if the driver can do 10A :)
led4power, what would be required to make this driver reliable at 9-10A? How much would that increase the cost?
I'm sort of confused at this point... Triple parallel LED's should be no different whatsoever to a driver. You are just running each LED at a lower current (non-turbo), so for high mode, take 5A / 3, and you have current per LED. The LED's operate more efficiently with a lower Vf - everything is fine. If in turbo mode (effective DD), you will be able to achieve higher total current because of the lower Vf -- this is where it can get dangerous in theory... If my max amps is ~5.4A now to a single XM-L2, it could be much higher to a triple XP-G2, or triple XP-L.
Also in theory, driving a sinlge LED with 3-4 cells in parallel, in theory you should be able to get a little more amps to the LED than you can from one single cell. Again, the driver doesn't know or care.
For the driver heat sinking issue -- I'm also confused now. It's the components that will get hot -- which components? You mentioned the MOSFET, so that should be the focus... Adding contact surface to the ground ring may help some (standard pill mount method), but it's probably not very effective. You really need to heat sink the specific components, and maybe the solution in post #53 would work well for that. I guess we need to come up with some sort of guideline or criteria of at what amp level we need to heat sink at.
Heat is problem,nothing else,so if you want 10Amps,just like in other electronic applications mosfet should be mounted on "heatsink".In flashlight world that probably means mount it on the bottom of pill(but it must be electrically insulated for body).