It makes sense to me to start with something feasible and then move on to more difficult projects. How about starting even bigger with a 25 or 30 mm board for 26650 lights or multicell lights and then move onward and inward. The effort would not be wasted as there are plenty of hosts that could use a larger driver. Let’s do the driver circuit and then work on miniaturizing it.
If you hung in there and mulled over your available parts selections and options, would you be able to estimate a reasonable maximum height for a stacked 20mm version? If it seems doable, Im sure that many of us can start taking some of our lights apart and get a few measurements.
That pic certainly does put things into perspective for the 17mm and how difficult a 20mm version would be to design. Sheesh! Although a larger board, does anyone know if the DQG triple xp-g2 driver (I assume is a boost) might offer hints for parts and construction? Im beginning to wonder if we need to start with a fresh clean slate or if we could possibly garner a few hints from other designs. Yeah I know… Im starting to think like the Chinese in “borrowing” other designs. 
I doubt that's a boost driver, 1250 lumens from 3 XPG2s only needs 4.0-4.5A total depending on whether that's including losses from the optics or not.
Here’s one:
C8 long reflectored flashlight.
Pill cavity is 20mm x 7.5mm
I modded it to accept a 20mm contact board with 8 x 7135 driver.
People have reported over 6A in turbo.
From this review and the flat regulation curve, it sure looks like a boost. Thats on a Panasonic NCR18650A… definitely not the best performer in terms of voltage sag and yet it maintained.
[Review] DQG Tiny26650 (Triple CREE XP-G2 R5 CW, 1x 26650) there at post 35, that’s the one, right?
That actually is a boost converter, a ‘misused’ LTC1871
That’s a quite big driver. The 8 pin part near the switch is 5mm long, so the board has to be like ~30mm?
The inductor looks like ~7.5x7.5x5(?)mm or something in that order.
I bet if I calculate the inductor current a little optimistic, take just the next fitting value without some ‘bonus’ and simply ignore the loss of inductivity at higher currents, I could get away with a similar sized one. With the same ‘optimism’ one could probably also get away with a little so8 sized switching FET.
Still takes a lot of space to make that voltage converter constant current.
That was’n a cheap driver either, the main IC alone is like 8 bucks (if you buy only one, of course).
So, you see, with enough space and budget, everything is possible. 
I’m still curious what a 20mm pill looks like, maybe I can find some pics.
Thanks, I missed that. Im just reposting the driver pics to help keep the gears turning.
Thanks for your comments. If you were to throw a ball park number out there, how much would you guess it might cost to reproduce this driver state side in a run of 100 (completed, not as a kit). Not that this particular driver would be one to duplicate, but it would be interesting to have an idea of the cost factors involved and the subsequent viability.
At 100 I’d say 11…12$ for parts & pcb plus whatever the assembly house wants (have to quote them when a prototype with complete BOM is ready.)
At 10 maybe 13…15$ with my free hand assembly*1 
Example RS & Farnell:
Oshpark medium run ~1.5$ (2.5$ at 10) for the PCB
SMPS Controller: ~3.5$ (4.3 at 10)
Power FET ~0.55$ (0.65 at 10) (2 for buck/boost plus some small parts as driver)
Power Inductor ~1.3$ (With a fast switching IC we can make that small now, like XAL7030-332)
Power Diode ~0.25$
Microcontroller ~0.9$
Current shunt resistor ~0.5$
misc parts ~2$ (just a guess, depends on the actual circuit)
————————————————————-
Have I forgotten something?
Overall, even with some cost savings, I guess it will end in the 13…15$ range.
Remember that is just an rough estimate from someone sitting very tired in a Railway station, waiting for his train.
Good estimates can be made when a well performing prototype is ready.
While for me this is all just relatively cheap entertainment (Cinema is 10 bucks/hour, including advertisements ), I am not sure how big the interest in that price range actually is. 100 seems an awful lot to me. Maybe we will know when some real-world performance data exists, but I’ll still like to find a more affordable solution.
*1 Remember: I can only supply partly assembled ‘testing kits’ strictly for evaluation purposes and further developement for free or net cost price and production data for free. I cannot bring anything in circulation that could be considered a ‘electronic device’
Yes, it is a sad and well regulated world we live in.
The price range you conjured actually sounds very reasonable if the product can deliver to spec without overheating. I also think the demand would be there once a performance comparison proved the benefits of flat +3A regulation and a store front were established for distribution. Like you, Id just like to see this come to fruition for the benefit of the community. Do you still think you might take a crack at building a prototype?
I’d like to point out the Zebralight SC600 has a linear/boost driver, not sure if it would help knowing that but just in case
http://budgetlightforum.com/node/27914
Nice…have you seen TexasPyro 20mm Nanjg yet?
Good idea and a great driver. I actually considered that before but havent found a thread where anyone was able to pluck the driver. I have the Zebra S6330, and the driver is firmly glued. vinhnguyen54 destroyed his when he tried to remove it and had to return it to the mfgr for factory repairs. I didnt get much farther with mine. BTW, the S6330 has 3 separate boost drivers on the same board… one per emitter with thermal management that works extremely well. In fact, the best booster I think Ive ever seen for XM-L.
I saw that but its a linear driver. You can buy these adapters and save a whole lot of hassle, although they are a rip-off for what you get. :bigsmile:
Keep the ideas rolling guys. Someone will eventually crack this nut so someone like Dave can have an example to study.
A 26mm 3-4.5A max boost driver is on sale on LPF: The "SXB-2" (Super X-Boost) 3A+ Boost Drive for 9mm 445nm diodes | Laser Pointer Forums - Discuss Laser Pointers | High Power Lasers
Thanks ryansoh. After reading the speculation for XM-L2 and watching the vid, it looks like typically voltage sag (3.6-3.8V and then decline from there) is going to be a problem in developing <3A with this driver. Any ideas how to interpolate?
Also (as you had mentioned) above 3A looks like it will far exceed the ability of a single 18650 to deliver. In fact, as jayrob mentioned, it will thrash a good 32650, although he was pushing 2.3 Amps above 5V.
Do you own one of these drivers?
Have you or anyone else tried this driver with XM-L2?
There’s some great pics of the driver in that thread as well.
And darn you for posting!!! This pic is way cool:
Unfortunately one of the reasons why this driver is not fit to drive LED’s is because direct driving XM-L2’s with a single li-ion can deliver much more current than what this boost driver is capable of. (Although the voltage will drop rapidly due to sag.)
9mm laser diodes have a Vf of about 5.5V at 2.4A, and since LED’s have a lower Vf, I don’t think it will pull too much current from a li-ion.
I don’t have this driver, but Lazeerer’s products are awesome and one of the best. These are perfect for laser diodes but not for LED’s due to the forward voltage that can be either above or under the battery’s voltage depending on sag. A buck/boost 5A driver in a 17/20mm package would be stunning. 
@FlashPilot
I’ll make a little proof-of-concept board at least, and we’ll see if I can get it to a performance level that’s worthwhile the trouble.
(At the moment my project is a 100% duty cycle capable buck driver (=dd capable at lower battery voltage levels). A comparison between the buck, buck/boost and a standard linear driver would be quite interesting for me.)
@ryansoh3
I’m actually not sure how much fun that would be. At the lower battery voltage levels, you would be looking at something like ~11…12A input current. That’s a little much for most single cells, I guess.
I’ve had decent luck boiling glue off, depending on what kind it is, if you destroy it internally in the process it should be no big deal as long as its physically intact
Thanks, I didnt realize lasers had such a high vF.
Ive tried all the tricks, including preheating the light in an oven to 250 degrees F and then heating the circumference of the host with a blow torch where the contact shelf and driver meet. The solder melted on the MCPCB and the SMD’s began floating on the board… and yet the glue held. Freezing and heat cycling had no effect. Zebra designed their lights to prevent them from being copied by their competitors, and it seems that they were rather successful… just ask vinh.
Thanks for keep us in the loop and not giving up. I cant wait to hear of your assessments between the three driver types… what a great idea! Have you started a thread on your buck/dd driver?
Yeah, I have a pack of those coming in…I just like the way his design looks, and like to tinker and whatnot…