If the calculation is inaccurate please correct me.
Damian linked to an example on how to calculate it in his post above:
Many thanks for your feedback all of you.
So far I thought my ideal flaslight would be with an SST40 with most efficient driver possible in the optics, size and budget of a C8+ or M21A (or at most the optics, size and budget of an FT03) WITH options for 10° TIR optics but that seems only possible for S2+ size torches?
I am looking now at your Osram options from a throw/watts perspective. So far I really liked the beam of the CULPM1 in Funtastics video you linked. Especially at 300lm: enough throw and light closeby but low enough closeby so as not to spoil night vision (which I find also more relaxing while riding).
Also (OSRAM CSLNM1.TG & CULNM1.TG 1mm², CSLPM1.TG & CULPM1.TG 2mm² - Flashlight Modding and DIY Parts - BudgetLightForum.com), the osrams are clearly less efficient than the SST40 in lm/w, but e.g. the CSLNM1 bests all those Agro linked to above in terms of lm/watt versus cd/mm2.
I’ll get back as soon as I can if I may. Many thanks meanwhile!
Led4power’s calculation in that link is actually incorrect. It is taking the average of the starting and ending efficiency, and that implies that the efficiency changes linearly with time which is not the case. Even if the cell’s discharge curve was linear, that doesn’t lead to efficiency increasing linearly because the expression for efficiency has the cell voltage in the denominator. A linearly increasing efficiency implies a discharge curve with a concave-up shape which is not what you see with Li ion cells and that is why the calculation is incorrect.
The correct way is to find the time-average voltage of the cell discharge, and divide the LED voltage by that average cell voltage. Which is what ggf31416 did.
You can sort of estimate the average voltage just looking at the discharge curve, or a more precise way is to use HKJ’s data. At the discharge rate of interest, divide the measured Wh by the measured Ah.
Thanks for the feedback! It makes sense that it does not change linearly over time! The rest I’m not sure I understand: I’d have thought that in this case, the led is driven at a constant current and hence the numerator (led Vf) is a constant. Now, if the denominator (cell voltage) halves/triples/etc… than the total value of the ratio doubles/triples/etc…, no? So that aspect aspect it is linear, only the discharge does not happen linear over time?
If by ‘time average’ you mean this: Temporal mean - Wikipedia ; than that seems like the best way to do it, but also a lot of sampling calculations, no? With battery discharge, does the result differ a lot from the “the supposed linear” result?
EDIT: The discharge curve of the Samsung 35E at 7A draw looks extremely linear though (except with drops at the start and end of the safe range wich are however also quite symetrical). So if I only use that battery, I would be quite accurate with LED4power’s simple method, or is it a lot less symetrical at lower current draws (I can’t find data on those lower draws)? Samsung 35E
Yes, cell discharge is sort of close to linear. But I’m saying even if the discharge is linear that calculation is wrong. A quick example of what I mean about the efficiency not being linear with discharge is: with a 3.0V LED and battery going from 4V to 3.8V the efficiency changes by 3.9 percentage points. The same change in voltage at the bottom of the discharge, going from 3.2V to 3V results in a larger change in efficiency, 6.25 percentage points. The end result is that calculation overestimates the efficiency by 3-5 percentage points.
OK, lets outsource the voltage curve integration to HKJ. Samsung 35E . IIRC 7135 loses at least 0.1V even when Vf is below battery voltage.
For Vf = 2.8V we don’t need to be concerned about what happens below that voltage because that’s already the cutoff. We need 1A but we can use the data for 2A to account for resistance outside the driver (that’s arbitrary). We got 11.838Wh/3.327Ah = 3.558V avg. under load, but it can be done without using that voltage, we have 3.327 * 2.8 = 9.315Wh out. 9.315 out / 11.838 in = 78.7% efficiency or 21.3% loss. 2.8V / 3.558V gets the same number. I know I’m neglecting significant figures.
For Vf = 2.9V, lets assume that we lose a constant value of 0.1V in the unregulated region, we have 3.2Ah above 3.0V and 0.127Ah below. The regulated region contains about 11.5Wh, 11.5/3.2 = 3.59V. So we lose (((3.59 - 2.9) / 3.59) * 3.2 + 0.1 * 0.127) / 3.327 = 18.8% loss (with many sources of errors).
For Vf = 3.3V I think the regulated region ends at 3.4V, have 2.35A, 8.5Wh above = 3.62V avg, 0.977Ah below, so (((3.62 - 3.3) / 3.62) * 2.35Ah + 0.1 * 0.977Ah) / 3.327 = 9.2%. I took the wrong data as the current is about 3A rather than 1-2A but that correcting that will only increase the efficiency.
The point is that linear drivers are quite efficient when the Vf is about 3.3V or higher, for example when running a SST40 at 4.5A, but they lose efficiency at low current where the Vf is lower (overall driver+led efficiency still increases).
That’s the first chart with many colors, 2 lines for each load as there are 2 samples.
Note that in the case of Convoy the CULMP1 driver should also be suitable for SST40. OTOH the linear / regulated FET drivers are also suitable for CULMP1 if you feel that the 8A driver mode spacing is inadequate, but they are less efficient.
You can request such changes from Simon, I’m not sure what’s the current procedure for the price adjustment, but you can contact him first.
The 1mm^2 is better for throw at the cost of lumens, but check whether the hotspot is big enough. The difference in lumens will be maybe 10% at 1A, but it will grow to over 20% at 3A. The ramping driver is not ideal for it as during adjustment it will be at 8A for some seconds, it will probably survive if your sample is as resilient as the one tested at BLF, but I wouldn’t recommend it (or use a battery below 40% to adjust it). Running the 8A driver with the 1mm^2 at 100% is a bad idea as well.
I think the 20.5mm LD-29 buck driver might fit in one of the Convoy 21700 hosts, or just take the stock 5A/6A regulated FET driver and accept some efficiency loss. There is a 7135*3 driver with 7 modes between 10% and 100% but the perceived changes will be small and leaves you with no options when you need high output.
Could also change the sense resistor on the convoy configurable “ramping” driver to make it a lower current so we not to damage an LED during adjustment. Its easy customizable modes make it great for when you really want to control your runtimes.
I don’t think the OP is comfortable with changing driver components. I’m not sure if Simon is willing to do that.
Edit: to that list of some throwy flashlights add the “Noctigon KR1”: Noctigon KR1 review | EDC Thrower flashlight with 1300 lumens | 1Lumen.com . The runtime was 170min at 275lm using 3000mAh battery. Not really sure how the efficiency compares to other models. There is also the Noctigon K1 but probably outside your budget.
Also see Convoy S2+ Comparison (Runtimes, Brightness, Throw)
Thank you all so much for your imput! I needed some time to digest and research it all a bit.
My impressions are that the osram 2mm² is at the limit of minimum beam width that I’d need.
I am considering the osram in an M21A it has 1.5x more throw. I’m asking Simon for a ramping version.
Alternatively the osram with a buck driver in an M21B, or else the M21A with an SST40 and ramping.
My D25S arrived so I’ll get some batteries for it and try that out first before I decide.
Meanwhile many thanks again for your feedback!
M21A is a 21700 C8+ with 44mm head. M21B has 35mm head. M21C has 55mm head. Of course the S2+/S21A are much smaller.
Bigger head = higher throw.
Yes I know. The 8A driver is 22mm and hence doesn’t fit into the C8+/M21.
That’s why I’m considering the M21B (M21C, L21A are a bit too heavy/large as a bike - head light; the Acebeams etc beyond my budget 
)
I don’t think Simon has any high current buck drivers?
The 8A CULMP1 driver is a buck.
Edit: disregard my ramblings, the M21B with Osram 2mm^2 is looking like a good option, 40g lighter too. I don’t longer think the SFT40 will be a real improvement over the Osram.
Original post:
If you are unsure between the SST40 and CULPM1 because the Osram is too small and don’t plan to purchase soon you may as well wait until the end of March, when the SFT40, a dedomed version of the SST40, is expected to be released.
The old XPL HI would be suitable as well with a 3x7135 driver (a 8x7135 will have poor regulation) or the regulated FET drivers at low current if Simon is willing to build it, and should survive the ramping driver because it won’t go above 5 or 6A with Li-Ion voltages, but the only advantage would be the choice of warmer color temperatures, due to its high forward voltage the output will will reduced when battery is low. However the CULPM1 in the M21B reflector should be somewhat similar to XPL HI in the larger M21A/C8 with better regulation. Both will output about 400lm at 1A.
The advantage in output of the SFT40 will be mainly at high current, at low current the main advantage will be the low Vf allowing better regulation or higher efficiency with a buck driver.
Great summary of options! The 3*7135 will have PWM, no?
It may be a good idea to wait for that SFT40 (I was reading up on dedoming sst40 but without much confidence: reading about faulting led/bondwires )
I’m awaiting Simon’s response and meanwhile I will experiment with the S25D: I’m looking for smooth reflectors or throwy TIR options for that light.
Many thanks again!
Disregard my ramblings, the M21B with Osram 2mm^2 is looking like a good option, 40g lighter too. I don’t longer think the SFT40 will be a real improvement over the Osram at low current, the throw will be 1.4x lower than the Osram because of larger die size, maybe a small improvement over the XPL HI with lower Vf. It will be between the SST40 and the CULMP1 but the CULPM1 in the smaller M21B reflector will be in between too.
Edit: between both Osram 2mm^2 LEDs the CULPM1 will be more expensive but the bin is higher (1690lm vs ~1400lm at 6A but I’m not sure how it will translate at low current), it will also handle 8A within specs instead of 6A ( but the CSLPM1 should still survive the brief time at 8A out of specs from the ramping driver but I’m not sure if Simon is willing to risk it).
Edit: look at this youtube channel , he compared several Convoy flashlights, at least look at this one.
Edit again: about the steps in the ramping driver, from EasyB Convoy configurable constant current driver discussion
Thanks again (it’s been a while - had much work in the house).
I think I’ll stick to the ramping driver (only fits the C8+ apparently) and wait for the SFT40.
Meanwhile I still have to order 18650 batteries. Since the Samsung 35E’s are sold out most of the time, I’m thinking of buying Liitokala’s.
Both my Lii-100 charger and C8+’s ramping driver (and Sofirn D25S?) should have the necessary protections, no?
Are there in my user case reasons to buy or not to buy Liitokala protected/unprotected and buttoned/flat topped?
Will the Convoy C8+ fit with protected button topped Liitokala’s (69.5x18.5mm) vs unprotected button topped (67.5x18.5mm) vs flat topped (65x18.5mm)?
I suspect they wont fit my Sofirn D25S, whose tube I measured as 73x19.5mm, but the tube’s lenght is reduced by springs of 7mm and 4mm. Am I right?
Which model should I chose and should I buy from its official store or any high rated ali store? I guess it’ll be unprotected in order to fit the D25S?
Cheers and goodnight (at least over here )
The SST40 ramping driver is available im both 17mm (C8+) and 22mm (M21B/M21C). Also there is an adapter from 17mm to 20mm (M21A).
If you can buy from USA consider the sale on the Molicel M35A , untested but they make good cells.
If not maybe the Vapcell M35
Just put this in my C8+/LH351D 3500k build. Not a bad driver so far! Anybody know if it will tolerate 2xCR123 at all?
I have no idea, but it won’t get any brighter than at 4.2V if it survives and all the extra volts will be wasted as heat. Also several amps from a CR123 is a bad idea. Trying a single CR123 if a 18350 tube fits at low current for emergencies would make more sense, but it would hit LVP fast, using high current with primaries gets very expensive anyway.