The s21b is pretty close to being a 21700 s8. Shallower reflector and integrated shelf though
What’s the Lume X1 size? The S21A uses a 20mm driver, I suppose the S21B does as well. The S12 uses a 22mm driver, but for a triple.
Ok, now you have my attention. Looking into that, seems like the S21A has a pill but no fins, the S21B has a fixed shelf but includes fins. Those little fins are surprisingly effective.
I would want to swap to a TIR optic so I need the brass pill.
Hmm, if that $1400 stimulus comes though looks like I will be ordering a few things from Simon. Finally something worth considering as a replacement for my EDC.
I could not find specs on the drivers that come in these lights, anyone know who made them? What current they output or what firmware they use?
Although at this point I should almost just see if DB custom is still around and can be talked into customizing a 21700 host for me lol.
Is the x1 even a product yet? Last I saw it was just a single driver that was made as a proof of concept for a particular light. I did not see any generic drivers being sold or plans to build them?
So a bit green then.
Mr. Texas_Ace, I agree with the need for high power turbo, but please let me explain what I mean. I have two main idea:
- For triple led in small flashlight, it seem to me that even with direct drive, maximum lumen is maybe about 3000 but only last for 30s. My suggestion is that instead of running 60W for 30s, is 40W for 60s better? 40W for a 90CRI xhp70 is still something like 4000 lumen I think!
- After turbo, a switching driver will allow longer runtime or same runtim e with more brightness, and allow you to use led in different configuration (maybe you can use boost and xhp50 or 70 which are more efficient but maybe not as nice tint)
For triple 3535 or 5050 led, if you use high cri, even with fet it is difficult to run much more than 3000 lumen.
For small single battery flashlight, I don't know any flashlight that can run at 90s at turbo (let assume about 3000 lumen). For example, Emisar d4 with 4 high cri led has 2350 lumen on turbo but only for less than 30s (https://1lumen.com/18650-reviews/emisar-d4v2/). With another battery maybe it can do 2600 lumen but less time. FW3A is another example with triple led and it can do about 2800 lumen with lower cri xpl hi led: https://1lumen.com/18650-reviews/lumintop-fw3a-blf-tlf/. For these two, they are smaller than your convoy so I imagine them to run for shorter time at turbo than your flashlight. What is most disappointing is the sustain thermal stable output is very low, only 160 lumen with FW3a.
When I say lume x1 or any switching driver will have longer or same runtime at higher brightness, you are right that sustain total power remain the same regardless of turbo, but this sustain total power is LED power + driver loss. For linear driver, I made a comparison with experiment data from Mr toobadorz to compare between K9.3 and E12. I know the comparison is not exactly fair but the math show that switching driver like lume1 is almost 40% more efficient for using total energy in battery to led, and it provide 1.5x to 2x brighter operation in sustain state: https://budgetlightforum.com/t/-/54052/479 .
I think is is ok to compare because both flashlight are about the same size (k9.3 is slightly bigger and heavy). Even if the experiment is not accurate and we half the efficieny improvement, the lume1 driver is still 20% more efficient than linear.
I am hoping loneoceans will have a new idea because he says he wants to build a small and powerful flashlight. Maybe you can post on the thread to ask him about his plan. I am still using loneoceans gxb172 with my convoy for many years but I think I will want to try a new 'e-switch' flashlight to replace it.
But maybe you are looking for slightly bigger flashlight with more runtime and lumen, then I agree that maybe 60W is better because your flashlight will be bigger to run turbo for longer, but maybe more led could be better, like 7led. I am thinking fireflies has new E07 version 2 coming out but is now in some delay because of lunar new year. It use 21700 battery. https://budgetlightforum.com/t/-/64022/382 The k9.3 and E12 are also bright enough for your need but I think it is too big?
1: yeah, 3k lumens is about the most you can expect and while it does drop off that is mostly due to heat which will effect a constant current driver as well. Cell voltage drop plays a role but less so then heat. I am going off personal experience more then anything, I just know that with XP-G3’s I could get around 90 seconds of turbo before heat became an issue and output would remain plenty bright during that time (started around 3k lumens, forgot what it was at the end). During testing running another test on the same cell would show only minorly lower current draw.
An XHP70.2 would indeed give the output I want at 40W easily but the issue with it is lack of optics. I am a hardcore TIR optic fan, pebbled tint mixing TIR optics can make even sub-par LED’s look great. I only use reflectors in large lights or situations where a TIR is just not practical. If someone knows of a ~20mm XHP70 TIR optic I would love to try it out but the smallest I have found was 30mm.
2: No argument on steady state usage being better with a constant current driver. I have been campaigning for a high quality boost driver for years. I just want something that allows me to have my cake and eat it too. A driver that could handle 60W for 90 seconds and then drop down to 40W would be fine for me and should be possible for the components as well if designed right. I am guessing that the X1 could be run higher as well, Loneoceans tends to run his drivers pretty conservative compared to what we normally do around here pushing things to the limit lol.
It doesn’t surprise me that the FW3A can’t run turbo very long, that is part of why I didn’t work on that project, it doesn’t have enough mass to allow high outputs for every long. Note I said if the light is built right is can do longer. My S8 is much longer and heavier then an FW3A, the tradeoff is that is can run turbo for much longer as almost all that mass and length is converted into thermal management. The solid shelf also reduces turbo runtime as it transfers the heat too quickly, using a pill and isolating the body from the LED’s can drastically increase turbo runtime.
That driver test is flawed and doesn’t directly measure the driver efficiency. An FET driver on turbo will actually have nearly 99% efficiency as it is just directly connected to the LED and the driver has virtually no losses. Now if you are comparing PWM modes then yes there will be a difference but not nearly as much as some people think. I did some of my own testing and found the efficiency losses under PWM to be surprisingly low. Less then a high quality switching driver but not as much as you would think in my testing although there are many many possible edge cases and gotyas in both directions. Overall IIRC it was within 10% on average.
All of this really doesn’t matter all that much in reality though, all my testing and experience really taught me is that nit picking the details about drivers and LED’s generally doesn’t matter much in the real world. I spent a lot of time working with and developing constant current drivers but after testing side by side, realized that the old FET driver worked just fine and in some ways better.
PWM FET direct driver generally nets you a much nicer tint then a constant current driver for example. That alone has been enough for me to just stick with FET drivers in most lights that use LED’s with tint shift over the power range (such as the SST-20 I am using now)
I am not hard locked on 60w, more locked on needing 2500-3000 lumens on turbo. I know that this will require around 60W of power in most cases, hence I would like to see a 60w capable driver.
If you look back at Loneoceans threads I have been one of the biggest proponents of them 
I still am as well. Just waiting for a proper driver to be released to jump in. Knowing LO, I can assume he has a pretty good safety margin built in and I am betting I could bump the power up to a point I would be happy with. Just got to have a driver to try it with lol.
Also have not had a host to put it in until now, the S21 looks like a possible replacement for my EDC. Will just have to see what the pill looks like, if it has enough mass to allow for some good runtimes. Then there is finding a good tint LED and optic combo. I have an optic that I think will work well with an XHP50.2 but have not tested it in the real world yet.
These S21 series are perfect for triples. Either will need a custom spacer to convert to a triple, but the bezel opening won’t block any of the optic unlike the 18650 tube style Convoys. That can equal a lot of lumens OTF just right there!
Does anyone make said adapters?
I would be debating between a triple/quad or an XHP50.2. I guess it would largely depend on what tints are available on the XHP50.2.
I did testing a long time ago by milling out the edge of an S2+ so it did not block any of the optic and I think I saw about a 10% increase in output. I sold that modded light to someone on here a few years back.
Kiriba-ru supplied me with the copper spacer to make my s21a into a triple
Good to hear he is still around, yeah I have gotten basically all my spacers from him.
Has anyone tried the S21B XHP50.2? I could not find any reviews? I am particularly interested in the driver it is using.
Probably this one . I guess then the S21B must use a 22mm driver rather than 20mm as there is no other 12-group XHP50.2 driver in the store. The listing for the host doesn’t say anything the driver size, someone should contact Simon to confirm. The XHP70.2 driver would make more sense if you want maximum output (it looks like it would fit).
So 3A, not great but pretty good for a factory driver at that price point. Too bad it is not using BLF firmware, kinda surprised as Simon generally works with TK for firmware.
All the new 12-group drivers use a closed-sourced Biscotti clone. I guess Simon’s hardware suppliers and open source don’t mix.
Odd, up until now they have all been open source. I wounder if TK did it or if it is literally a clone.
No, TK didn’t make it . This is Simon’s explanation . I don’t really buy the manufacturer claim, they would need to rewrite it from scratch to create a legal clone, quite often developers (illegally) reuse GPL code without following the GPL licence, but it would be very hard to prove without access to the clone source code.
I don't actually feel great about this one. Using an inferior heatpath to get better turbo runtimes doesn't appeal to me. If the driver is still on the pill, then you're still having to really crank the thermal management up if that driver has temperature regulation. The real gain is that you're insulating your hand, so you can keep the light at a temperature that is comfortable (or maybe just not-painful) to hold for a bit longer, at the cost of cooking your electronics a little harder.
IMO the issue (well, for heat anyway) with FW3A is that the shelf is super thin. D4 and KR4 have thicker shelves. FW3A did this intentionally to get the size down of course, and the cost was sustained outputs.
Ok, that is kinda what I figured happened. Yeah, I am guessing they re-used some of the code but almost impossible to prove but then again I have seen china shops insist on writing code from scratch just so they could keep it all in house over used an open source version before.
I felt this way early on as well but after a few hundred light builds and examining the actual temps and tolerances of the electronics, I no longer worry about the temps of the internals that much.
The LED will be the hottest thing by far and is still generally rated to around 150c. Even at higher temps it will last longer then most flashlights are used.
The driver components are generally rated to 105c and rarely get anywhere close to that.
Basically the real world ramifications for causing higher temps internally are very minor for a run of the mill flashlight. Within reason of course.
My basic mentality for heat path on small lights now is making as long of a heat path as possible before reaching the body. This creates time for the heat to move to the body meanwhile the copper can absorb a fair amount of the heat. The actual heat dissipation is only mildly worse but you get much longer before you feel the heat.
Going a step further and putting a rubber coating on the body to protect your hand is the next step and also valid if done properly.
The electronics are perfectly happy running at 100C, your hand on the other hand doesn’t like much above 50c.