TLF/BLF/Lumintop FW3A review (18650, 3x XP-L HI cool white)

edit 2019-01-31:
added stepped ramp default (78 lm) runtime
added a comparison of hotspot intensity with different optics
measured maximum output with 50% drained 30Q (1980 lm @ 0 sec, 850 lm at 30 sec)
filled the CRI table with rest of the modes
added tint vs. mode CIE graph
Measured standby current at a negligible 28 µA
Added thermal image of the light running on turbo
added optics beamshot comparison TLF/BLF/Lumintop FW3A review (18650, 3x XP-L HI cool white) - #44 by maukka
edit 2019-02-03:
added beam angle comparisons between three Carclo optics (10508, 10511, 10507) TLF/BLF/Lumintop FW3A review (18650, 3x XP-L HI cool white) - #62 by maukka
edit 2019-02-10:
fixed an error in calculation for the throw of 10507 and 10511 optics

Disclaimer: The FW3A was provided for testing by Lumintop free of charge

Note: this light is from a pre-production batch and might not be identical to the final version. For example, this light came with cool white XP-L HI emitters, even though only neutral white will be available after the official launch. Also the optics used is the 10508, not the narrower 10511 as planned for final version.

The FW3A is a compact EDC flashlight with some unique features. The concept was originally designed by Fritz, but it became a community effort at BLF and TLF. After almost two years, it has reached production status. Some original ideas were modified during the process, but it still has several points to justify its existence.

Main features:
–3x LEDs (Cree XP-L HI ~5000K CRI70 or Luminus SST-20 4000K CRI90)
-Carclo optics, frosted 10511 was chosen as the default edit: this sample has an even floodier, probably 10508 optic
-Compatible with 18650
–1x7135+8x7135+FET three channel driver with Anduril firmware
-Electronic rear switch
-One of the smallest 18650 triples on the market

If you’re here only for the numbers, here they are (missing data coming later):

since the light is direct driven on turbo, the output depends on the internal resistance of the battery used. Samsung 30Q was used in all of the tests. Several measurements also depend on the user configurable thermal limit (set to 55°C in my testing)

Measured dimensions and weight
Length: 92.5 mm
Head width: 25.5 mm
Handle width: 21.5-25.5 mm
Weight: 52 grams plus 47 grams for an 18650 battery

Box and contents

My light is not the final product you’ll be buying and as such doesn’t have the regular accessories. It only came with a pocket clip.

Physical appearance

The FW3A is exceptionally well balanced and slim for an 18650 light. Even though the finish isn’t raw/clear aluminum as originally intended, the design is very attractive and minimalistic.

The electronic rear switch is one of the special features. The distinctive clicky feel of it is great and requires no precision as it will activate even when pushing at the very edge. The light tail stands stably since the switch is recessed just enough.

The three emitters (XP-L HI cool white) are behind a frosted Carclo 10511 optic (Narrow Spot Frosted) edit: probably 10508 in the review sample which makes for a floody (2.8 candelas per lumen) but very even beam. For more throw, it’s easy to swap in a clear 10507.

Even though the tailcap is removable, the battery should be changed by opening the head. The tailcap with the switchboard isn’t retained in any way and the insides will drop out if you’re not careful. There’s an inner cylinder that conducts the rear switch clicks to the driver. The setup bears some similarities to the Olight M2R even though the Olight has a two stage e-switch.

The driver has six 7135 linear regulators visible on the PCB backside. There’s two more on the other side. The spring is non-metallic (BeCu?), short and stiff. I recommend only using unprotected flat top batteries. Even though very short button tops might fit, there’s some possibility of denting them. I noticed some marks on the negative end of my button top 30Q batteries which are only 66.3-66.4 millimeters long.

The bezel comes off easily. A glass lens protects the optics. The three emitters are in parallel on a 20mm MCPCB. This makes it an easy swap for any emitters you’d like.

User interface

The FW3A is operated via an electronic rear switch. It uses ToyKeeper’s Andúril software.

The Andúril UI is very intuitive in basic use. Give the light to someone and they’ll figure it out right away. There’s even a more simplified muggle mode in which the output range is limited with turbo and all the blinky modes deactivated. The light is ready to go with the stock settings, but it offers lots of things to configure to you liking. Such as setting the minimum and maximum points at the end of the ramp and configuring the temperature limit. In addition to the smooth ramp there’s a stepped ramp with however many steps you prefer. Default is 7 + turbo.

ToyKeeper recently wrote operating instructions for the UI which can be found here:

UI graph

Source and additional information:

I recommend calibrating the temperature sensor when you get the light. It’s easy to check how accurate the sensor is by doing the following when the light is at room temp:
From off

- 3 clicks: light is in battery check mode

- 2 clicks: sunset mode

- 2 clicks: beacon mode

  • 2 clicks: temperature check

The light will blink out the current temperature in degrees C. If it deviates several degrees from your ambient temperature you should calibrate it by:

- 4 clicks: light is in calibration mode indicated by a blink and buzz (fast strobe)

  • Click once per degree, e.g. if the ambient temperature is 24°C click 24 times

After the calibration mode you can adjust the maximum temperature. Temperature limit configuration mode is indicated by two blinks and a buzz. The default temperature limit as in where the lights starts throttling output, is 45°C. To change that click once for every degrees over 30°C. For example, to set the limit to 55°C, click 25 times. That was also used when making the measurements.

With my setting of 55°C limit the light always stabilized at 45-48°C after the initial peak on higher modes. This is still hand holdable.

Size comparison

Next to other multi emitter lights. The Emisar D4S uses a 26650 and the Fireflies E07 a 21700 battery.

FW3A, Emisar D4, Emisar D4S, Fireflies E07

Beamshot comparison

Note: tests done with what is most probably Carclo 10508 optics. This is floodier than the 10511 that will be used in the final production light.

Photos exaggerate differences. Your eyes will fairly quickly adapt to whatever color temperature the light is. The white balance in the camera was set with the Emisar D4 as reference.

Individual beamshots
E07 SST20
E07 219B

Individual beamshots
E07 SST20
E07 219B

With the floody optics, the FW3A is not a thrower. This is a zoomed in photo.

Individual beamshots
E07 SST20
E07 219B

Beam and tint

Thanks to the frosted optics, the beam is very even. There’s no artifacts or discoloration, just pure cool white goodness, or neutral white when it actually comes for sale. Lower modes are objectively a bit greenish in tint, but visually that didn’t seem to bother me.

I’ll add more data when I have time to test with 10507 (clear narrow spot) and 10511 (narrow frosted spot) optics.

Here’s the effect of different Carclo optics to throw. My sample had the 10508 installed by default, but the production light should have the 10511.

With the 10508 the variation in tint within the beam is non-existent.

Beam angle comparisons between three Carclo optics,
10508 (Medium Spot Frosted) - This was installed in my light when I got it
10511 (Narrow Spot Frosted) - Default optics in the production light
10507 (Narrow Spot Plain) - Clear optics with throwiest beam but more artifacts

Throw comparison

Color temperature shift within the beam. Limited to 2% output, that’s why 10511 and 10507 only cover up to 31°

Duv comparison. (Green-magenta axis)

Tint deviation (deltaxy). How much does the tint change within the beam. This illustrates the 10507 color artifacts the best as it has a big shift right in the bright part of the beam. The further from the hotspot the shift is, the less noticeable it is.

Spectral data and color rendering

For spectral information and CRI calculations I use an X-rite i1Pro spectrophotometer with HCFR, Babelcolor CT&A and ArgyllCMS spotread for the graphs and data. For runtime tests I use spotread with a custom script and an i1Display Pro because it doesn’t require calibration every 30 minutes like the i1Pro.

Explanation of abbreviations

If you have an hour to spare, I recommend watching this presentation on IES TM-30-15 which also shines light into color rendering in general.

CRI data on turbo measured from the hotspot (at 15 seconds before stepdown)

CRI data on 1x7135 (119 lumens)

Output and runtimes

It’s not productive or reliable to compare maximum output numbers between with FET/direct drive lights. Output depends heavily on the current path and the battery used. Even the slightest difference in internal resistance or state of charge will affect the results. Contact resistance also plays a role, so dirty surfaces and the amount of clamping pressure could mean a difference of several percents. Don’t assume your light will have the same maximum output as someone else’s.

At turn on I got maximum output between 2200 and 2800 lumens with a number of 30Qs, VTC5As and GAs. A lower current battery such as a Sanyo NCR18650GA is a good choice if you are planning to swap in Nichias for example. More measurements are needed to gauge whether it’s safe to swap in for example sw45k 219Bs without doing any firmware modifications.

When I first measured the FW3A with a 30Q, I got a maximum output of 2800 lumens, but could replicate that only with a VTC5A after all the runtime tests. That’s why it says 2800 lumens on the table, but 2500 lumens on the graphs. It doesn’t seem that the emitters suffered any damage though since the 1x7135 level stayed exactly the same at 120 lumens.

Muggle mode (six clicks to turn on/off) has an output range of 5-240 lumens with no turbo or moonlight. This is great if you want to hand the light to someone not familiar with high power flashlights or just to make sure it will not step down or heat up much. In muggle mode the user configured temperature limit is bypassed and in my testing the light stepped down after 38-44 minutes before the light even got very warm. This is very useful, since you can just activate muggle mode when handing the light over even if you have set the temperature limit to a very high level.

Turbo is only for very short bursts as output starts declining immediately after turn on. In theory the FW3A is regulated all the way up to the ~880 lumen level where the eight 7135 regulators are active. With a full battery, the driver heats up fast as the regulators burn off the excess voltage. As can be seen from the runtime graphs, the output starts slowly climbing up as the battery voltage approaches LED forward voltage and less power is wasted as heat. This is more pronounced if the light gets some external cooling. Without any cooling or heatsinking (air, hand) the maximum sustained output is below 300 lumens if you want to keep the surface temperature below 50°C.

Otherwise oscillations are minimal. ToyKeeper did a good job on this one!

Zoomed in

Zoomed in

Comparison on ~100 lumen level, 1x7135 mode on the FW3A, S54S, ROT66 and D4. Test done with a single 30Q. I retested the worn 30Q used here, and the actual efficacy of the FW3A is 93 lm/W, but used the same assumed capacity keep the result consistent with the others.

Efficacy is on par with other lights with linear current regulators. Sustained output isn’t as high as buck/boost lights, which heat up less.

Standby drain

There’s some parasitic drain on the batteries when the light is switched off. This is understandable due to the electronic soft switch. The drain negligible.

Standby current: 28 µA

Low voltage protection

Low voltage protection at 2.89V.

The light will drop output gradually as the battery voltage drops until the voltage drops below 2.90V and it turns off.


There is no visible flicker, but PWM is used on modes other than 1x7135, 8x7135 and full FET (turbo).

I measure the flicker optically from the LED with a Thorlabs DET36A/M photodetector and an oscilloscope. Parametrek’s helpful script calculates a snob index, which indicates whether PWM allergic people will notice the flicker. Any light with a snob index value of over 10% may be annoying, but YMMV.

Flicker frequency: 15.6 kHz below 8x7135
Snob index: 0% (not visible)

Mode 1 (3 lumens)

Mode 6 (between 1x7135 and 8x7135)

Mode 7 (8x7135)

Brightness is varied using PWM on every brightness setting except for three modes where either one or all eight 7135 current regulators or the FET are fully active. In smooth ramp mode the default mode after inserting the battery is the 1x7135 level. With default settings the 8x7135 are used as the highest mode (ceiling) in both smooth and stepped ramp operation.


My testing was done indoors in 22°C ambient temperature.

The temperature sensor was calibrated before testing to match room temperature. Maximum temperature was set to 30+25°C=55°C via the configuration menu.

On the 1x7135 level the light will be hand holdable indefinitely. On higher modes, after the initial heating up the surface of the light stabilizes at about 46-48°C, which is hand holdable. Maximum sustainable level without stepdown or temperature exceeding 50°C is about 280 lumens.

Muggle mode has a conservative temperature limit and can’t stay at the initial 240 lumen level forever. In my testing it stepped down to 5 lumens after 38-44 minutes while not being hot to touch.

The FW3A on turbo with a 30Q heats up to about 50°C when the thermal limit is set to 55°C. It stabilizes at around 46-48°C. This is a good setting for not burning your hand. Also loving how quickly the heat transfers to the battery tube from the head.


The basic operation principle of the FW3A is nothing out of the ordinary. Most are familiar with the community favorite user interface, this time Andúril is just implemented in a rear switch light. The design is attractive with high quality machining, which is not surprising considering Lumintop’s track record on previous projects. Considering the extra tube inside the body, the size is marvelously small. I’m fond of the great switch with a tactile feel and the fact that it’s, well, easy to press even with gloves. Of course there’s no trouble finding it.
If you enjoy the beam profile of a triple/quad, the FW3A doesn’t disappoint. The frosted optics is very floody, but can easily be changed to something like the 10507 for more throw and distinctive beam. That’s just a matter of preference. I can’t say much about the final emitter choices, since my review sample had whatever LEDs the manufacturer had in hand at the moment. But it’s safe to say, that the FW3A will probably be modded to suit the user’s need. There’s no glue to hinder swapping in whatever you like.

+ Excellent form facor

  • Nice tactile switch
  • One of the smallest 18650 triples there is
  • UI with lots of configurability
  • Muggle mode is very easy and safe
  • I don’t know, give me a buck/boost driver and a deep carry clip?

Awesome. Good Idea starting a new review thread.

Edit. What do you feel is the sustained output for this host size?
Looks like 500 lumens could be possible. On par with my D3.

Excellent review as usual, maukka.

Thank you so much for the review and the measurements, I appreciate the time you took to bring us the excellent details.

Since a Noctigon 4x MCPCB is about 20-21mm - would a quad Carclo optic fit in this light? They are roughly 4mm larger in diameter than the triple.

Wow. A lot to take in if you’re looking for specific specs, but thanks for cutting through and sharing the most important details in a way that I can understand.

I greatly appreciate the “Temperature” section. It gives me the details I was hoping to get before buying this one.

I’m particularly excited about the small size and the UI. It’s going to be a great hit.

Thank you for putting this into one post. Great info for us while we wait for the final notification! :+1:

Thank you for all your efforts. Great review! Very detailed.

Thanks for the review! Don’t want to sound like a know-it-all, but I think you meant to say “non-ferromagnetic” instead of “non-metallic” for the springs? You put a lot of effort into this, very nice to see such a detailed review!

Thanks for all the work again maukka! Great job.

Think from off 4 clicks is lockout. Did you mean 3 clicks

Thank you maukka, very well done!! :+1:

Very impressive you were able to perform your usual, very thorough review so quickly. Thanks!

It sounds like muggle mode thermal control is perhaps overly aggressive. I’m sure it would be fine to just step down to max 1x7135. However, since it’s not intended for the primary user of the light, ~38 minutes of max muggle output should still be ample for most loaners.

I suppose the big drop off might also encourage muggles to be a little more disciplined with their output level and less likely to hand you back a light with an empty battery.

Excellent review!!!

Amazing work as always. I especially like the charts, very useful!

The beam’s a bit flooder than I expected tbh so will be interesting to see how it looks with a clear or polished optic.

Is there much room in the head for an MCPCB and additional cables? Would love to see aux boards for this light! Provided a kind soul makes them of course :slight_smile:

This was meant to be a production sample, but it looks like some of the internals don’t match their final form. Specifically, I’ve noticed a few things out of spec on mine… things which I hear will be corrected for actual production:

  • Emitter CCT (yours is ~6000K, mine is ~4000K, actual target is ~4875K)
  • Optic (mine looks like it’s 10508 or 10509, target is 10511)

If the beam has zero visible artifacts with XP-L HI, it’s probably not 10511. To check, turn the light on and point it at a light-colored wall from about 20 cm away. Rotate the light around its beam axis. There should be visibly brighter spots around the edges like the hour markers on a clock, which rotate around the hotspot. If there aren’t any “hour markers”, it’s not a 10511 optic.

Mine also had some other minor differences:

  • Firmware version is slightly older than the most recent one I sent, so double-click-from-off still turns off before going to the ceiling level. That may or may not be updated on production drivers.
  • Someone played with mine before sending it, so the ramp ceiling wasn’t default. It may be worth manually resetting yours to make sure it’s actually at the highest regulated level (21 clicks during the ceiling option).
  • Thermal calibration on mine was way off, so it thought room temperature was over 40 C.

So I’m planning to get a new optic before I post any throw data. The 5A/5D-ish tint on mine is not what I consider ideal, but it’s a lot nicer than the SST-20 4000K 95CRI light I used to have.

Thanks for the review maukka. :beer:

Thanks :+1:

That does not sound very production-ish :zipper_mouth_face:

Differences in off-the-shelf parts, as long as the expected parts will be implemented in actual production, don’t concern me too much.

If this is a production test of the custom parts: machining and final driver, tail PCB, MCPCB, etc, that still seems significant enough to reasonably call a production test.

The firmware difference is a bit of a puzzle though. I hope they didn’t have 1800+ drivers built and flashed with an older version of the firmware.

There isn’t much difference between the version it came with and the latest version. The only user-visible differences are:

  • Double-click from off no longer turns off before going to the ceiling level.
  • I added a potential workaround for moon flicker issues, but those have only been reported on Fireflies lights, not on any other brands, so it probably doesn’t matter for the FW3A.