Convoy 3x21A SFT-40 6500K first impressions, info, mods

I just received the Convoy 3x21A SFT-40 6500K today, paying full price (no free review sample or discount, etc.). I haven't bought a Convoy light in quite a while but I do own, and sold quite a few earlier Convoy's. Many of the new models are re-labeled older models, but with different LED's and drivers, so I wasn't too impressed with many, since I've modded most Convoys with drivers and LED swaps to my liking anyway.

Some first impressions:

  • overall excellent quality - clean outside and inside the reflector, fit and finish is there
  • threading is Convoy quality (excellent)
  • nice SS bezel, AR treated lens, nice threads, nice anodizing, good grip, relatively heavy
  • interesting there's nothing included, except for the flat top adapter, but no lanyard, no spare o-ring, no manual
  • The SS bezel has softened edges, except where the notches are for fitting a tool - very sharp edges at all 4 of these machined out notches
  • The dimensions and weight as listed, match up well to what I measured:

69 mm head diam, 158 mm length, 570g w/no batts, 777g w/batts

  • If I don't compare the beam to other lights, I would say the tint is pretty clean, and a decent white. But side by side with ~5000K LED lights, it does have some green/blue coloring that shows, but no tint transitions like you see in the CREE xxx.2's.
  • This light has deep reflectors, reducing the overall flood area width but brightening it up as well. You can see that in some of the beam shots below. They did a good job in maximizing the reflector widths in bringing the outer edges of the reflectors right to the bezel, but they could have been made even bigger by increasing the overlaping which reduces the walls between LED reflectors. Other multi LED reflectors are designed that way and it seem to work well.

I haven't seen much in measurements or specific details on the driver on the new Convoy's, so I thought I would delve into that here, and add to this OP as I learn more. From the start I found Simon's listing intriguing, spec'ing lumens, throw and amps, but adding verbiage like "theoretical" and "for refence only", so making it as non-committal as possible.


Interesting review of 3X21A w/SFT40 (in Mandarin): is external)

Beamshots start around 11m 30s. Credits to TLF member helicoil who frequently discovers latest news on Baidu.


For the amps measurements, I only used a single cell, a charged 30T solder top without the battery tube, so the spring loss's are not included, but with 3 cells, the amps would probably be higher than 1 cell. This is all on the stock light, out of the box.

  • Amps at tail of the 30T cell:16.8 amps
  • lumens: on charged 40T flat tops w/adapter: 5515 at start, 5220 at 30 secs (mauuka calibrated #'s)
  • throw: taken at 5 m indoors: 286 kcd (1070 meters)
  • parasitic drain: 32 uA (nice and low!)

Comparison with a couple old lights of similar size, the FandyFire SP03 and SupFire L1:

It's quite large compared to a standard BLF Q8:

Comparing the Q8 XP-L V6 3D, ~5000K, (left) with the 3X21A (right). This is where it gets dicey. Clearly the SFT-40's are showing green in these pics, maybe a little more than in real life but I can't deny the green can be seen:

By themselves, Q8 first:

Now compared to a SST-40 6500K light, BTU PK26 on left:

The PK26 always looked blue white to me in this light, but the 3X21A is still showing a strong green. It really doesn't look that green green as the pics are showing:

The 40T's are about as tight as possible, still the tube is slightly wider than a 4x18650 setup:

The driver - really like what's done here to make it easily accessible, removable. 2 screws in recessed positions, easy as could be. No driver wear (like on a Q8) since the battery tube makes contact to the bare aluminum of the housing. I really like this simple design:

The driver shown below - didn't have to remove any wires to get this picture, just tilted it out. Interesting design. I believe the 2R2 coil is part of the USB charging circuitry. The 4 FET's in parallel is the most I've ever seen in a driver - this is probably the main high amp circuit. There's probably a low amp circuit as well, but I didn't trace it fully yet. What's obvious here is those two low value big resistors (0.02 ohms) sitting between batt- (ground) and the ground input to the FET's - these resistors act as an amp suppressor to limit the max amps. Of course my first thought is to simply jumper the resistors out to boost the amps, because ~17 amps for 3 SFT-40's is limiting. There could be easily 24-27 amps feeding the LED's, and lumens in the range of 6500 instead of 5200, and kcd would get an equal bump in percentage as well. Also the springs don't look like quality springs - bypasses need to be added. The bypasses alone should boost amps based on this driver design, unless there's some current sense resistor circuitry I didn't notice.

The LED wires are thick ones, I'm guessing 16 AWG - this is well chosen.

I'm not a fan of this UI. It's ramping is clearly linear and goes way too fast at low levels and way too slow at high levels, so my preference in modding this light would be to add a piggyback driver and remove the driver components off the board, trying to keep the charging circuit in tact. It looks do-able at this point.

I did test 40T's with solder tops and no adapter and actually got slightly lower output #'s. This could be from a difference in cells, or perhaps with the adapter there's a bit more spring compression that lowers the spring resistance. More evidence bypasses may get a nice bump.

The adapter I got is pretty beat up. I had to trim some plastic slightly covering the brass buttons to ensure good contact. Here's what it looks like, cracks and all, out of the box. Though it looks pretty bad, it's still usable, and I have enough 21700's with solder tops to not use the adapter anyway:

I'm not fond of the dimples in the center of the brass buttons. I'll sand those out. The more surface area of contact, the better.

Both sides have the dimples.

That's it for now. I'll be adding more info, and modding details.

I started on the mods. First added 22 AWG spring bypasses:

Then some minor changes to lessen resistance for the screws for mounting the spring PCB board. Sanded off the anodizing to make contact with the conductive surface of the PCB:

Replaced steel screws with brass ones, and NO-OX-ID treatment:

Couldn't measure any noticeable difference in output. With the stock driver, the batteries were not a bottleneck, but with a direct FET based driver replacement, these mods will all help. Was hoping I could keep the charging in tact but there's too many interconnected singles - I counted 3 between the driver MCU and the charging controller.

So, this is what it ended up as. All surface mounts removed, piggybacked FET+1 driver. There's plenty of clearance for the mounted board. I used a good quality 2 part epoxy to mount, keep it insulated from the lower driver board. Notice the new pads I created on the R and K pads? This is for the RED switch LED and switch to wire from the new driver to the old driver. There's a blue 28 AWG to connect batt+ to the new driver - doesn't need much in power, the batt+ wire to the LED's will be 18 AWG

Here's the switch PCB. Small resistors for RED and GREEN LED's (100/500 ohm), but I'll only use the RED's. Somewhat odd they chose to use 2 RED LED's and only 1 GREN LED. The SMD LED's are dual color.

This is somewhat disappointing. The grease is as thick as silly putty and doesn't spread, even when secured down by screws. I hope the surfaces are flat, and the MCPCB has some contaminants on it. This will get a sanding treatment and MX-4:

That's it for now..

2021-07-27 - continue on, and completed the full mod.

Cleaned up MCPCB:

Nice size:

Excellent thickness:

Shelf cleaned up before sanding:

After sanding:

I enlarged the wire hole, using a metal grinding tip on my rotary tool, here shown from below. It made a big difference. I can now fit either 16 AWG or 2 pairs of 18 AWG without a sharp bend, like it had stock:

Now from above:

Also I enlarged the pads on the PCMCB. Looks like the reflector had clearance for it:

Here's the driver with all wires except the main LED pairs. I tested it out in this form by adding temp LED wires. Notice I tinned the pads and added a 2nd pad for the LED+. Blue wires are 28 AWG, black batt- wires are 20 AWG:

Decided to go with 2 pairs of 18 AWG. The 16 AWG is a lot stiffer and was concerned about the thin small pad for the LED- wire on the small driver. The 2 18 AWG wires should be more flexible, less stress. Had to dremel the wire opening a little on the MCPCB:

Time for the final assembly. Using MX-4, I wanted to concentrate it in the middle and verify it spreads out when assembled:

Wires inserted, carefully pushing down the MCPCB into place:

Last view of the driver with all the wires. Made the pair of 18 AWG little longer than stock so I had room to work with. The 2 LED- wires were a tight squeeze:

View from above, screws tightened down. I can see some of the MX-4 oozing out - good sign:

Assembled view from the backside:

USB-C connector is now gone, giving you a view of the piggybacked driver edge:

Kapton tape applied just-in-case, and stock centering pieces. Was thinking of sanding them down but didn't yet. It can help or make it worse:

The reflector is notched out to allow space for the wires:

Wall shots of the "new" 3X21A. As you can see, the beam color is all over the place. My camera only does auto white balance, high output:

Low output:

So pick your color - white, blue, green!

Fiddling with some camera settings:

It's a nice looking flashlight, even better look'n now with the switch LED controlled by Anduril2

New Measurements

I wasn't 100% careful in taking these #'s, didn't have the light meter at the right scale initially, and when I changed it, I recorded that # as the start number. Also my timing might have been off by a couple seconds.

  • lumens: on charged 40T solder tops: 7330 at start, 6180 at 10 secs, 6000 at 30 secs (mauuka calibrated #'s)
  • throw: taken at 5 m indoors: 335 kcd (1158 meters)

It's certainly a nice bump. Looks like a fast drop over the first 5 secs or so, but very stable slow drop from 10 secs to 30 secs. It's a well designed host, plenty of size for these hot new LED's.

2021-07-28 - another set of measurements, more complete, thorough

  • amps measured on a single 30T at 4.20V: 24 amps
  • lumens: on charged 30T flat tops w/adapter: 7151 at start, 6696 at 10 secs, 6636 at 15 secs, 6424 at 30 secs (mauuka calibrated #'s)
  • throw: taken at 5 m indoors: 360 kcd (1200 meters)
  • parasitic drain w/switch LED OFF/LOW/HI on a single 30T at 4.20V: 35-90-361 uA

These are significant numbers. With my prior used calibration factor based on manufacturer's ratings, the lumens would be:

  • 8024 at start, 7208 at 30 secs

With the mods, the SFT-40 matches or out throws, and exceeds lumens over the 4X18A SBT90.2, based on reviewer's measurements.

Congrats on your new toy. It looks like you went specific for throw with SFT-40 6500K. Any specific mods that you’re looking at? I’m a flood type with some throw. It’s Cree XHP50.3 for me, though Simon mentioned that a triple XHP70.2 was in the works.

Yea, for throw, but also lumens. I'm thinking triple SFT-40's can out throw and out lumen a single SBT90.2 for at least $20 cheaper. Maybe the throw is close, or less, not sure. The $20 becomes $40 when they are in a light it seems.

Does the XHP50.3 solve the ugly tint shifting of typical CREE .2's?

Great review, Tom. :THUMBS-UP:

I must admit, I did not do any side by side comparison with other lights when I posted about my 3X21A sample. If I compare it with some XP-L HI V3 1A 6500K lights, the SFT-40’s white still looks pretty nice. :))

By the way, I have no idea how much beaten up my adapter looks now that I have screwed it on with the batteries installed in the tube, lol. It was totally warped already when I put it on top of the batteries. I didn’t dare to take a peek yet. :D

I expect the 50.2 HIs will, but unfortunately they’ll likely only come in very high colour temperatures.

I have 3x21A Sft40 and 4x18A Sbt90. I bypassed springs and sanded down the middles from flattop adapter. Negative end pcb contacts to the flashlight body only with small steel screws. Three in 3x21a, four in in 4x18a.I added three more bigger screws to 3x21a to accomondate that. Still the 4x18a outperforms 3x21a clearly and tint is nicer. I use bundled Liitokala batteries on both.

For those who might be interested I have few comparisons. Be warned, I’m not a photographer.
Left side 4x18A, right side 3x21A

Bottom of the ramp

Top of the ramp

You are right - the 5700K SBT90.2 has a more pleasing beam tint, plus comparing a single SBT90.2 to a triple SFT-40, you don't have all the artifacts associated with 3 reflectors. Those deep reflectors show their effect in that last wall shot.

Did you notice or measure a bump with the spring bypasses?

Jumpering out the resistors should be a bigger bump, but does have risks. I'll probably try it first before piggybacking in a driver replacement. I don't have to worry about blowing out the driver circuit. I don't care for the UI in this light at all. The press&hold from OFF should still allow you to ramp. If last used is max/turbo, you'll get blasted with it if you need anything but the lowest setting. In Narsil and Anduril, we accommodate that by adding a delay at the lowest setting, then ramping up is engaged. The linear ramping drives me nuts -- I thought Simon was asked to fix this over and over, I don't get it.

Truthfully I didn’t notice anything happening after spring bypass. Might be very small increase if anything at all.

The disk quality is disappointing. I have the 4x18 and the disk is super strong.

That’s one area I wish Simon would innovate: UI. I can live with quality eccentricity, but UI’s are easy to deal with, especially since pretty much every manufacturer is using Anduril these days. Great lights let down by wonky UI’s. To be fair though, Convoy has gotten better with USB C charging and great single cell lights with e switches.

The greens in the pics really surprised me. It doesn't look that bad in real life, even on a white wall. If I put it side by side with an old SST-40 6500K, the SST-40 looks blue and this SFT-40 is more on the green side, but not the strong greens in the pics.

Looks like this light is still one i want to (try) to get, with SFT40. I agree my pictures give it an unfair green look as well, which doesnt show as bad in real life. Sad to hear about the UI, i have been reading about it on the Convoy thread too.. I don't know how i would fix something like that, so' i honestly might hold off on this light until Simon can straighten it out or until i have the pleasure of watching your piggyback mod and determining if it's something i can do. I mean sounds wroth it to buy a $23 astrolux ec01 (Anduril) and rob the driver for a piggyback? is that how it works? lol. i obviously have no idea. Thanks for the insight Tom

It's common to see our smooth ramping poorly implemented. Back in 2016/2017 during the Q8 development, the Thorfire engineers went off on their own and decided to clone our UI on a different MCU, rather than use the ATtiny85 and our firmware. The ramping in that prototype was linear as well. Thorfire ended up firing the engineers from what they told us. There's been several other smooth ramping lights with the same problems, and typically they make the ramp speed too long, longer than our 2.4 secs (16 msecs x 150). Our ramping tables are 150 entries and ramping increments/decrements every 16 msecs.

Love those ThorFire’s





Updates in the OP and post #2 for mods.

I was wrong I didn’t mean XHP50.3 but XHP50.2 3 volt. The 6500K have the worst green/yellow. 5700K a bit less green same amount of yellow. When outside not much of a factor compared to white wall shots. It’s a pretty drive such a shame a little resistor mod could do it. SFT-40 at 5.66 amps each is factory nice but not around here. 8-9 seems very possible. It does look identical to the 4x18A driver running the SBT 90.2 including the two big resistors rating. Looks like a one size runs everything.

I guess he amp limited the 90.2 version as well? I think there's something in the circuit to limit amps, not just those big resistors. Not sure what type of current sensing but all those components to the right of the big resistors look like they are doing that, but not sure.

I hope someone gets to compare sst-40 to sft, in this same light, with such good photos as OP. I have the sst40 5k on the way, but i am wondering more about beam shape than tint, just so I can understand when I should or should not choose sft.

Thanks Tom, great pics and details. I figured I could just use a convoy, but I will look it over for any issues when it arrives, thanks to your tips.