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

I'm very mixed on the opinion of the tint, same as the pics show. The real tint to my eye seems to match closest to that last pic, but if you look for the green, you can see it, but it's subtle. I would say it's more green at low output that high output.

The XHP35 HI is great with availability of warmer tints, but it's difficult to work with because of the 12V requirement and high Vf for a 12V LED. You really need a boost driver even if you have a 4S battery setup, as DEL found out with the BLF GT driver. I think the SFT-40 will out throw the XHP35 HI as well and I'm pretty sure the dated XHP35 HI is less efficient when driven hard than the SFT-40. I think the XPH35 HI is based on 4 LED's comparable to the XP-G or XP-G2 - it's based on older designed LED's.

It very much looks like a linear driver just like their 5 and 6A driver, bridging the two 20mΩ sense resistors will result in full direct drive output with no modes. You can decrease the sense resistors in order to achieve a higher output, assuming that the mosfets can handle the additional thermal stress.

Ok, interesting. I wasn't sure if those were sense resistors because they are directly in the Batt- path before the FET's. Yep, that's what happened when I bridged them - modes were all messed up, or rather the ramping was about useless. Thanks thefreeman! Wish I knew more about the various LED driver circuit designs.

This is an example of a linear driver :

The op-amp would be what looks like a SOT-23 package at the top right of the board, the Vsense trace starts under the left R020 resistor, goes to a resistor (R8) then the op-amp.
The PWM signal out of the MCU goes through a RC filter to smooth the signal into a voltage, probably a two stages RC filter actually seeing the number of passives next to the op-amp, and then a voltage divider to set Vsence.

So you have I=Vsense/Rsense and since you know I=17A with Rsense =10mΩ you can calculate what Rsense you need if you want 24A or something else.

Dang, a little too late for me - driver stripped, but good to know, specially if others are interested in tweaking more output. Does this mean there's no noticeable PWM's on all output levels?

No PWM indeed.

Edit :

Ah yes somehow I missed the modding part in the second post, with the components removed we can clearly see it is indeed a linear driver.

Thank you Tom E for your very detailed and comprehensive review. After reading it I realized that 4x21A SFT40 is not able to overcome the 4X18A SBT90.2 so I really believe that I will not buy it instead I am waiting for news for a 3x21A SBT90.2 that I would buy with my eyes closed.

Sorry...4x18A!

I swapped them 3x21A SFT40 and 4X18A SBT90.2, I was writing and thinking about more ...

I don't know if anyone did lumens/throw measurements on the 4X18A SBT90.2, but based on his listings the SBT90.2 is probably out throwing the 3X21A SFT-40 but can't say what the lumens are, but thinking they are roughly the same. His posted specs on lumens are not accurate but thinking the throw #'s are fairly accurate. This makes sense because it's easier to measure throw. He also doesn't state what the amps draw is of the SBT90.2, so can't compare that, but I suspect the SBT90.2 is drawing more amps.

The 3X21A triple reflector could have been designed better for throw by making the cups wider and having more overlap.

But certainly the beam's tint of the SBT90.2 is better.

Ohh - tech reviews:

https://www.reddit.com/r/flashlight/comments/obhlqm/review_convoy_4x18a_only_50_notes_for_a_1122m/

https://www.reddit.com/r/flashlight/comments/o3si19/tactical_grizzlys_convoy_4x18a_sbt902_review_the/

4920 ANSI lumens mentioned, and throw measured at 367 kcd and 337 kcd.

damn.. this is kind of disheartening to hear, I was looking for every reason to buy the 3x21a sft40 over the 4x18a sbt90, but looks like I would go with the 4x18a. Of course, I need to figure out how to piggy back a driver like that. I think if i can figure most of it out, except what to take off the original driver (did you remove everything?)

also if you don't mind me asking, what driver did you use for the piggyback? Or what driver is a good candidate? (Fet w Anduril and flash pad)?

Yes - removed all parts from the stock driver. Driver I used was designed by a BLFer Quadrupel, uploaded to OSHPark, then I reflowed the parts. It's a FET+1, forget if 20 or 22 mm in diam. Actually it's the driver he designed for use in the Amutorch E3/E3S.

For driver, 20-22 mm is a good size. FET+1's are easy to work with. Flash pad for pogo pins don't work out well unless it's exposing the pins on the top (SMD) side. I was gonna use one but on the driver PCB, only the backside pogo pin holes were conductive, on the top side they were covered over.

awesome, thanks for the info! I have something to consider now

Very nice modding job Tom. What cpu is on the board you used? Is it pulled from another light, or something you built? Do you know what the old cpu was? It looks in your pic like they sanded off the markings!

I wonder if you’ve tried TK’s very latest ramp code, from just yesterday or so. See the current thread about it: better low modes with dynamic PWM

It is incredibly clever from my perspective.

The MCU is a ATtiny85. The driver I used is something BLFer Quadrupel designed for the Amutorch ES3, but I think I uploaded the design files to OSHPark and bought them off of there. I reflowed the board, adding the components. It was just a driver I had laying around, but any 20 - 22 mm or so e-switch driver could be used.

TK's latest ramp tweaks I think only apply to particular designs that have issues at low output levels. The driver I used is what we call a "FET+1", meaning it uses a 7135 (0.35 amps) regulator/limiter, and then a FET for high amps. The 7135 "raptor claw" version does really well at low amps so doesn't need those tweaks. The problem with FET+1 drivers is you always get PWM flicker, though we use a high frequency, about 16 khz, and using the FET and 7135 using PWM's to control output is not power efficient or have the ability to maintain a constant level of power as the battery drains. Buck, boost, buck/boost, or CC linear regulated drivers are considered a better more efficient design. Of course they can be more costly, more complicated circuit designs, more complicated to program, require more space because of a higher chip count, and are usually max amp limited compared to FET+1's.

Hi Tom, I like your mod, I don’t know why Simon doesn’t use Anduril in his designs, Anduril is such a perfect User Interface IMHO

Anyway, I am thinking of maybe using an Astrolux EC03 Driver, to have USB-C charging dock and also the Anduril UI.

Do you think the EC03 driver can fit piggyback on the 3X21A board like you did and be able to position it in such a way to still have USB-C charging?

Thanks!
AlexGT

This is the EC03 driver:

I don't know how you can physically get the USB-C connector in the place of the host cutout - maybe flip it over but not sure there's clearance for that. Unless some how you can add a few jumpers to the 3x21A board's USB connector. Re-using the 3x21A connector might be the best option, but dunno how, and how easy/hard that would be.

For me that would be crazy fine cut/jumper modding, but things like that have been done.

I’m wondering about using one or two of these in parallel for mods like this, space dependent. They are single sided. Using stock connector and jumping wires to the charging boards. Limited to 1A each, but better than nothing and having a useless port.

Interesting, maybe possible. I think there's not much vertical spare space. Built in charging is great for an EDC, like the SP31 Pro that can run Anduril (and 2) though.

Convoy 3X21A custom handle - YouTube have a 3X21A SFT40 to which I have added a handle to make it more comfortable to use. Here you have the result. What do you think?
I also bypassed the springs and got +10% lumens at Max