Haikelite HK90 tear down

Yes. I would sand it until it’s the same thickness as the centering gaskets.

It’s really unfortunate they routed the traces that way. No reason for it.

Mod continued

Prep'ing Surfaces

Started at 400 GRIT, notice the machining marks are totally gone now:

The 600, 1000, and 2500 GRIT:

Same done with the copper plug cutout:

copper washer, as JaredM suggested. Overlaps the traces, slightly thicker than the centering pieces:

Original washer thickness:

After flat surface sanding, 230 GRIT, then finer to finish:

It's clearing one trace now, thinking of leaving it overlap the other since the overlap is too big:

Preping the 18 AWG replacement wires, slightly longer than stock wires:

Sanded to 2500 GRIT as well, burrs are all gone, flattened and level:

Mirror finish on the plug from 2500 GRIT, though not all the machining marks are gone, still probably over 95% flat:

Would love to see what numbers you are getting after the mod.

Dang Tom E, you are good! Thank you for this entertaining and informative post.

I’ve always been impressed with members like you that take a perfectly good working expensive flashlight apart for the benefit of the BLF community. And in the end you’ll end up with a better flashlight than what you started with.

Thanks I'm not expecting much of an improvement, but better overall efficiency. Tomorrow morn I'll download Anduril 2, then hopefully can assembly it all back and do some testing.

The Finalization of the mod, and tests

At this point with the bare driver, it's a good point to do the firmware update. So, got online, retrieved TK's latest Anduril2 rev 579, dated 2020-10-19, updated my Win 10 based Atmel Studio 7 Anduril2 project with the updated files, and set up a new config file for the HK90. Used the Q8 hardware config file (same FET+1 w/switch AUX LED's), but used the ramping table from the Mateminco MF01S config file (high lumens)

These type of lights with a reflector retaining screw inside the driver cavity, you have to make the wires long enough at a minimum to be able to move the driver out of the way to get a screwdriver on the screw to loosen up the reflector. You can't start with de-soldering the wires on MCPCB because the reflector is locked down. So here's the new 18 AWG wires threaded up

Now the wires soldered with plenty of flux:

Then cleaned up (psst: notice anything wrong yet?):

Applied kapton tape to protect against possible ground outs on the reflector, and the copper washer placed with MX-4 coating on both sides:

Next is adding the copper plug/heatsink, MX-4 on contact surfaces:

Heatsink placed, reflector secured down with the screw (copper screw is shipped but coming from China), and screw also treated with MX-4:

The resistor change made the difference I was hoping for - brighter blue switch LED's:

Ok, doing the first power-on test, heard a slight crackle when the LED's blink in tightening up the battery carrier and first contact is made, hhmm. Then clicked ON the light -- Ooops! One LED is out, dead. Ouch! opened her up, removed the reflector, and noticed the problem:

  • Stupid me, the bad LED has the Red/Black wires reversed! I know I checked this and somehow totally missed it! The pics above captured the mistake, and still didn't notice it! :FACEPALM:

Ok - hoping for no permanent damage, re-wired, re-assembled - LED is working and doesn't appear to have any side effects. So, test results:

  • lumens start-30secs: 18500 - 15880 (Tom E @'s), 16320 - 14010 (maukka cal'ed)
  • 737 kcd

So it's a bump but not much. Kind of expected this but still there's something holding it back. Then realized I didn't look thoroughly at the springs:

What is this? Ohh, looks like a quality outer spring, and double inners of very thin, very soft copper or copper clad, soldered on the top, but not the bottom -- they have to go... So:

Removed the dual inner springs, adding 20 AWG bypasses:

When re-installing the spring PCB, added NO-OX-ID on all contact surfaces of the brass screws - nice brass screws came on the stock light.

Now we about done with the mods I wanted to do, minus the copper reflector screw replacement.

Doing just a quick lumens and throw measure, I see about a 5.3% bump on the 40T's (solder tops), lumens (Tom E/maukka): 19480/17190 at start, throw: 785 kcd

Now for the ultimate test - solder topped 30T cells, fully charged:

  • lumens start-30secs: 20500 - 17580 (Tom E @'s), 18090 - 15510 (maukka cal'ed)
  • 820 kcd (1811 meters)

So I think what happened here is the FET and 18 AWG wire change didn't do much because the springs were still holding down the amps, so once they were bypassed, got a decent bump with all the mods combined. Now I'm not seeing/measuring benefits of the thermal improvements. At 30 seconds on turbo, the housing in the hot area gets pretty warm but still ok to hold in the hand.

In the Anduril config file, I add these settings:

#undef DEFAULT_THERM_CEIL
#define DEFAULT_THERM_CEIL 60 // from 45 -> 60
#define MAX_THERM_CEIL 90 // from 70 -> 90

#define USE_2C_MAX_TURBO

// Make standard mode default, not Simple mode
#define SIMPLE_UI_ACTIVE 0

So I'm bumping the thermal ceiling, make getting to turbo easier, and defaulting to the regular Anduril UI (not simple or muggle). So I cal'ed temp, and set the ceiling to 85C. Also had to tweak the voltage reading a little, using the config UI (new feature).

Overall pleased, and really really glad I didn't mess up an expensive LED

Wow great work! Thank you very much! :+1:

Can you please share the Anduril2.hex file that you made ?

Sure, plus the source code - may have to wait til this evening.

So for those wanting to get the most bang for your buck here it’s looking like the spring bypass and UI temp configuration tuning are the way to go.

Great job TomE. Very curious how hot it gets after 60sec of turbo. :smiling_imp:

Thanks for all the info TomE. I wonder if upgrading the spring would be the 'quickest' upgrade.

As JaredM says, I'd be curious on the runtime graph of at least the first 3 minutes or so :)

Springs are the easiest, and maybe best bang for the buck, but once the springs are done, then maybe the LED wires become a bottleneck. Not sure til it's tried. Would be interesting to see how many amps are being drawn, though that's probably not easy to do. I'm guessing in the 70-75 amp range, based on the measurements.

I kept the driver glue free, and at first it was difficult to load the batts and screw down the battery tube because with those thick and long 18 AWG wires, they wanted to push the driver up. Now though, the driver is staying down in place - maybe just needed time to be in the assembled state to get in shape, and maybe getting some sticking on the driver edges.

Also fyi, I did not add back in the copper heatsink aluminum cover - there's no point since removing all that thermal grease piled up inside of it.

I'll try to take a shot at runtime graphs, might not be for a couple days though.

Hey great job and some meaningful performance gains! I get nervous whenever I I ter a jumble of wires like that in tight spaces! Shows there is potential in this light after all. Maybe add more copper in that open part of the copper slug? You could run a countersink flathead screw through for clamping with some thermal paste. That would help hold turbo or higher settings longer I think.

Thanks! That big M5 center screw goes thru the whole stack: copper slug-shelf-MCPCB, then screws into the reflector. That's why I'd like to use the copper one I ordered and not the steel stock one. Yep, more could be done with the copper slug, but that space is pretty crowded with the wires.

Originally was gonna de-lens the LED's, but the centering pieces sit on the PCB and appear to use the lens to center on, so I don't think it would work out.

Thanks Tom E :+1:
You guys are a wealth of info and experience for us to soak up ! :sunglasses:

This weekend I want to modify the HK90. It would be great if can run Anduril 2 on it. :sunglasses:

Posted in the Anduril2 thread, will duplicate it here:


Here's the Anduril2 build source code for Atmel Studio 7.0, and the Anduril2.HEX file for the Haikelite HK90 I used:

Google drive share for HK90

Anduril2 Updates

Source Code Base

Summary - List of Changes

  • TE 2020-10-31: Make the version # v1.2 (in version.h)
  • TE 2020-09-25: add optimize setting: “-fwhole-program” to reduce code space

CFG file Changes

  • Thermal ceiling default changed from 45C to 60C (DEFAULT_THERM_CEIL)
  • Thermal calib extended from 70C max to 90C max (MAX_THERM_CEIL)
  • Add USE_2C_MAX_TURBO
  • #define SIMPLE_UI_ACTIVE 0 (default to standard Anduril w/full features

Operational Changes

None at this time – TK incorporated a couple of my prior custom tweaks

Build Notes

  • 2020-10-31: HL-HK90 built at 7908 bytes

The copper M5 screws arrived this week and they are look'n good!

AAA light:

Here one is installed. This is how it should have shipped

Latest Beamshot:

Ok, not my HK90 but something close to it!

Hey Tom,
What distance did you measure lux? Rick sent me his to look at and at any real distance the beam looks out of focus, out past 150 yds or so the center of the hot spot shows shadow. At 5M distance I’m getting 732.5Kcd, pretty sure that would greatly diminish on a 50M test.

Nice work optimizing yours!