APEX 5T6 NW with upgraded Heatsink and 4x KD7135x8 drivers ~50W

Hi everyone!

So this is my first real post here and since my flashlight buying and modding craze is primarily the fault of this fantastic forum I thought I’d stop lurking and sucking inspiration from this place and hopefully return the favour :slight_smile:

So this mod of my Apex 5T6 NW is my most involved to date, before this I did a few driver/emitter swaps on various hosts but this one needed a bit more planning. Not to mention it took a bit more conviction for me to tear apart and potentially destroy a rather fantastic flashlight.

Aims where:
-Boost output per emitter from around 1.7A (the Apex version of this light is claimed to run 8.5A total shared between the 5 emitters) to at least 3A each (15A total).
-Remove annoying PWM for medium and low modes (supposed to be ~120hz). This alone makes this mod worthwhile.
-Beef up heatsinking beneath the emitter MCPCB by filling in the driver cavity and improving heat transfer away from the LEDs.

So it’s nothing too spectacular and I’m certainly not the first to do this kind of mod on here but I’m still rather pleased with how it turned out and maybe it helps someone like me along who wants to turn this great light into a more spectacular and more usable one :slight_smile:

The Driver:

For the driver upgrade I knew I was going to go for the 8x7135 3A drivers from Kaidomain, specifically the 50ma low mode versions. I had used these before in other builds and since I’m not yet able to reprogram Atmel based chips these seemed to offer the best mode options off the shelf. I only really want 3 modes with sensible low and mediums but with these drivers you can just as easily have access to strobe modes in a semi hidden way. Perfect.

The only question was how many of these could I fit inside that driver cavity and hit both my output target as well as maximize the amount of extra heatsink mass I could fit into the head. I really didn’t want to leave any empty space inside that head if at all possible.
I also had to figure out how to wire up the drivers in a master/slave configuration but thanks again to this forum that wasn’t a problem for very long :slight_smile:

I find it useful in planning to be able to visualize things in 3d so I built a rough model of the Apex head in Maya and tried to get a handle on what was possible. It also let me work out a bunch of stuff before even having to tear apart the flashlight.
That looked something like this when I was done with it.

(The green pucks are KD7135 drivers to scale with the height given by adding an extra 7135 on top of each of 4 chips. so the total height of this driver assembly is 7135+pcb+7135+7135 which I worked out to between 6.8 and 7mm)

From this I saw that the most compact approach for the driver would be to squeeze 4 pcbs side by side wired in parallel for a total of 12A and an extra 7135 stacked on as many of the up facing chips as I needed to reach my goal of at least 15A…or better yet 16A…hehe can never have too much power :slight_smile:

I could then also run any thicker connecting wires between the top facing stacked chips and not exceed this total height of 7mm.

The alternative of using either 3 or 5 drivers as a base would have meant a total driver height of more than this ideal 7mm. Using 3 pcbs would have been cheaper but would have meant triple stacking some 7135s and as I said I didn’t want to sacrifice any vertical space inside.
Using 5 drivers was really only a backup plan in case I totally sucked at chip stacking :slight_smile:

with a 16A Goal and 4 driver pcbs

16/4 = 4A each — 3A from stock driver so needs 1A additional current for each driver
1A/0.380 = 2.6 additional 7135s per driver // 10.4 ~11 total extra
1A/0.350 = 2.85 additional 7135s per driver // 11.4 ~12 total extra

I only had 0.350A 7135 chips anyway so that was an easy choice. In total I ended up soldering 18 chips, 12 onto 3 8xKD drivers and another 6 in 3 stacks of two onto a 4x KD board. (fun day :P)
This whole mess worked out as 28x0.380A and 18x0.350A for a total of 16.94A maximum current. Strange… wonder where that extra amp came from?…hah :wink:

Here is a view of the top of the assembled driver array. The white wires are LED - and the red wires are connecting 7135 right pins to parallel the slave boards. Most of these are on the bottom but I ran out of space for the last one and the short one fixes a damaged pcb path on the 4x board by connecting that bank directly. The Copper strip connects to the + pad on the master controller (right pcb) and provides power for the array from the middle + vias on the Apex driver board.

And a look at the bottom. Here you can see the microcontroller on the masterboard and the red wires connecting the slave chips in parallel. The ground rings of all the drivers are soldered together and connected to the - on the Apex contact board.

Speaking of which here is the Apex driver board with most of the components removed and showing where I pulled battery + and - from. The + section in the middle of this board is directly connected via a number of vias to the battery contact rings on the other side and I didn’t feel I needed to beef up the connection here at all. For the -/ground I had to find a way of getting a solid point to connect to without increasing the thickness of the board much. The ground ring sits flush on a shelf in the head so if the board gets even a little bit too thick there it becomes very hard to screw the body onto the head and won’t bottom out at the right spot.

To solve this I decided to solder on a thin copper strip that connected to and covered the main outer ring and also connected to a number of the ground pin pads that are in the area marked with the black border. I could then file down the part of the copper strip that was covering the outer ground ring and minimize the thickness I added to the board.

This shows the contact board and driver puck, with the copper strip extension of the ground ring before I filed it down as flat as possible, not too happy with that method but it works ok. Maybe a better way would be to remove the dielectric coating on the whole area and solder to the pad directly.

And finally here’s the fully assembled driver ready to be installed. Getting everything lined up and soldered in place was quite fun as there is very little room for error and the copper strip connecting the + to the master board was very fiddly to align. The -/ground rings of the drivers were soldered to the copper strip ground pad using a short piece of 18awg black wire. Also of note here is that the + coming off the Apex board is directly wired to the LEDs + on the led board. Thankfully the + pad on the Apex board is slightly offset from center so that aided in letting me align the + 18awg wire straight up into the hole drilled through the heatsink block…but more in that in a bit.
Suffice to say this was all really cramped and took a fair bit of fiddling and rearranging to align everything. I’m pleased to say I didn’t have to resort to decreasing the calculated size of the heatsink block and most of the 7135 chips are flush up against the alu block when the driver board is fully seated against the groove it’s supposed to sit in. So the body screws on just as it used to.

Heatsink Mass:

Now for that heatsink improvement that constituted such a tight squeeze for the driver.

It’s really nothing special, just a large chunk of aluminium bar stock 40mm diameter, lapped to make good contact with the led mcpcb and screwed in to suck as much heat off it as possible. This lack of contact area below the leds is definitely the weakest part of this flashlight design so anything you can do here will be an improvement. Of course a solid chunk of copper would be even better but I didn’t have anything like that on hand.

I worked out that with the above driver height of 6.8mm I needed a piece of aluminum of around 15mm height to get a very good fit. Luckily I had one exactly that size in a box of bits and didn’t need to track down someone who could machine it for me. After lapping and cleaning up it was precisely the right size to give me that tight fit on the driver cavity. I forgot to take an exact measurement of the block before final assembly but if you’re keen to try this mod 40x15mm and the driver configuration I gave above should make for a good fit.

To attach the block to the mcpcb I drilled and tapped an M3 hole using the middle hole of the mcpcb as a guide. Then drilled two further holes at I believe 4mm for the two LED wires.
To assemble I used an aluminium foil wrap to get a tight fit in the body and applied some thermal compound to aid heat transfer between it, the pcb and the body of the head. To screw it down I used an M3 hex head bolt and made sure everything was nice and tight.

It works well but here you could make some improvements for sure. As I said a big copper block would be better and machined for an interference fit into the head of the light would be ideal. But how much difference this will actually make to output I have no idea. As it is it definitely helps control the heat from the mcpcb for a little while and as an added bonus balances the light out nicely with a good increase in heft to the head.

For final assembly I glued the driver board into the head using Fujik and applied some thermal compound to the tops of the 7135 chips where they make contact with the aluminium block.

Testing so far has been really fun, on high this thing is crazy, crazy bright and combined with some really nice tint NW emitters it’s an absolute joy to use outside at night. I’m not going to attempt to speculate on total Lumen output or lux and don’t have the equipment to make accurate measurements so I’ll just say it’s a heck of a lot brighter than before and the low and medium modes are considerably more usable due to not having that annoying flicker. I did do some indoor comparison shots but unfortunately the difference is not all that representative in those photos, should have done a test outside. Maybe if I get a second one of these lights at some point I can do a better direct comparison.

The tailcap readings taken across the switch with a beefy Watt meter that I trust more than my multimeter are as follows.

Test Setup:

Note: This is with a somewhat tired, 1 year old set of Senybor 2400mah 18650s,

High: 14.5-14.9A
Dropping slowly but steadily from there on down, not surprising considering the heavy load. I believe average stable current after a few minutes was around the 12-13Amps. Not bad
Med: 5.3-5.7A
Good light level for this light, close to being as bright as High mode was before the mod and can run indefinitely without cooking itself to death.
Low: 0.225A
Quite low but very useful, I have it set to come on in low without memory so this avoids blinding anyone nearby when you go outside and lets your eyes adjust.

I have some kinoko IMR 18650s coming from Illumination supply which should provide a bit more juice for more testing. Even just one of these cells can power this light to 8.5A so I bet 4 will do great.

In the meantime out of curiosity I wanted to see if all my 7135 chips were alive and well so I hooked up a beefy rc lipo battery that won’t sag like the 18650s and got a solid steady current around where I was expecting ~16.8A. This test also made me keen to build a version of this light using an external Lipo power pack for extended high power :slight_smile:

So that’s about it, sorry if it was a bit long but hopefully it’s useful to someone attempting a similar mod.
In any case I can wholeheartedly recommend doing something like this on the Apex 5t6. It’s a lovely light and deserves some extra power!

Cheers
Linus

nice!!

What a fantastic crazy mod. That electrical work is incredible and well done on getting it all to work. It must make enough heat to warm an oven.

Thanks, certainly warms up quick now! :slight_smile: I may do a thermal runtime test to see when or if it stabilizes but it certainly hits 55+ deg C fairly fast even with these saggy batteries. Outside I’d be confident to run it on high for the life of the batteries. Nothing has melted yet…:stuck_out_tongue:

Anyway you do some incredible work! I’m especially intrigued by your recent mt-g2 host and testing of the 5/9A ios drivers. Was planning on using one of the 9A versions to do a triple MT-G2 BTU shocker next but not sure now after your overheating experiences. Do you think enough heatsinking would keep it running more stably or is it just outspecced and too inefficient?

Cheers

This may save you some grief. This is a PM from 18sixfifty and I’m sure he wont mind me copying it here.

I tried three different drivers all together. The DRY and two from IO the 7 led one and the “up to three” one. All three worked fine for a while and then froze up and locked in place. You could turn them off for a while and they would work the IO ones anyway the DRY driver just froze and stayed that way. The two IO ones both ended up crispy critters and totally non functional. None of them lasted long enough to be worth trying again.

I too was going to build a triple MTG-2 with some help but have put it on hold until something comes up that will drive it. If you find anything let us know. Cheers.

Oh well damn, that doesn’t sound promising. I have one on the way anyways so I’ll approach it with caution now that I know they fail and maybe I can figure out which component gives out and why.
Thanks for the heads up!

Great build! That driver/spiderweb thingy is awesome! :slight_smile: You are an inspiration to us all (or me, at the very least)! :wink:

I have been thinking about doing something similar to my SRK….man, now that you’ve planted the idea it is going to be hard to get it out. Been reading Dr. Jones’ site about some of the 7135 driver possibilities and it seems doable, even with the electronic switch.

Wow! What an amazing job, Linus!

Cheers guys.
Hah, I’m beginning to think there’s not much you can’t do with those little 7135 fellas. :slight_smile: The fact that they are just as happy driving mt-g2s as in your build has really got me rethinking my BTU triple MT-g2 build now!
Especially since the current driver options for this configuration on 3s seem to be going up in smoke! :stuck_out_tongue:

Thinking out loud right now but how does this sound to the experts on here…

BTU Shocker host with 3x MT-G2s on noctigons wired in parallel.
-Some heatsink improvements if possible i.e Thermal Path from the LED mounting plate to that massive reflector… maybe over a copper gasket that fits between the noctigons?

Beefy external belt Lipo 2S power pack
-Thinking about building this from an Otterbox 2000/3000
-Lots of capacity and won’t sag much under high drain. Plus I have loads of lipo cells I can salvage from my RC helis.
-Loads of room inside the body if I don’t have to/want to run it on 18650s. As an alternative it may even be possible to find a bigger 12v LED buck/boost driver that fits into the body of the light.
-I’m a big fan of handles on bulky lights so maybe add one to the BTU and find a way of integrating a thumb clicky switch.

-Drivers based on 7135s with MCU modded for >6v, maybe 3 in parallel stacked to high heaven! :slight_smile: Total output on high around the 15A maybe.
-That should get it close to being a 100W light…right? :open_mouth: …Toasty!
-Maybe run it with 4 modes on a custom programmed mcu to get 5A per emitter on Turbo, 3A each on High and then sensible and usable Mid and Low
-I suspect I will need to figure out very good heatsinking for the drivers to stop it all melting, no idea how hot these would get running at over 6v @ 15A and dissipating the voltage excess of fully charged 2S lipos (8.4v). Need to study those great Mt-g2 emitter tests on here again! Has anyone done tests on the 7135 drivers to see if they get much hotter/less efficient/quirkier running at 6v vs 3v?

It’s so much fun coming up with crazy, silly torches like this! hehe
Linus

RMM wrote:

Great build! That driver/spiderweb thingy is awesome! You are an inspiration to us all (or me, at the very least)!

+1. Well done.

Linus, your ideas sound feasible. The packaging would be tricky, but you’ve already shown that you are more than capable of pulling it off.

In regards to heat; my understanding is that the 7135 chips basically burn off excess voltage in order to reach the constant current. Given V=I*R, Watts=I*V, it is the difference between the input voltage and LED vf needed to meet the given 7135 current output that is burned off as heat. So…the less the batteries experience voltage sag the more heat will be put into the chips. More heat will be put into the chips at higher mode levels (higher duty cycle) than at lower modes, because when the 7135 is switched off it isn’t producing heat. It is possible that the 7135s are running hotter at 6v than at 3v but I’m not sure by how much, would need to get some measurements to know for sure. Despite all of this speculation, however, we know that the 7135s can survive driving a 6V MT-G2 so it should be possible to do what you want to do.

In my light the 105c is pressed into the pill and has a decent thermal path, at least from the pcb to the pill, so it is getting some cooling, even though the 7135s do not have a direct thermal path. Like you suggested, you may have to figure out some way to transfer some heat from the drivers into the reflector or body. 5 amps/board is definitely possible.

Glad to hear I’m not complete barking up the wrong tree here, will be giving this a go as soon as the parts arrive.
What’s the general consensus on here about flashlights that use external power packs for big lumen output, is it considered cheating? :stuck_out_tongue:

On the 7135s that’s a very clear way of explaining how they work and confirms what I’ve been seeing in terms of heat dissipation and efficiency. Will be interesting to see how they handle what i’ve got planned. I suspect I’ll have to stack them vertically inside a heatsink tube or brass pill that can suck the heat off directly. Actually I’m excited I can do something else with these drivers, still have 40 of the little 0.380A fellas in a drawer and they need a home. :slight_smile:

Today I was playing around with making a flexi-cord from various power cords I had lying around. Trying out various thicknesses and materials. Ultimately I want to have a spiral wound flexi-cord to connect the flashlight to the lipo pack and at 15Amps it needs to be pretty substantial to not suck up all my precious volts before they even get to the regulators!

Made them all “spirally” by winding tightly around a dowel and heating them as much as I dared with a heat gun then let cool. Works ok and they stay wound, but the materials found here are not quite as keen to return to their fully wound state as I’d like.

I found that even for a relatively short reach of around 45cm (in it’s semi wound state, quite a lot longer unwound) the resistance losses really add up quickly. My best result of the day was with a shaver power cord that has what looks like 18awg copper core wires inside. Best result, as in it didn’t weigh half a tonne because of excess plastic coating (white cable) or heat really quickly and drop almost 1.7v under the test load (thinner black cable). That said it still dropped just over 0.8v on my test rig at 15A which I hope will come down once everything is soldered tight and the excess cable trimmed. It’s certainly something I’ll need to keep an eye on as it will have a fairly big effect on the whole power setup and I don’t think I can go much thicker on the cord without making it too heavy and cumbersome. Obviously there’s no point in having a stout battery pack that doesn’t sag when you lose all the benefits getting that power to where it’s needed!


The winner

The test arena

The rejects :slight_smile:

Should probably start a new thread for the BTU planning and tinkering…
Cheers

Well you definitely got me beat.
.

.
.
Nice job LinusHofmann. I feel your pain :wink: Mine’s only running 4 amc7135 2800mah drivers. Should give 11200ma, but only does a little over 10amps in a 4-xm-l kung using Panasonic 2900’s laptop pulls. Really didn’t take as long as I had thought it would to throw these boards together. Thanks for sharing and great mod.

Oh cool, and that’s running the electronic switch similar to the SRK right? I assume you had to flash some custom firmware to get it to work right with these drivers. I should probably read up on how all that electronic switch stuff works, could come in handy for sure!

007, please share the juicy details of your setup. :slight_smile:

Great mod and work!

Sounds like you made an awesome light.

:beer:

There’s a little more detail in this thread.

Great build.Cool Of course, we still need beam shots. Wink

Working on it :slight_smile:

OK so today I received 3 more of the Kinoko IMR 18650s from IlluminationSupply and I thought I’d do a quick runtime test on the Apex to see just how toasty it gets when driven by some stout IMR cells.
I was also keen to see if my driver heatsinking would be adequate over an extended period on full or if I would see some throttling back of the current from the 7135 chips as they get too hot.

Incidentally the Kinoko cells have an internal resistance of between 85-95milliOhm which seems pretty good to me, in comparison my year old Protected Senybors (2400mah) are up around 155-170 each and the Protected Trustfire Flames that I’ve tested are also in that range. Tested on my iCharger306B with relatively long and sloppy leads so take them only as comparative figures.

So anyway here are the results.

–4 freshly charged Kinoko IMRs reading 4.22v each
-Run on high with IR temperature readings taken at the head every minute.
-No fan or air movement of any kind, flashlight lying on it’s side (didn’t want to risk burning a hole through my desk :P)
-Tailcap current readings taken with my trusty Turnigy Wattmeter with 12Awg leads soldered directly to the switch tab solder points (see above for a picture of the test setup)

I called it a day at 5mins because with temps still rising steadily and close to 70C at the head I started to get a little nervous! Didn’t want a runaway meltdown scenario! :wink:
That said the temperature reading of the body was obviously at a lower level, inside the battery compartment taken from the driver PCB was at a safer 45C and the batteries themselves came out of the flashlight at 47.5C.

In terms of the output stability and how these batteries performed I’m really impressed.
The current readings where much more stable than I was expecting and didn’t exhibit anywhere near the sagginess of my Senybor or Trustfire Flames. Seems ~4A per cell for these IMRs is cruising speed :slight_smile:
I also don’t think the steady drop in current as seen has anything to do with 7135 chips overheating and throttling back so that’s reassuring. It all looks to me like solid battery performance with an expected gradual drop.

Conclusion is this light loves IMR cells, and can safely(ish) be run on high for <5min bursts. Again outside at night with cooler ambient it’s a different story and should stabilize at a lower temperature. Plus the handle makes holding the light at high temperatures fine although it’s easy to forget and accidentally touch the head with your index finger for a nasty surprise!

Thoughts welcome but I think I’ve achieved the aim of making both a camping light and heating stove all in one! :slight_smile:
Hopefully I can get some beam shot comparisons done tonight.

Linus