The goal
Compared to the original, it will be both significantly brighter and dimmer, it will have more output levels (7 levels instead of 2), it will allow starting in the highest or lowest modes and easily stepping up or down, it will have much faster PWM (~9 kHz instead of ~150 Hz, I think), and it will show colors more vividly than pretty much any other light available. To me, this multi-tint approach looks even better than a high-CRI neutral white like the Nichia 219.
Parts
The original had XM-L T6 1A emitters on aluminum, and the 3-toroid driver with only two levels (100% and 10%, plus hidden strobe). I’m replacing this with three XM-L2 emitters on copper, with tints chosen for relatively optimal spread across the ANSI white spectrum. Plus the new BLF SRK driver from the Oshpark thread, which RMM now sells pre-assembled (I’m not skilled enough yet to build a driver like this on my own, so it’s nice that I could just buy one). The driver is a 32 x 350mA 7135 model, for 11.2A total or 3.73A per emitter.
All parts came from RMM’s store, http://mtnelectronics.com/ .
Tint mixing
I originally wanted to do 6700K (1A tint), 2700K, and something in the middle. The U2 1A tint is out of stock right now and RMM didn’t have anything as warm as 2700K, but I think I found a different suitable configuration instead — T4 7B3 + T6 4C + U2 1D. This works out to about 4700K and 3058 emitter lumens when driven at 9A. It’d actually run at 11.2A, but the 9A value seems reasonable for OTF lumens after loss from optics. To estimate this value, I used the Cree tint chart and output bin specs and came up with this…
>>> (6300 + (3100*0.7998) + (4350*0.9329)) / 2.7327 = 4697 Kelvin
>>> 895 + 1044 + 1119 = 3058 lumens
Although I’d prefer a 1A (~6700K), this might be a better balance since the warmest tint (~3100K) will also not be as warm as I’d like either (8C1/8B4, ~2700K).
For reference, here is the Cree tint chart. It might help explain what things like “1D” and “7B3” actually mean. (click for a bigger version)
Step 1: Taking it apart
I tried to take the bezel off by hand. It wouldn’t budge. I tried to wrap leather around it and do it again so I could grip tighter. Still no luck. So I asked BLF, thinking I’d need a strap wrench or something, and felt a bit dumb afterward because the solution is really simple. Just push it down against the carpet and turn. It comes right off:
Easy.
But it occurred to me that I didn’t need the bezel off just yet, so I screwed it back on gently to keep everything in place for now and avoid damaging the reflector.
Next I needed to get the driver out. I hear that the easiest way to do this is to remove the button, stick something in the hole, and press it out from the inside. The button’s outer ring unscrews if you have the right tools. I used my machinist calipers (a.k.a. dividers):
In retrospect, I think it probably would have been better to use pliers to grab the outer ring and turn it just far enough to make it loose. After a quarter turn or maybe half turn, it can be hand-loosened. Pliers would have gotten a better grip than dividers.
With that off, the next step was harder… pulling the button itself out.
Mine was pretty wedged in there, and I had to grab parts of it with needle-nose pliers or my fingernails and just wiggle it until it started to come out. It took a while. Eventually it popped out and gave me access to the inside of the light though.
Note, in the above picture, that the switch’s O-ring is behind the circuit board for the switch. That means the switch itself is completely exposed to water if the SRK gets wet. When putting this back together, it might be a good idea to wrap part of a balloon or condom over the switch and through the hole its wires go through, to make the switch waterproof.
With the switch hole open, I was able to reach in and pop the driver out. I tried a few things, but what eventually worked was the back end of a crochet needle.
Note that the driver did not come out easily, and I was half-convinced I would break something in the process. This is because the driver was glued down in three places. This shows one of those glued spots:
Here’s what the other side of the driver and inside of the head looked like after getting the driver out:
This is the new driver I want to put in:
At this point, I was ready to start un-soldering things. But first I took a bunch of macro shots of the old driver so I’d have a good reference for where things go if I needed to revert any changes.
So, let’s get soldering! Er… un-soldering.
The first thing I took off was the switch wires. It may not quite look like it here, but they have just been removed and still have their old blob of solder stuck to the wire.
Then I thought about removing the other wires on the driver, the ones which drive the LEDs. But on second thought, it seemed a much better idea to leave those attached to the driver and simply remove them from the emitter stars instead. So I started removing the reflector:
That was pretty easy. Just unscrew it, unscrew the bezel, and it all falls right out.
At this point, don’t forget to remove and save the emitter centering rings. You’ll need those again later.
Then it’s safe to un-solder the old emitter stars. Pretty simple, with tweezers to lift each wire while touching an iron to it.
And then the old driver easily pulls right out, complete with red and black tentacles.
The emitter stars came off easily, since they were stuck with thermal paste instead of glue.
The last thing to do in the “taking it apart” phase of the mod is to clean off the nasty old thermal paste. For this, I use isopropyl alcohol (rubbing alcohol) and a tissue. Try to ignore the expiration date on this bottle… it may be five years past its expiration date, but it still works just fine.
And now the pill is clean, and ready to put new stars on. It is, however, rather a rough texture… and I don’t have any of the right equipment to smooth it out. You may need to click to zoom in to see how rough it is.
Step 2: Getting new parts ready
The new parts need a little bit of preparation before I can put them into the light. For starters, the driver doesn’t actually fit. It comes with tabs on both sides to allow you to custom-fit it to make it tight. So, I get to file the tabs off. I don’t have a dremel or anything like that though, so I’ll be using a hand file. This part might take a while.
I’m not sure how long it took, but eventually I got the driver to fit. I had to remove pretty much the entire tab on both sides, to make it look like it had never been there.
Then after cleaning up the mess I left from filing, I got out the new emitters. There are three different tints, from three different brightness bins. The warmest is a 80+CRI model, if I recall correctly.
Each star needed a bit of preparation too. They usually come with a few rough edges, and I don’t want anything getting between the star and the pill to interfere with heat transfer. So, I filed all around the bottom edge of each star to make sure no rough edges would stick out. A small diamond-coated file works well for this, and if you click to zoom you should be able to see one of the rough edges I was getting rid of.
Then to get each star ready, I pre-tinned the pads. I used flux this time, since I finally managed to open my flux bottle after running it under hot water for a while. And now I understand a little better what people mean when they talk about putting solder on the iron and then touching the iron to something. The flux pops and smokes and gets replaced by solder. It’s kinda cool.
Step 3: Hooking up the driver
Since I left the old wires attached to the old driver, and since I wanted to use thicker wires, I need to measure and cut some new wires. The old ones were 80mm long. I decided on 100mm for the new ones, since I want enough room to pull the driver out and reflash it if I want. This might be a fun host for testing firmware. It’s 22 AWG wire, so I don’t think the extra length should significantly reduce overall output.
Also, this will be the first time I get to try out my new wire stripper.
I made a set of three red wires and three black wires.
The driver only has one contact for LED + and one contact for LED -, and the batteries are in parallel, so I need to wire the emitters in parallel too, three wires to each contact. To make this easier, I twisted the ends of the three wires together and pre-tinned them.
My first attempt to tin the wires didn’t work. I tried holding solder up to one side of the bundle and the iron to the other side. Nothing happened. Then I remembered I have flux, and painted flux on the wires. Afterward, it worked quite easily. Hooray for flux! (I think this bottle might be older than I am)
The next issue is that the bundle of three 22 AWG wires is fat, and the hole they need to fit in might be a little too small. I guess I’ll find out later if I have too much wire…
I’ve never soldered anything like this before, so I wasn’t really sure how to do it. Eventually I decided to try holding the wire to one side of the hole while the iron itself was pushing on the other side of the hole, with the wire and both sides of the hole fluxed.
It seemed to work… I think.
For the next bundle of wires, I decided to make the stripped portion shorter.
… and it looks like both wires got soldered on okay.
After soldering, I could’t really tell much from the back side.
With those connected, it’s time to put the wires through to the front part of the pill. However, it seems that two 22 AWG wires just barely fit through the holes. I had a hard time cramming these through, and I’m glad I didn’t strip the ends first.
On the other side, it occurred to me that the LED wires may be extra-long, but I didn’t swap the switch wires. I hope they’re long enough to let me pull the driver out for easy reflashing later.
… and the negative wire bundle fell out. Oops. I guess I didn’t solder that on very well after all. It’s good I found out now though, instead of later.
So, I resoldered both wire bundles, this time making sure that they actually went through the holes. I think they’re solid enough now.
On the back side, I can see at least one wire sticking out, and I think the other pad has the wires all the way through too.
Next I connected the switch wires to the holes marked “SW+” and “SW-”. This was the same basic approach as the LED wires, but a lot easier.
… and the back side shows they are indeed all the way through the hole.
Step 4: Hooking up the emitters
Back to the front side, I still needed to strip the wires leading to the emitters. This was fairly straightforward… just pull a pair of wires forward, use the wire stripper on them, and push them back down into their hole. I tried to keep the stripped portion reasonably short, but not so short that the wire insulation would get in the way while soldering.
Next I decided where the three emitters should go, using a UV light to help me figure out which one was which so I could arrange them. It doesn’t show well in the picture, but in person the warmest emitter lights up bright orange, the neutral-ish emitter lights up yellow, and the cool white emitter lights up in more of a blue shade.
After arranging, I put thermal paste (Arctic Alumina) on the back of each star one at a time, placed the star with the “+” mark toward a red wire, and then slid the star around to help thin and spread the paste. I used tweezers quite a bit here, both for placing the stars and for sliding them around with pressure applied.
Then I tried to push the wires back even more, and bend the ends to get them roughly into the places they’ll need to be.
From there, it was a relatively simple matter to solder the wires to the star pads. However, with wires this thick (22 AWG), it was a bit difficult to get the wires exactly where I wanted them… and the wires kept pushing the stars out of place (or even lifting the star up off the pill). This is a little annoying for now, but it’ll all get fixed while putting on the reflector. (also, my soldering was pretty messy… still not very good at this)
Remember those centering rings from earlier? It’s time to put those back on. In my case, I found that my sloppy soldering kind of got in the way on a couple stars, so I had to file down two sides of the ring to make it fit better. I didn’t file those until after this picture though.
Step 5: Sanity check
At this point, every electrical component was connected, so I decided to test it to make sure I hadn’t botched anything badly enough to prevent it from turning on.
It lives!
When I took the battery tube off after the test, I noticed something. My little baby now has her first positive battery ring scratches! I’m a proud mother.
Step 6: The reflector
So, then I put the reflector on. It was kind of a pain, since the LEDs weren’t even close to centered. I had to put it on loosely, reach carefully into the reflector with needle-nose tweezers, press down hard on the centering ring, and try to drag the emitters sideways to fit. Then flip the light over, tighten the reflector screw a bit, and repeat. Eventually I had it tightened down with the reflectors centered into place.
And I put the battery tube back on. Or, at least I tried to… but every time it even started to thread in, the light would turn itself on. I didn’t press the button or anything…
So I had a short. My guess was that the reflector was shorting ground to one or more of the emitter leads. D’oh. My sloppy soldering and thick wires were probably too thick.
I removed the reflector again for a second quick sanity test, and it still worked fine with the reflector off. So, it’s a reflector short.
Normally, I’d cover the back of the reflector with Kapton tape. But I don’t have any Kapton tape, and don’t know any places to get some on short notice on a Sunday evening. So I used electrical tape instead. The easiest way I found was to put a piece over the reflector hole, then cut it out with an Xacto knife. Tweezers were also necessary, to pull and provide enough tension for the knife to cut effectively.
I did the reflector gradual-tightening-and-centering thing again with tweezers, and eventually got it back into place. I think I got a few small scratches onto the reflector in the process too, near the narrow end, and some fingerprints on the flat front part when I wasn’t careful enough. And some dust and hair. I did not attempt to clean it though, aside from simply blowing it off. I’ve heard too many stories about reflectors getting scratched even by microfiber lens-cleaning cloth. Anyway, here it is after putting it back on:
Note: Cereal_killer gave me a really helpful suggestion for handling reflectors… Wear a glove. ZOMG, why didn’t I think of that? No more fingerprints! Disposable rubber “exam gloves” are only a few cents each, and would be perfect for the job.
Step 7: Put it back together
Putting the rest back together is easy. Well, aside from the driver wanting to pop out because I didn’t get its press-fit quite tight enough. In any case, it went back together quickly. The only odd thing is that the battery tube now rotates to a different position when it’s fully tightened. My lanyard used to attach directly below the button, and now it’s about 100 degrees tighter. Not a big deal though, I can just re-thread it.
Tints and Beam
After getting everything back together, here is what the front of the light looks like when it’s on at the lowest setting:
… and how it looks when I shine it at a wall from nearly point-blank range:
If I pull the light back a little (less than a meter, IIRC), the hotspots all blend together into a white tint:
The overall tint came out just slightly warmer or more yellow than I intended. It might be better to use a 3* tint for the middle instead of a 4C tint. Or perhaps use a 1A tint at the cool end instead of 1D. It still looks very nice, but my BST-wide puts out a slightly more pleasing beam. But then, if your eyes prefer warmer tints, maybe this would already be “just right”.
Diffuser film
After using it for a little while, I decided that the SRK was too throwy. Normally it would be okay, but it seems like the wide-spectrum output is a bit pointless when there’s such a huge difference in brightness (and color temperature, I think) between the hotspot and the spill. So I added some DC-Fix to the lens to smooth out the beam. This makes the SRK much more of a flooder, and more useful up close (which is where I actually want to see colors vividly).
If you make a wide-spectrum light, I recommend making it floody to get the best and most vivid colors.
Adding the film is really straightforward. Take out the lens, clean it thoroughly, put it onto some DC-Fix after removing the protective backing, then cut around it with a thin, sharp knife. Rub it thoroughly to push out air bubbles, and it’s good to go!
Here’s the front of the light with the diffuser film on:
Or, approximately the same thing with some ambient lighting from the room:
As aoeu pointed out, the picture above looks an awful lot like Mickey Mouse with his face lit up by a smile. I think this mod made my light happy. 
After adding DC-Fix, I re-tested the SRK in my light box, and found that the results were about 3% lower than without the DC-Fix. So, this type of diffuser apparently sacrifices about 3% of the total output in order to make the beam smoother. Of course, I also didn’t recharge the batteries first, so it could potentially be a little bit due to less-fresh cells.
Output
Before any changes, my light box measured the light at 2112 lumens OTF. I don’t have the light box fully calibrated, but I think that’s at least in the right ballpark. I’d expect about 2200 lumens after calibration.
After finishing my changes (but before DC-Fix), I measured again. The results were:
- Level 1: 44.6 lm OTF
- Level 2: 86.2 lm
- Level 3: 300 lm
- Level 4: ???
- Level 5: ???
- Level 6: ???
- Level 7: ???
At level 4 and above, for some reason, my light meter bugged out. It flashed some sort of error at me. But I still wanted to know what the highest output was, so I did a really simple ceiling bounce test. I tested level 3 via ceiling bounce in a small white bathroom, then tested level 7, and based on a simple linear ratio, I got the following:
- Level 7: 3098 lm OTF
Very close to my earlier theory-based estimate of 3058 lumens!
I’ll have to get better measurements after building an integrating sphere, which probably won’t happen for a few weeks.
Also, I’d love to get tailcap amperage measurements, but the SRK makes that kind of a pain and my DMM only goes up to 10A. I’d probably blow up my DMM if I tried to measure the highest mode, or if there were any amp spikes during measurement.
So I think I’ll just trust that it’s actually 11.2A since the lumen measurement came out so close to my estimates. If any of the 7135 chips aren’t fully connected, or if something else isn’t fully up to par, I’ll probably never know.
Final thoughts
I did this mod during the day, so I haven’t had a good chance to really test it at night yet.
So far, I really like it but the interface can be a little confusing. I’m accustomed to Zebralight’s UI, where a short press turns it on in the high mode and a long press turns it on in the lowest mode. The STAR momentary-switch firmware does the opposite though, so I keep getting it backward.
I didn’t glue the driver in, and I made the wires extra long, so I hope I’ll be able to experiment with firmware on this host.
Update 2023 (9 years later)
This light hasn’t really had a lot of use since modding it. The new driver is nice, and the even newer driver is nicer (I gave it a Q8 driver with Anduril 2), but the LEDs never quite mixed right. I’d recommend using a single type of LED in this light instead of a mix.
Also, the BLF Q8 exists now, along with several derivatives. It’s basically the same thing, but pre-modded. I don’t think its host looks as nice as the SRK, but it works better overall and has a handy lighted button.
That said though, I don’t really use any “soda can” style lights very much. They’re just too big, not very throwy for their size, and I almost never need the amount of lumens or runtime they can provide. And when I do use one, it’s generally an old (no longer in production) FF ROT66 with a bunch of 219B LEDs inside, because it has a fantastic beam. Or sometimes a MT18S that I removed the optics from to make it a mule. If I ceiling-bounce it, the mule beam turns the entire room into an integrating chamber, which works great for photography since light comes from everywhere and there are basically no shadows.
Instead, I use smaller lights for general-purpose daily stuff, or throwy lights for distance. Like, instead of this particular SRK, I now have an Emisar D4Sv2 in XP-L HI, with tint ramping from 8A (warm white) to 1A (cool white). It’s smaller, about the same brightness, has more and better features, and the beam looks a lot nicer. Less runtime, since it has a 5,500 mAh battery instead of 12,000 mAh, but that’s okay. I almost never need to recharge it.
Anyway, this SRK mostly just looks nice on a shelf now. (and in BLF’s banner image… that’s this exact light up there in the logo) I’d re-mod it to give it consistent LEDs again, but I don’t really use it anyway so I haven’t bothered.
