a fully modded sk68 clone

There have been many threads about how to mod a sk68 clone (the last week the number of threads seems to have exploded even ). And whenever I join in with the replies my answer is: just don't, the sk68 is fine as is, any modding will make it worse or at least not better. The reason is that the XR-E emitter combined with the used aspheric is a very well matched combination. You can change to more modern leds, and that gives more output and better tint in flood modus, but throw modus is a different story, and let's face it: that sk68 has become famous for its throw. Most newer leds have a bigger die, just the die of the XP-E (and XP-E2) is as small as the XR-E, but it does not have the narrow beam angle of the XR-E, so light losses are bigger and despite the higher output, you loose throw. That old-fashioned XR-E is very hard to beat.

Well, I ignored my own advice and tried to improve the sk68. To do this I tried to address all issues that this light has. First a short summary of the sk68:

first the pro's:

*great emitter/aspheric lens combination

*cheap cheap cheap

*driver can handle both alkalines and 14500 no problem

*led is hard driven, most clones 1.5A+ on 14500

*in most clones one mode, I think that is nice for this light

cons/issues:

*the old XR-E is not as efficient as more modern leds, and the tint is very cool (to me that is not )

*hollow pill in the Sipik version and most clones, no thermal paste, thin led board

*even worse: very nice finning on a nice and thick battery tube, but that is not where the heat is: from the pill the heat first has to travel through a tinfoil thin tube of aluminium to get to those nice and beefy fins. I believe that to be a bigger bottleneck than the hollow pill.

*non-servicable pressed-in clicky, who nows how bad the electrical connections in there are?

*just one mode, some people don't like that

*14mm driver gives very few options for upgrading

OK, let's start optimistically with a nice hotrod coloured sk68-clone (under 5 dollars, shipped, from Banggood), this one has a xp-e led, making throw a bit less than one with the old XR-E, but well, it is going to go :-) :

I cleaned out the pill, drilled small holes an tapped m2 threading to screw the new board tight, made some room on the side for the screw heads. Not the tidiest job but it will work (*sigh*, tip : never try out tight fit new threading with a brass screw, the head will break off and if you try to drill the remains out, guess where your drill wants to go if it can choose between brass and aluminium ).

The above dx-driver was the only 7135 based 14mm driver I had lying about, one mode is what I like, I have just learned myself how to stack chips, so it's time for some beefing up :evil:

So that was two 350mA chips stock, I added 4x380mA chips, in total 2220mA worth of chips, I tested it with a Efest 14500 IMR on that XM-L and measured 2.2A :-) . Because it will be press-fit into the aluminium pill I made two little protrusions opposite to each other on the side of the driver from cram-shaped pieces of copper wire through one of the via's and along the side; soldered it tight and filed it to shape. this will keep the driver in place and provide a good electrical connection for the minus of the driver.

The led is going to be a dedomed xp-e2, dedoming is the only way to beat the XR-E in throw. It is a R3 binned 1D tint from Illumination Supply, in my knowledge the only easy source for the xp-e2 at the moment. I reflowed it on a 16mm Noctigon board and during cooling I dedomed it using a scalpel. I soldered and screwed the led-board and driver into the pill, Arctic Silver between the led board and pill, and tested the current: 2.15A :-) . I do not worry about the hollow pill: the Noctigon board is much thicker than the stock board and copper has twice the conductance for heat.

Watch that piece of silicon left around the bond wires of the led, you will see it later on in the beamshot :-) (not affecting beam quality at all by the way)

And now for the bad heatpath from the pil to the battery tube-and-fins. I had a brass soldering sock that had exact the right dimension to make a heat-conducting ring between the pill and battery tube (I know aluminium or even copper would have been better but this is what I had and I think it will do a proper job).

This picture shows how short the distance is from the pill to the battery tube, if they had just bored it out a little less deep so that the pill screws directly onto the thick part of the body...

From left to right: without the brass ring, with the ring in place, with Arctic Silver added, and a picture showing that the pill just screws tight onto the brass ring leaving a tiny gap under the pill.

And now the switch; there's not much I could do, found it too hard to get the press-fit switch out, so I sufficed with soldering some copper wire from the minus contact plate of the switch to the top of the spring to reduce the resistance somewhat. When measuring the amperage of the light there was no difference with or without the switch in the circuit. That said, when there is enough voltage overkill, this lineair driver should compensate for extra resistance and output a constant current, so that test did not say much about the improvement.

At last some cosmetic improvement of the beam. A dedomed led has much less back reflection of light, that is good for the beam already, but to make it better even I spraypainted the top side of the pill with matt heater paint. First I covered the led with a plastic cup (it was an internal part of a cheap ballpoint pen) and taped off the pieces that did not need paint. Two layers were sufficient to get a good black cover.

So here it is, next to a untreated clone with a XR-E, nothing much to see exept a bit thicker black o-ring between the lens part and the rest of the head (because the led board is a bit thicker than stock, the lens had to move up a bit for correct focus in throw mode).

ok, the top view is different :

Performance,

compared to a stock sk68 clone, that draws (tail measured) 1.53A with a Efest 14500 IMR battery (all measurement were done with this battery). The exposures of the beamshots vary to show the different aspects of the beams:

flood compared at 1 meter from wall, mod light left:

throw compared at three meters from wall, mod light now right (sorry):

throw mode beam quality compared at 30cm from wall, mod light left:

Outside, trees at 25 meters, mod light on the right (cheated a bit, tiny bit over-exposed compared to reality ) :

Some numbers :

current on fresh Efest 14500 IMR: stock 1.53A, modded 2.16A

output in lumens zoomed out: stock 220 lumen, modded 250 lumen

output in lumens zoomed in: stock 135 lumen, modded 138 lumen

estimated efficiency on 14500 zoomed out: stock 35 lumen/watt, modded 26 lumen/watt (ouch, that is bad! )

throw in lux at 1meter, measured 1 minute after switch on: stock 16 klux, modded 39 klux (yes! )

Runtimes compared:

The drivers in the two light are quite different, you expect a flatter output from the lineair driver in the modded light, but heating-up effects will show in the output as well. I find it difficult to interpret, . (I should have measured a few more minutes by the way, it looks like the battery of the stock light is just about to give up after 23 minutes, I also expect that from the current/battery capacity, a few more minutes would have shown that).

The modded light gets hotter, and much more quickly. It draws more current than the stock one, but I think that the attempt to improve the heat path was also succesful. Both lights can easily handle a full battery-drain-run from an IMR without any cooling. The modded light gets painfully hot after 10 minutes but survives well (at the moment I have done 7 such runs with it and it still works fine).

Conclusions

Well, I think the modded light is a winner: assuming that throw is the most important aspect I think I succeeded in improving the sk68, throw went from 16klux to 39 klux. Also the beam pattern is much nicer, the led tint is definitively much nicer, and a speckled hotspot (still meh) I find less ugly than a striped one. Some things have gone worse though: I needed a much higher current than the stock light, so the runtime is shorter. And the efficiency of the modded light is the worst of any of my led-flashlights (but I expect no different from a hotrod light ).

But there is a different and perhaps more significant conclusion. Look at that under five dollar stock sk68 clone. I think the performance is top-notch! Starting at a respectable 220 lumens, with its lack of heatsinkpaste, its hollow pill and the dramatically bad heat path to the body tube, its output just goes strong throughout the 23 minutes run. No-one will notice the only 25% output drop over 20 minutes without a lux-meter. And nothing melts, it just keeps working! It is just laughing at us at BLF, all worried about heatsinking issues, thinking of mods to improve that, THE DESIGN WORKS AND IT WORKS FINE .

Thanks for reading. :-)

Very nice mod djozz. You have got your creative juices running. Thanks for sharing.

nice!

Certainly hearkens to the day of candy-apple red hot rods! Put a lot of love into that!

Nice mod thanks for sharing.

Probably the best mod thread I have seen on a SK68!

Great write up and information.

Thanks for sharing.

Thanks guys, it was fun to try everything to get the sk68 right. (and I still don't like aspherics for their ugly square hotspot )

Cool

Nice work!

Exactly the same findings as me, my test of hollow-pilled UF-838 also showed little drop in lumens output throughout the run. And that’s XM-L running close to 3A.

So really, regardless of what the general opinion is, hollow pill does NOT necessarily equal bad thermal transfer. The hollow pill design works just fine. IMO it even works better than flat pill using silicone type fujik if not pressed properly. It really is much more important to ensure proper contact of the metal.

Oh, very nice build by the way. :slight_smile: I did one similar to this but using this:

http://www.buyincoins.com/item/29943.html

The head size is similar but much easier to mod - 16mm driver AND much beefier 18650 cells for the runtime. My result is quite similar to yours.

Thanks!

Here is a drawing how a hollow pill will transfer heat better than a solid pill with a too thin alu plate under a board that is not completely flat (it sounds theoretical, but this is the case with one of my cheap flashlights). It is actually the same as what happens with the blob of Fujik that you describe

Not only is the star curved but the center of the pill tends to be raised because of the lousy machine work.

Nice mod, BTW!

I don’t understand how this light can perform so well…
According to this SK68 comes with 17.7 mm lens (the convex part only).
I took the Enderman’s calculator, entered:

  • 92% glass transmission (about what is expected for uncoated lens)
  • Clear Aperture of 1 (a perfect lens which this is clearly not)
  • 17.7 mm diameter
  • 172 cd/mm² surface luminosity

I got the 39 kcd that you did. But….

  • I assumed a perfect lens - according to these calculations you’d need more than 172 cd/mm² to get 39 kcd.
  • According to sma dedomed XP-E2 Torch maxes out at 141 cd/mm² and does 138.6 at 2.2A.

So… 29 kcd would already be very high according to these calculations.

Does anyone have an idea what’s going on?
Maybe the calculations are wrong?
Maybe your SK68 has a bit larger lens?
Maybe the lux meter that you used 6 years ago significantly overread the output?
Maybe it’s an editorial mistake you actually measured 29 kcd?

wow, a blast from the past. I have one of these lights somewhere but could never go back to it :slight_smile:

Sure a blast from the past, six years ago :sunglasses:
I still have the light around somewhere, I will re-measure it.

Did you take into account that the led is dedomed, which with those old XP leds doubles the throw?

Yes, sma tested it dedomed.
It’s very nice that you’re able and willing to recheck, thank you. :slight_smile: