Mission Impossible: XML conversion for the Philips SafeRide (parts 1 through 6)
OK, so the original challenge went something like this...
“I see that you’re making custom lights. What would you think about taking a decent Philips SafeRide (PSR) and turning it into a fire-breathing monster by retrofitting it with CREE’s XML emitters and new electronics?”
At first glance I thought, “No way, can’t be done,” but then after careful examination, I began to see that this might actually be doable, and not just by me, but by anyone with some basic shop tools and some DIY determination. So, in the spirit of documentation for anyone who might be interested in performing such a modification, the following is a step-by-step process (and lessons learned) that I went through to pull off this conversion. At the end of the writeup, I’ll show some comparative beam shots with a high-performance 1300 lumen, uniform beam bike light.
Step 1: Disassemble the light
It becomes very obvious when you open the light that you must be able to remove the entire front lens and reflector assembly in order to do this job. This realization will be quickly followed by a lot of head scratching and comments like, “how in the heck do I get this thing apart.” At first I thought there were some plastic “tangs” that were holding the clear lens cover in place, but after a couple of exploratory drillings, it became obvious that it was simply "glued" in place.
Finding the right tools before trying to pry the clear cover off may be the most important step. I happened to have some small Xacto-type chisels, but I suspect several small to medium sized flat-blade screw drivers (jeweler's type) may also work. The miniature chisel idea does work best though. In other words, something with a sharp edge that quickly widens. Getting started is the hardest part. I recommend starting from the middle of the UNDERSIDE, since this is where the most damage will occur. With the light body held firmly in place on a firm surface, you need to tap the first wedge in between the plastic and the edge of the aluminum housing. Leave it sticking in place, and move over slightly (maybe 0.5 inch) and tap in another wedge. Keep adding new “wedges” with only small separation and eventually, you’ll be able to “break” the bond between the plastic and the housing. Once the bond is initially broken, it becomes easier to work your way around the perimeter, gradually prying as you go until the entire piece pops out. There is no other retaining mechanism other than the glue. Carefully set this piece aside in a clean zip-lock bag as you will want to avoid touching the inside surface as much as possible.
The metalized reflector is held in place primarily by the two obvious screws, but there are also two holes in the LED board that act as “receptacles” for mating “pins” on the reflector (these will have to be removed later).
Removing the electronics and battery holder is straight forward (4 screws), but the LED board itself is simply adhered to the housing with some marginal adhesive. The board will pry off very easily, but then the remaining black adhesive will need to be removed and the area thoroughly cleaned with rubbing alcohol.
Last edited by pethelman; 07-13-2012 at 06:48 PM.
Mission Impossible: XML conversion for the Philips SafeRide (part 2)
Step 2: Modify the XML star boards for mounting onto the housing.
The XML boards that I used came from LEDsupply.com for around $10 each. They have cool, neutral, and warm. I used two “neutrals” for this project which yielded a very pleasing and natural looking beam, without being overly “warm.”
The major trick here is that the star boards are too thick to be used directly, and they must also be trimmed to fit. Essentially, they need to be the same thickness as the original LED board and also not extend beyond the front edge of the original board. I made a small paper template (see picture), marking the centerlines of the original LEDs to assist with the sizing and positioning of the new modified star boards. Keeping the new LEDs in the same spacial location as the original LEDs is very important for the optical alignment to the reflector.
A dremel tool with a cutting wheel will suffice for the trimming bit, but the hardest part by far was reducing the thickness of the star board. To facilitate this, I made a small aluminum jig, with a receiving hole for the LED and two 2-56 threaded holes for holding the board in place while performing some very crude “fly cutting” on the back side of the boards. Doing this with the drill press was not fun (oh for the love a mill), but not impossible. I took very little material on each pass, and basically “eye-balled” the final results. The picture shows the difference between the original thickness board and the final thin board. The final surface needs to be as flat and smooth as possible for the best thermal contact between the star board and the housing. I would recommend more than a 2 fluted end-mill bit like you see here in the picture. By going slow and then rotating and remounting the star board to get the areas that were initially around the screws, you can eventually get it done.
Last edited by pethelman; 07-11-2012 at 09:14 PM.
Mission Impossible: XML conversion for the Philips SafeRide (part 3)
Adhere the LED boards to the housing and modify the reflector to fit.
At this point you should have two XML star boards thinned and trimmed as shown, ready to be permanently adhered to the light housing. My thermal epoxy of choice is Arctic Alumina. It's extremely effective, non-conductive (electrically), and very permanent, so it's definitely a step on which to take great care.
Using the paper template as a guide, I was able to use a fine scribe and mark the same centerlines on the aluminum as I had drawn on the paper, this gave me a nice visual indicator as to the final alignment of the center of the LEDs. The epoxy cures fast, so once you apply a very thin layer to the star board and lay it roughly in place, you'll have only a few minutes to get it positioned correctly. You should use something like a flat blade screwdriver or both tips of a some pointed pliers to apply some decent pressure to the boards to get the epoxy to "squeeze" to a thin film under the star boards. The idea is to have only enough epoxy under the board to fill the "voids" or other surface imperfections. I definitely recommend doing only one at a time. The Arctic Alumina isn't cheap, but you can get the small serving syringes (two parts) for about $9 shipped. I get mine from these guys: BestByte Computers.
One of the major changes here is the overall size of the LEDs, and because the plastic reflector piece basically slides straight in with very little vertical free-play, you have to make some modifications to the reflector to clear the extra height of the XML domes as well as the conductive pads where the supply wires will be soldered to the star boards. You want to only shave away as much as is necessary, so this is pretty much an iterative cut/test, cut again, test again process. The pictures show the permanently adhered boards as well as the reflector both prior to and after trimming. In two of the shots, it is not all the way pushed in, but in one shot, you can see it sitting in its final position. At the end of the day, from the front, you should see the domes of the LEDs sticking up slightly beyond the inside surface of the reflector as in the last shot.
Mission Impossible: XML conversion for the Philips SafeRide (part 4)
Step 4: Fabricate the electronics mounts and adhere the driver board.
There might be some flexibility here in the choice of drivers, but I went with the TaskLED driver: the H6Flex. This is a switch mode "buck" regulator, and as such requires that the input voltage be higher than the required forward driving voltage of the LEDs. With both LEDs in series (HIGHLY recommended vs. parallel), the H6Flex works brilliantly with an 11.1V li-ion pack. I'm not going to explore all the other possible options here, but suffice to say, this solution represents a very robust and efficient approach.
The trick with the H6Flex (see it here on TaskLED.com) is to get it mounted in a robust AND thermally conductive way. I did not want to take the thermal tape approach, so I opted for a two-piece solution. First, I cut a simple rectangular plate from 1/8" thick aluminum stock, and drilled holes in the corners to line up with the mounting bosses in the housing. Next, I cut a special "L" shaped piece which could be permanently adhered to the bottom of the H6Flex and provide not only superior thermal heat transfer, but clearance for the incoming wires and mounting ears for screwing the assembly to the larger sub-plate.
In the pictures, you can see the two aluminum plates (both 1/8" thick) along with the general stack up arrangement. I'm noticing in the picture that I didn't quite have the board rotated enough, so the "IN+" terminal would actually be shorting to the aluminum in this case. When properly placed however, all of the through hole pads had clearance outside of the bar. The small surface mount component nearest the front of the light is the micro-processor in which the temperature sensing operation is performed, so I wanted to have this component oriented as close to the LEDs as possible. The electronics board is permanently adhered to the "L" shaped piece with Arctic Alumina.
Mission Impossible: XML conversion for the Philips SafeRide (part 5)
Step 5: Wiring and switch configuration
Getting down to the fun part now... connecting everything together. But before you can do this, you have to decide on your external wiring/connector. I used the standard 5.4mm/2.1mm connector common to several lighting systems on the market, and as such, I drilled an appropriately sized hole in the back side of the housing with small countersink for an o-ring seal. The hole is also slightly smaller than the diameter of the wire insulation, so it's a nice tight fit. I use 3M weatherstrip adhesive to hold the external o-ring in place. This is also one place where I screwed up. I would have liked to have put the exit hole for the power wire on the other side, so the power wire would be closest to the stem when mounted on the left side of the bars.
I wanted to keep the light as stock looking as possible, so I really wanted to maintain the original power button and status indication window. Upon disassembly, it was obvious that this was not a waterproof assembly, so I used some flowable silicon to remedy that situation. As it turns out, a standard tact switch epoxied to the back side of the aluminum base plate has the PERFECT spacing such that the original rubberized power switch literally rests right on top of the tact switch button with zero clearance. Got lucky on that one, but I'll take it. The tactile feedback from the switch using this setup is REALLY good. Nice sold feel with a very audible click. To help properly locate the tact switch on the base plate, just install the plate with the original rubberized switch removed, then take a pencil and trace out the area on the base plate using the hole as the guide.
One of the last major challenges was figuring out how to implement the status indication LEDs. With the original circuit board, these were surface mount components (three total) and you can see that they would have normally been "shrouded" by the white plastic on the switch assembly. I learned, quite by trial and error, that this was to keep the back spill light inside the housing from leaking into the status window. My solution for this was to "embed" three small red LEDs down inside the white shroud and then pot the whole assembly with black RTV. Works like a champ. With the 11.1V source and three LEDs in series, I ended up using a 680 ohm current limiting resistor in the status LED drive circuit. This gives just the right amount of brightness on the status LEDs without being obnoxious at night.
The wiring is very straight forward (refer to H6Flex technical info). I would only add that I prefer to use silicon insulated for good resistance to high temps at the LED boards. I should also point out that you can never have enough good strain relief on the wires (various blob-looking spots of clear glue). And lastly, you should use some high quality thermal paste, like MX-2, between the stand-offs (you can see this in one pic) as well as between the aluminum base plate and the permanently affixed electronics adapter piece.
More to come later....
Last edited by pethelman; 07-11-2012 at 11:08 PM.
Hi, very nice, Great job!!!! ....now a few of beamshots
Greetings -. Saludos
wow, some serious work going on there! What were the original emitters? Just wondering if it was possible to reflow newer emitters (Nichia 219s for example) onto the existing board..
As amazingly awesome as the mod is, I have to admit I have a twinge of disappointment when I saw you going with an external battery. I was hoping for an internal battery and USB charging - this is the way I'm heading with my commuting lights for shear ease of use and convenience (strapping a battery to your frame with snow sliding down your neck gets old real quick!). If you'd used an Lflex and a 2s2p 14500 pack, do you think it would have all fitted? As for a USB charger, the only ones I know are for 1s packs, but there might be others around.
Originally Posted by mattthemuppet
I'm not saying it'd be impossible, but it'd probably take a lot more custom work, and likely a lower power solution. The primary goal for the individual that I did the work for was to see just how much light we could get out of one of these things. I really wanted to stay away from the linear regulator, just because I knew that heat was potentially going to be an issue and I really wanted to get the run times as long as possible. With a small 3-cell, 11.1V pack, we were seeing almost 3 hour run times on the nominal high setting (I'll talk about drive currents and heat management in another post). One of the positives that was reported was that the overall lighter weight of the assembly allowed it to be mounted on the bar more securely (less vibration). I've got a few more steps to post, then the beam shots...
Oh, and the original emitters are the Philips Luxeon Rebels. I'm betting it'd be pretty tough, if not impossible to reflow an XML (see the size difference in Step 2) onto the existing board. Definitely outside of my abilities in any case...
Nice work! Will wait for beamshots
true, drive current would be more limited due to the reduced battery capacity and it would most likely be even more of a pain to mod. I just like all-in-one lights For someone who doesn't need to constantly remove their light an external battery pack is not a problem anyway.
Very cool upgrade for that light! I like what I have seen from beamshots from the saferide. I hope that the larger die size doesn't mess up the shaped beam much. I would have been tempted to salvage only the reflector and build a new housing for it. Expensive way to get a reflector though. Maybe someone on the forum knows someone at Phillips and can source the reflector, ah the stuff of dreams!
it's a shame no one has one of those fancy 3D mapping thingamajigs. Then they could make a 3D model, send it to someone with a gajillion axis CNC machine and we'd have more of them than we'd know what to do with. <ahem> does that count as dreaming?!
Very nice work, I wish id ordered some spare xmls when I was building my aquarium light.
Mission Impossible: XML conversion for the Philips SafeRide (part 6)
Step 6: Button everything up and re-install the front lens cover.
Really the hard part is over at this point. After making the few necessary solder connections, all that's left is to screw down the main aluminum plate, then screw on the electronics board, using a bit of thread lock at every location. Note the application of MX-2 thermal paste at all of the non-permanent mechanical connections. Be especially careful when using thread lock around the reflector hold-down screws. It WILL attack the plastic.
One of the last little modifications to the case will be the new lid hold-down scheme. When the original circuitry went bye-bye, so did the threaded insert for the lid hold-down. To facilitate the new screw holes, I first marked my best guess for the screw locations that would give the most "meat" underneath for tapping, then installed the lid and taped it in place. I then clamped the light body on the drill press so that I could drill exactly perpendicular to the outer housing. This aluminum has been "molded" so it's a bit like "pot-metal" in consistency. After seeing that there was only enough material to catch about 3 threads, I decided it'd be better to have at least 3 screws for added strength. You definitely DON'T want to over-tighten here, and for sure use lock tight.
In this picture, you can also see the interface between the clear front cover and the aluminum body. The same flowable silicon that was used on the power switch was used around the perimeter of the cover and should be a nice semi-permanent solution. Once dry for 24 hours, it will be very difficult (but not impossible) to remove the lens cover.
Cautionary note... you will inevitably get some dust or foreign material inside the surface of the metalized reflector. Lesson #1 learned here: Do NOT used canned air to try and blow it out. This air is not clean enough and will leave a hazy film on the reflector. Lesson learned #2: No matter how soft you think that cotton cloth is, it will leave scratches on the reflector. You really need the same kind of tools that you might use on high dollar camera lenses here. Or, in my case, I found that a 3M micro-fiber dustless cloth worked very well.
I really like the final brightness of the status LED. I fussed a good bit with the series resistor to get the brightness down to the noticeable, but not obnoxious level.
Beam shots and riding impressions are next...
One of a kind light!
I am the lucky owner of this supercharged light. Pethelman, thanks a bunch for all the ingenuity and hard work that went into this. This is truly one of a kind super light that is not available from any manufacturer at any price
I really liked the stock light for its beam shape. The key to this light is its reflector which puts out progressively more light towards the top of the beam with a sharp cutoff at the very top of the beam. This results in very even and long throw beam and does not waste light above the horizon. It is an excelent light for road biking. I just wished for a little more brightness (maybe twice as bright) when dealing the glare of oncoming car lights. Also the runtime was a bit lacking (only 70 mins on high). So, off went the unit to Pethelman who graciously took up the challenge of hot rodding it. An external Li-Ion battery solved the runtime issue. Regarding the brightness, the hot rodded version delivers in spades! My eyes estimate at least a 4 fold increase in brightness over the stock unit at the highest settings! It is much brighter than my 2001 BMW's Xenon low beams (although car lights throw a wider beam pattern)!
The light has 5 levels compared to 2 on the stock unit. Level 1 (50mA) is basically a "get me home" setting. Level 2 (220mA) seems comparable to the "low" setting on the stock unit and Level 3 (730mA) appears to be a bit brighter than "high" on the stock unit. Level 4 (1595mA) is where it really shines (pun intended). Level 5 (2800mA) is a bit brighter, but not really needed since Level 4 is so satisfying with 75% more runtime.
I was concerned that swapping in the larger XM-Ls in a reflector designed for the smaller Luxeon Rebels would seriously compromise the beam shape. But, I am happy to report this is not the case. The beam is a bit more diffuse - this actually helps fill out the near field - and the cutoff at the top of the beam is not as well defined. But this doesn't seem to increase glare to oncoming traffic a whole lot (at comparable brightness settings). But I am sure the light is no longer StVZO compliant!
My other concern was increased heat as result of more power. It is in the mid-60s at night here and the unit gets barely warm on Level 4 and slighly warm on Level 5, but never hot to the touch. The big aluminium housing seems to be a doing a great job in heatsinking.
The light is powered by an external 2900mAh Li-Ion battery. I performed some runtime tests at my desk with a fan blowing on the light for cooling. The light ran for 2hr 45min on Level 4 before turning itself off. This is really nice given the brightness and the small size of the battery. After this, it ran for an additional 6 hours on the "get me home" Level 1.
I will try to post some beamshots comparing the stock and hot rodded versions when I get a chance. Once again, a million thanks to Pethelman!
Last edited by NiteBiker; 07-15-2012 at 01:28 PM.
What an awesome conversion!
Kudos for all of the hard work and thought that went into this
Mission Impossible: XML conversion for the Philips SafeRide (Beamshots)
Thanks to "NiteBiker" for the opportunity to "reverse engineer" the PSR. It was definitely an enjoyable project, and I hated to see it go... snif snif
But on to the last details. The last great questions to determine for this project were how best to drive the LEDs, and would the plastic reflector take issue with the extra heat? After a lot of experimentation, it appeared that the optimal "sweet spot" for driving the LEDs was roughly 1.6 amps. In order to achieve this, I had to set the maximum drive level from the H6Flex at 2.8 amps. As NiteBiker nicely detailed, this means that the optimum run mode would be power level 4, with level 5 simply existing as a "boost" mode for occasional use.
The other factor (heat), was a major concern, since the original LEDs were fairly low power and utilized a remote temperature sensor (RTD). The temperature sensing now is on the microprocessor of the H6Flex, so there is definitely a time-constant involved for the heat to "spread" throughout the entire case and finally work it's way up to the circuit board.
I ran a test with the light running on level 4 (1.6 amps) with NO air flow and the H6flex temperature trip point set to it's lowest value (50 degC). The result was that the housing would definitely get hot, but it generally stayed below 120F before the temperature sensor kicked in and dropped the output to the more sustainable level 3. Although it's possible to run at level 5 for some time with air flow, I believe that turning the light on at level 5 in a zero-air-flow condition would probably allow the temperature to rise too quickly at the LEDs before the heat could reach the micro-controller. The good news is that the reflector appears to be made of polycarbonate and is very resistant to heat.
As usual, beam shots are a poor representation of what the eyes actually see, so I'll try to expand on the pictures. The first two shots show the PSR on level 4 and level 5 respectively. The first cone is 30ft from the front wheel, and subsequently spaced at 30 ft. intervals. From shot no. 2 (level 5), it can be seen that this serious bump up in power from 1.6 amps to 2.8 amps, only has a minor effect on the overall coverage and brightness. In fact, you can perceive a slight color shift toward "cool" at these levels.
The third shot is my own 1300 lumen headlight (DS-1300) on high. Right away, the big noticeable difference is the temperature color. Although the DS-1300 does not fall into the stark white "cold" category due to the influence of the warmer XPEs, the neutral XMLs are definitely a good bit warmer. The neutral color of the XMLs is VERY easy on the eyes, and watt-for-watt, I've noticed that it takes more power to achieve the same perceived brightness levels with neutral tint than with cool tint. So in that regard, the PSR is putting out an incredible amount of light. It's also easy to see that the progressive nature of the beam allows the light to have just that extra bit of effective penetration out around 120ft. The dark "dead-band" at 10-15 ft that was inherent to the original design (see shot no. 4) is also now largely filled in.
To my eyes, the beam from the modified PSR was nearly perfect. There was no sense at all of being "overpowered" by near field reflected light, and the throw was completely luxurious. Although the cutoff has been softened slightly, I did not perceive the light as being overly bright as I approached it from the front. I believe it would still function quite nicely in traffic at level 4 without causing problems.
The last two shots attempt to show more of the differences between the shaped beam and the uniform beam. Here you can clearly see the cutoff (particularly looking at the tree) as well as the reduction in side spill light (the only drawback compared to the uniform beam). All in all, the light is a complete joy to ride with, and if you enjoy projects like this, it's totally worth the few weekends of work required to make it happen.
Last edited by pethelman; 07-16-2012 at 11:26 AM.
looks really good and I love the NW tint. I'll be using those next time I make a mtbing light.
How's the side spill on the PSR and how does it affect riding on really dark roads or bike paths? It looks like it has considerably less than the DS-1300 (what optics?) and I rather like having some side spill so I can spot various creatures (2 or 4 legged/wheeled) coming out of the side of whatever I'm riding along.
The DS-1300 does put out a really decent amount of side spill with the frosted Carco optics on the XPG triple. You also get some light directly down around the front wheel as well as +/-180 due to the clear lens cover. It definitely has a very "open" feel.
Originally Posted by mattthemuppet
However, the modified PSR did have enough side spill to prevent the "riding in a tunnel" syndrome. The beam itself is quite wide, so it would probably do quite well on most trails. It's obviously not quite as "smooth" in the near field due to the facets in the reflector, but it did provide a good useable amount of side spill in the first 10 ft or so.
Both the stock and modded PSRs beams are excellent for road biking with good width and excellent throw. Of course, the modded PSR has brightness on steroids. The beam is a bit wider as well. I just got back from a MTB test ride on the trails and definitely the modded PSR's wider beam and spill gets rid of the slight "riding in the tunnel" effect of the stock PSR. The modded PSR works very well as a MTB handlebar light when complimented by a helmet light (anyway, the helmet light is considered a necessity for MTBing to look at turns and obstacles).
Awesome light! Perfect handlebar beam.
great feedback. The modded PSR beam certainly does have a lovely large smooth shape in the middle, right where you need it. A friend off BLF has done some beamshots of the Ledil Rita, which is supposed to provide a similar upper cutoff, so I'll post them when I get them.
That is an impressive mod! As I think someone has already touched upon, if we're at the stage where such ingenuity and effort is considered worthwhile to upgrade a commercial light because of its ideal beam shape then hopefully we're getting closer to some nice cutoff reflectors trickling down to the DIY/modding crowd. 3D printing is getting pretty cheap for prototyping, but I guess the barrier is getting the reflective surface, presumably through expensive vacuum metalizing.
It's threads like these that make me wish a crate of Philips or Lumotec reflectors would fall off the back of a truck someday...
I couldn't agree more as the reflector is this reflector is really good for a road beam.
Originally Posted by minisystem
Wow! What a great hack.
But after trying dynamo I've sworn off battery lights until the battery tech improves a lot, so what I'd really like to do is run the light off a dynamo. It needn't be 1300 lumens (not that that would suck), and Wouter Scholten claims 330 lumens from the stock LEDs using his custom dynamo driver (up from 0.7 to 0.9A if I recall correctly), but what do you think would be a reasonable option for LED replacement therapy for dynamo power? I wonder if the whole mechanism, including a standlight battery, could be stuffed into the case...