Improved high power LED bike head light with integrated heat sink

After scouring the globe for the right parts, the following is an updated version of my copper fitting based LED bike head light. See the original high power LED bike head light with integrated heat sink instructable for the details on the earlier version. The latest incarnation uses a cheaper and more full featured driver circuit, an in-line waterproof switch, a waterproof connector, a more elegant mounting solution, a more robust and waterproof enclosure, and gets your whites their brightest.

Step 1: Materials

The bill of materials for this light is more extensive than previous incarnations, but the result is an improved piece of kit.

Housing:
1x ~1.25" long section of 3/4" copper pipe
1x ~1" long section of 3/4" copper pipe coupling
2x 3/4" pipe cap
1x ~4" section of 12GA solid copper wire
Gear for silver soldering

Electronics:
1x Cable with Trail Tech connector
1x Waterproof cable gland
1x 3-mode boost constant current driver
1x 2 Ohm resistor in 1206 or 805 surface mount size
1x LiFePo 6.4V battery pack
Misc wire for connecting driver to LED
Gear for electronic soldering

Light engine:
1x Rebel Endor star 3-up or 1x Cree XPG Indus star 3-up
1x Lens for Endor 3-up
Thermal epoxy

Mount:
1x 5/8" x 6" Velcro face strap

Comments

[vallka]

Nice work, just want to get an update, are you using the same bulb or have you upgraded?
would bulb like these work with your driver?
http://www.ebay.ca/itm/Cree-16mm-XML-U2-Warm-White-12W-High-Power-LED-Light-Bulb-Lamp-Torch-12V-1A-/120976007494?pt=LH_DefaultDomain_0&hash=item1c2abb6146

http://www.ebay.ca/itm/Cree-XML-U2-Warm-White-20mm-High-Power-LED-Light-Bulb-Lamp-Torch-DC12V-1A-/110942545006?pt=LH_DefaultDomain_0&hash=item19d4b0e46e

[jmengel]

I've used the XML's (on this motorcycle e.g.) but not the XTE's.  XML's are the best available and due to their large die size have much better thermal performance. No multi-LED stars though so you need to run a single XML at high current which is a bit of a challenge since the cheapo drivers are not usually up for running 3000mA.

We'll see if I can find the time to document some XML bike lights in the future. Basically it is the same as above except you use a single lens and difference driver.

-J

[vallka]

Thanks for the reply, I am using this driver to run 2 xml's in series.
http://www.dealextreme.com/p/t6-2500-3000ma-3-mode-regulated-led-driver-circuit-board-for-diy-flashlight-4-5-18v-128269
I am going to us part of your design and a few others, am working on brazing fins to the copper tube as well as a few other idea's.
Thanks for all the work you have done and posted

[jmengel]

different driver

[vallka]

Or this one?
http://www.ebay.ca/itm/NEW-Cree-XTE-XT-E-White-LED-w-14mm-Round-Base-Up-456lm-5W-/390441439400?pt=LH_DefaultDomain_0&hash=item5ae81f90a8

[leeryan]

I really love the design of this. Very helpful instructable. I would be really interested in running this light off of a 6volt 3 watt hub generator, and possibly adding a capacitor for a stand light. Any tips.

[jmengel]

You'll need a different driver, a boost mode (like the one pictured in step 3) rather than buck mode ( like the one pictured here: http://www.dealextreme.com/details.dx/sku.13557 ) driver since the Vf of the 3 LEDs in series is greater than 6V. Most of the boost drivers available cheaply on DX are multi-mode, meaning the have hi, lo, strobe, etc modes that are cycled through by turning on and off rapidly. The problem with this on a generator driven setup is that when your capacitor stand light accumulator drops too low, the light will cycle the mode, so that when you start again the mode might be strobe, or even off. Kind of a pain.

Also, for a generator light, you'll probably need to set the current even lower unless you typically ride fast (>15mph). I lowered the drive current to 500mA on high for about 5W. You'll want to get down to around 3W, or 350mA. The good news is that you should still be able to get 350 lumens or so at 3W, and it will run cooler.

[tastydesignboy]

I am having a bit of trouble getting the stuff from batteryspace as the postal charge is nearly $80! This kind of makes the whole thing pointless for me as I am sure I can find a bike light as bright for less than the total cost. Unless I can get these parts here in the UK. So, I have 2 questions:
1. I don't really know what i'm doing so may need help on a battery pack and what to do to the driver as a result. 3x good UK sites are: rapidonline.com, http://uk.rs-online.com and http://cpc.farnell.com.
2. What do I use to charge this? I do have a multi power adaptor unit which has a volt selector for 1.5,3,4.5,6,7.5,9,12 volts.

[jmengel]

The battery pack you'll need depends on the driver as well as the LED and planned usage. I'll assume you are using the XPG 3-up star with a Vf of around 10V and max current to the LEDs of 700mA. If you are going with the multi-mode driver I specified, you will need a battery pack with less than 8VDC and 7Wh capacity for each hour of max current runtime. Watt-hours (Wh) can be calcuated by multiplying the nominal battery voltage by the amp-hour rating (Ah or mAh). In the case of the specified 6.4V 2400mAh battery pack, the Wh rating is 15.36Wh.

If you use a buck type driver (like the 13557 driver from DealExtreme) you will need a battery pack with more than 12VDC with the same Wh rating.

The charger will depend on the chemistry. Your multi-power adapter unit doesn't sound like a charger so I wouldn't use it for any battery. Lithium battery packs will require a lithium charger for safety and longevity. NiMH and NiCd will work with the same charger in most cases.

Use any source you feel comfortable ordering from to pick a battery pack and charger and I can take a look at it before you buy. I don't have the time to do your shopping for you unfortunately. I'd suggest a radio control hobby store as a good place to look. Or you can bundle a bunch of regular AA type cells together.

[tastydesignboy]

Thanks for the explanation. I am a little unclear. You say if I use your driver (your link provided in the shopping list) I would need less than 8VDC and 7Wh capacity for each hour of max current runtime. But the battery you are using is 6.4V 15.36Wh which has a higher Wh rating.

I did find this:
http://www.ebay.co.uk/itm/6V-1600mAh-NiMh-Battery-pack-RC-Car-Receiver-Futaba-/290560687908?pt=UK_ToysGames_RadioControlled_JN&hash=item43a6c43f24

Will this work?

My aim is to get maximum brightness for minimum 2-3 hours.

[jmengel]

Yes, the battery pack I listed has 15Wh of energy. Supplying 7W of power to the LEDs this will last a little over 2 hours. Since the power of the LEDs at the "high" setting is about 7W, you will need at least 7Wh of energy in your battery pack for each hour of runtime. If you want max brightness for 3 hours, you will want to find a battery pack with more than 3h x 7W = 21Wh.

The pack you linked is 6V x 1.6Ah = 9.6Wh which will only run on high for 1.3 hours or so.

[tastydesignboy]

The thing I am still a little unsure about is where you remove and add in a resistor to the driver. If I use a 7.2V 4200mAh battery then that's a very similar Wh to your battery but with a higher voltage. Will this be ok? Do I still need to mod the driver?

[tastydesignboy]

I found more batteries:
http://www.ebay.co.uk/itm/Overlander-2000mah-Eneloop-Nimh-Battery-RC-Aeroplane-Car-Rx-6v-/220893211351?pt=UK_ToysGames_RadioControlled_JN&hash=item336e42ead7

or

http://www.ebay.co.uk/itm/6V-3300mAh-SC-battery-pack-Radio-Control-Car-5x1-/170493664623?pt=UK_ToysGames_RadioControlled_JN&hash=item27b236dd6f

or will it be brighter with a 7.2V? I have a battery and charger like this one already:
http://www.ebay.co.uk/itm/Radio-Control-Rc-Car-4200mAh-7-2V-NiMH-Rechargable-Battery-Tamiya-Connector-M487-/190602324925?pt=UK_ToysGames_RadioControlled_JN&hash=item2c60c8abbd

Not sure what to calculate. Thanks.

[jmengel]

As I said, the Wh can be calculated by multiplying pack V and Ah.

For the first link, this would be 6V x 2Ah = 12Wh. Divide by the load wattage (7W) to get about 1.7 hours of runtime. I'd also mention that used battery packs are a bad idea as you have no way to know the pack condition and lifetime.

The second pack is 6V x 3.3Ah = 19.8Wh and thus 2.8 hours of runtime.

Both of these packs are suitable for a boost type driver for the star board with 3 LEDs in series (Vf = 10V)

EDIT: Seriously? You have been pestering me with all this and you already have a battery pack and charger? I suggest you use it. Over and out.

[tastydesignboy]

How many lumins do you think this is?

[jmengel]

I sold all the Rebel based lights I built before I made an integrating sphere for measurement of light output. My estimate is about 400 Lumens. Since I've built the integrating sphere, I've tested lights built with the R5 XPG 3-up stars from LEDSupply and when compared to a Cygolite MityCross 400 the XPG light put out about 50% more light for an estimated 600 Lumens at 650mA. Bright.

-Jon

[tastydesignboy]

Also your link to the lens is not working is it one of these that I need?
http://www.ledsupply.com/carclo-optics-3up.php

[jmengel]

Yes, this is the right lens family, you can choose the focus. I used the narrow focus lens. This lens also works with the XPG stars which I now use on the lights I build.

-Jon

[tastydesignboy]

I did speak to them about the lens and they advised the Elliptical to get a good mix of spot and wider light spread. What do you think? Also, can I use XPG stars in this spec above? Which on and where do I buy it?

[jmengel]

The focus on the 3-up lenses LEDSupply carries when used on the XPG is pretty wide to begin with. The elliptical is ~44x23 for the XPG and the narrow is 24 FWHM. When using the elliptical lens you will have to plan the orientation of your LED star as well as the lens in advance and then solder the housing correctly to allow the long axis of the elliptical pattern to align with the ground. Not impossible but in my opinion is unnecessary. This is due to the fact that with 44 degrees of half magnitude spread, the light will have very little throw and the effective candelas at a given point will be low and thus your visibility poor. Even the "narrow" conical lens at 24 FWHM is really a bit too wide for an all-around trail bike light and would be best supplemented with a head mounted light with a tighter beam and more throw to give better long distance visibility. For street riding, the narrow is good, but when doing a mountain bike trail at night you'll want some more throw when descending at speed. For that, I use a single XML with the tight 7 degree FWHM lens.

[tastydesignboy]

Thanks for all that I have gone with your recommendation. Just one thing. The link to the battery pack seems to be out of stock. What batteries are these? Is there an alternative?

[jmengel]

The listed 6.4V pack with connectors is really the best match and since it includes the connectors is really the best deal pricewise. If you can't wait for it to be in stock then the next closest match is a slightly bigger 6.4V LiFePo pack without the connectors. You'll have to buy the connectors separately and solder them on yourself.  A good coating of "tool dip" once complete is a good idea as well.

Good luck

[davidgroos]

I've been to many designs, I like this best and the comments are very instructive.

-- I bought 3 different lenses including this one: http://www.ledsupply.com/10507.php
--The 3-Up Indus Star appears to be the latest/greatest: http://www.ledsupply.com/indusstar-3up.php (talked to them via e-mail, very helpful).  Because of comments on the candlepowerforums I ended up getting the 5000K R2 since it seems this more neutral color is better for city biking than the 6500K, even though it is somewhat less efficient than the R5.
-- I too want to use an 18v Lithium Drill battery as power supply so looked at the driver you linked to in that specific comment.  As far as I could tell, that one couldn't deliver the power needed for the 3-Up Indus Star.  The person on ledsupply recommended the new 1000 mA buckpuck: http://www.ledsupply.com/buckblock.php mighty pricey it seems.

My question, will I be able to put in a variable resistor in series with the lights/driver to vary the light output?  How many Watt resistor? And I figure I can calculate how many Ohms...  Any other recommendations?

[jmengel]

First, I want to say that the heat sink capabilities of the copper fitting light described in this Instructable are not up to the task of cooling a 3-up Indus star when run at 1000mA which is 9.9W of output (3.3Vf x 1A x 3). I've run mine at up to 650mA which is 2W per XPG, and thus 6W total. My measurements with an integrating sphere show about 15-20% droop in light output when the LED runs in still air at 650mA. Based on the XPG datasheet that means a junction temperature of 115C or so. Extrapolating back this gives a heatsink performance in still air of around 10C/W which is to be expected based on the small size and relatively low surface area. Using these figures and the 1000mA driver at 9.9W total output, the junction temperature would exceed the specified maximum of 150C by a large margin. My estimate is about 180C which means the LED will die a quick death.

The buck driver I linked to (http://www.dealextreme.com/details.dx/sku.13557) is capable of driving the 3-up star at 650mA as long as the input voltage is high enough (Vf + overhead = 9.6V + ~2V = ~11V). The driver chip used on the linked buck driver is actually rated at up to 1200 mA drive, so you can change the resistor as needed as long as you increase the heatsink performance. The driver circuit is inside the housing with no good thermal path to ambient so derating the maximum current is needed. The IC has a thermal shutdown mode when die temp exceeds 160C, which I have never triggered at 650mA driving 3 LEDs.

If you use the linked (http://www.dealextreme.com/details.dx/sku.13557) buck driver with a 18V drill battery you can add a potentiometer to adjust the output current. Depending on whether you get a board with a PT4105 or PT4115E driver IC, the method of hooking up the potentiometer will change but it is possible with both designs. Just look at the datasheets for the ICs.

The buck block also has provision to adjust output current. Although it is not geared towards a simple potentiometer to dim, a circuit can be built to do so with a few resistors and a pot.

Your proposed method of hooking up a potentiometer in series with the lights/driver is not really doable since the constant current driver will try to keep output current the same within the limits of the supply. Putting the resistor in series with the LEDs will just be a heater and will not reduce LED brightness but will reduce battery life. You could put a potentiometer in parallel with the LEDs to bleed off some of the drive current, but it will largely be a heater. For example, if you wanted to halve the power to the LEDs from 10W to 5W, you'd have to shunt 500mA of the 1000mA drive current through the resistor or effectively 5W. So your battery life would always be at worst case even if your LEDs were very lowly lit. A nice handwarmer though.

[davidgroos]

I'm re-reading informational pages with this further info you have provided, able to make more sense of them... For fun I'll order a few of the buck drivers to which you linked (http://www.dealextreme.com/details.dx/sku.13557) since they are so affordable.  I'll experiment with controlling output current with 20k potentiometer as describe in the pages comment (and parallel capacitors and bypassing the rectifying diode).  Will look to the correct data sheet to see exact hook up as you described.

You mention that, with the buck block, one can build a simple current limiting circuit with a pot and a few resistors.  As I've already ordered this buckblock I'll need to build this circuit.  I can build from a schematic but don't have knowledge to design the circuit.  Do you know of an online schematic of this circuit or is it so simple a verbal description of it might suffice?

A few bicyclists on my block are watching with interest how it's going so your help goes to them as well.  Thanks.

[jmengel]

Check Figure 14 on the Buck Block datasheet. You only need a pot, unless you want to set a minimum dim level by adding another resistor in series to prevent the pot from turning the light all the way off. That would be to keep from inadvertently leaving the light on for long periods of time at very low light levels.

[hshyaminda]

dear sir, can i have a simple but relayable switch mode power supply circuit for use to drive a power LED = 3*1W. input supply voltage is 230v AC. and out put must be 7.2V , 800mA. thnxx. plzz..

[jmengel]

http://icanhascheezburger.com/

[hshyaminda]

And i try to down load pdf files from your site, but it dsnt wrk... can u plz tel me , dat how to down load those pdf s. thnxx..

[jmengel]

Click on them. Works for me.

[jmckittrick]

I'm looking to have a single "high" mode of about 750mA, can I simply use the top board only and solder a lower rated resistor in to get the 750mA? I will be using the light as a grow light on a timer and the on\off cycle will change modes if the board is left as is. Thanks for your instructables, they have been very helpful.

[jmengel]

You could probably set up the boards and only have one mode, although I am not sure off the top of my head if you could get that done on one or the other of the two boards alone. You'd have to look at the datasheet for the driver IC and reverse engineer the layout. The design follows the example application in the datasheet pretty much to a T so it shouldn't be too tough.

The other approach would be to use a cheaper single mode boost driver. I haven't found a good single mode boost driver on DealExtreme though.

Good luck.

[msmills]

After much searching, I've decided to build your light design but am confused as to battery choice. I'm opting to change to the cree triple xpg led which has similar power needs as yours (9.0 volts at 350 ma). I'm looking to get at least two hours on the high level. Will your battery run that long? I'm confused how you run a 6 volt battery on a 9 volt system. Thanks in advance for your help. Absolute newbie here.

[jmengel]

Using the XPG is fine. The battery has 14.4 Wh of which you should probably only get 10Wh out safely and maintain cycle life. The high current setpoint on my circuit was 500mA. With a Vf of around 10V at 500mA, this is around 5W. So runtime on high should be somewhat less than 2 hours when inefficiencies in the drive circuit (~90%) are included.

The driver is a constant current boost type driver, which takes the 6V battery voltage and steps it up to whatever voltage is needed to maintain the set point current. You must control the current in an LED, preferably with a constant current circuit or failing that at least a current limiting resistor (less efficient). You should never apply a battery directly to the LED unless you want to destroy it. The 9V is the forward voltage, the current when less than the forward voltage is applied is negligible. I suggest you read up on how LEDs work. The key is to understand the I-V (current vs voltage) curve.

http://en.wikipedia.org/wiki/Light-emitting_diode

[msmills]

I figured max watts of 9.9W @ 1000ma for led; then inefficiency of booster makes for a total of 11 watts needed. So if I want a run time of at least 2 hours I need a battery pack with a Watt hour rating greater than 22? If i use a 9.6 volt battery and the led uses 9.9 volts @ 1000ma, I would use a "boost" driver right? The other option would be to jump up to a 12 volt system and use a "buck" driver?

[jmengel]

One more thing, you can go with a buck-boost type of drive circuit, but the problem is that I don't know of any cheap ready made options out there for a DIY light. The LTC3454 from Linear is an option for single LED, and I'm sure there are chips available for more LEDs. You would have to build your own circuit though, either in dead bug mode or on a PCB. Typically the higher power chips have a thermal slug that should be connected to a PCB or other sink so dead bugging is tough to pull off, especially as you add more LEDs and current. As previous, it is easier to change your battery and stick to the cheap ready-built buck or boost drivers.

[jmengel]

If you want 1000mA, you're going to need a bigger heatsink housing in my opinion. The efficiency goes down with current, and the small copper housing I've shown will get REALLY hot with 10W being dissipated. Too hot to touch in still air. See the attached image for an option for increasing the power dissipation capability of the light. I attached an RC motor heatsink to the housing with thermal epoxy (it is aluminum, can't solder it on) which will increase the surface area. I have never run at 100mA so I can't say for sure if it will be enough, but the bigger the better. Increasing the body length and adding fins would be a good idea. It might be nice to have a hot light to cauterize any road rash if you have a spill though.

In terms of the battery versus the LED forward voltage (Vf), using a battery voltage so near the Vf is a tough situation. For a buck driver you need a few volts to drop, so that for a typical Vf of 10V you need at least 12V of battery. Similarly, you need a bit of room to use a boost circuit, so less than 8V battery for a Vf of 10V would be a good idea depending on the topology of the specific circuit. So I would say you stick with a 2-cell lithium battery (6.4 to 7.2V) or a 6 cell NiMH or NiCad (7.2V).

The battery capacity analysis is accurate. You'll need a bigger battery to get 2 hours at 11W, with more than 22Wh needed for 2 hours runtime. The amount of light you are going to get at 1000mA is going to be comparable if not brighter than a car headlight. How much of it will be usable is dependent on the optics, but the good news is that there are more options for the XPG stars for some reason than the Rebel stars. The lights that I have built lately with the XPG stars and the narrow spot optics have been really impressive at only 650mA. Good luck.

[jmengel]

Ooops, meant to say that I have never run a light at 1000mA, not 100mA above.

[msmills]

I actually plan on not modifying the boost driver; so my actual top wattage Should be 8.5W @ 800ma. That should help some with the heat and run time issues. I definitely like the heat sink idea and will use that, if nothing else the "wings" should help my bike fly. Using your switch; when you push the button does it automatically go through the different light levels with each click or do I have to alter the circuit ? I'm just looking for something like -1st click = "on" @ 800ma; 2nd click = 200ma; 3rd click = strobe; 4th click = "off". Is that the way this works or am I WAY oversimplifying. Thanks so much for your input so far and for inpiring me to take on this project, the copper should be a fun medium to work with.

[jmengel]

The driver cycles through the modes with "off" between if I recall correctly. So 1st click on high, 2nd click off, 3rd click on low, 4th click off, 5th click strobe, 6th click off, 7th click on high, etc. I think if you wait too long (a few seconds) between clicks you start over at 1st click again. This is based on my memory of the function, I sold all the lights I made with this specific driver and am waiting on the next shipment of drivers. My personal light uses a buck circuit a 12V drill battery and no switch. Keeps it simple.