March 06, 2004
A Discussion of the Kinkajou Light Source
The attached page contains an archive of an ongoing email discussion over the proper light source for the Kinkajou Microfilm Projector. The projector currently uses a LumiLEDs Luxeon Star V LED (5-watt Lambertian White, Part Number LXHL-MW1D) rated at 25 lumens. There are major issues with the LED heat and reliability. One proposal is to switch to a halogen bulb.
To join the discussion, please send us your comments either via email or by using the comments link below.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Tim McNerny
Date: Wed, 11 Feb 2004 14:27:18 -0500
I looked over the WPI projector report. It looks like they really did their due dilligence. Regarding the driver for the Luxeon LED, I noticed that there is currently no thermal feedback built into the drive circuitry. For the 1W LED, this may be less critical, but I predict that it will become an important issue for 5W LEDs.
I know from experience that Luxeon LEDs are easily damaged when they get too hot--either they fail "hard" or the gel inside gets cloudy. This means that "full brightness" is not a fixed voltage or current. Rather, it depends on the efficiency of the cooling system and the ambient temperature. A switching power supply will be a must for the 5W LEDs, either a clever analog circuit or a microprocessor-based design. The microprocessor can share the duty of driving the LED and charging and protecting the battery.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Tim McNerny
Date: Wed, 11 Feb 2004 17:46:03 -0500
> Heat is a concern with the 5W LED, hence the big heat sink
> and fan (you can see the optics assembly and the heat sink
> in the attached photo). I've also been interested in the
> possibility of cycling power to the LED at a high frequency,
> something not visible to the eye, as a way of reducing power
> consumption and heat loads.
Funny you should mention that, because the [0.1-watt LED] flashlight I showed you "chops" the power to the LED at just under 100kHz, and sure enough, your eye can't see it. But there's a catch: modulating the power reduces the apparent brightness. For a given cooling system efficiency and ambient temperature, there will be a an upper-bound on the DC voltage or the "chopper's" duty cycle used to drive the LED.
This is why you want a thermistor in close proximity to the LED to tell the driver if the LED is about to burn up (or lower its lifetime). The power supply can either reduce the voltage or reduce the duty cycle so it delivers less power to the LED, and thereby keep it within its specified temperature range. The good thing about a "smart" driver is that it can adapt. If the fan fails, or it is a particularly hot day, the projector won't be as bright, but at least it won't kill the expensive LED.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Tim McNerny
Date: Sat, 14 Feb 2004 02:57:24 -0500
I'm glad you got me thinking about reliability. It makes me want to recommend that we use off-the-shelf modules as much as possible. [...]
Using a fan in general bothers me somewhat, because it is the component that is most likely to fail in a dirty environment, but maybe it is a "necessary evil." At least they are cheap enough that we can probably afford to ship a spare with each projector. [...]
I'm starting to think that the Kinkajou product is 40% technology, 60% risk management. Then again, maybe I'm preaching to the converted.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Limor Fried
Date: Thu, 19 Feb 2004 19:21:23 -0500 (EST)
My reason for not using LEDs in this project is as follows:
Con LED:
1) really expensive (luxeon 5w is $40, 1W is $10)
2) 5w luxeons' life is not much better than halogens if white (a couple hundred hours) and are offcolour if cyan (dont know how bad this is)
3) they run at 3.5V, which means that either you run off of a 6V with a linear supply and lose 40% of your power to the resistor, or run a dc-dc converter, which is somewhat expensive ($5) and also is only 80% efficient. OR run 3 of them (1W?) in series on a 12V. thats probably the best way to do it with luxeons
Pro halogen:
1) halogens are about as efficient as LEDs
2) $5 a piece retail and $5 for a ceramic plug. (although perhaps cheaper for more than 10?)
3) are manufactured in various wattages (5, 6, 10, 15, 20...)
4) come with glass-cover lenses, and reflectors in different spreads (10-30degree are available online)
5) manufactured to run directly off of 6 or 12 volts, so no power is wasted to passives such as resistors or DCDC convertors.
6) dont need a cooling system as long as there are vent holes
going with a halogen can drop your cost more than $25 and can increase your efficiency, which means a smaller battery. I think this is worth it. another nice thing is that you can still dim the lamp with a resistor. i use a 6V 5W MR-11 on my bike lite about 1.5" dia. they are sold online. you can also get MR-16's, 2" dia. MR-16s are slightly more popular, but it depends on your space considerations. the only sad thing about these halogens is that they are bi-pin, so you cant really go around plugging & unplugging them.
in my opinion, led lighting is good for one thing: color and color changing. they're not ready to replace filaments in low cost applications. they dont even come close :(
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Limor Fried
Date: Mon, 1 Mar 2004 14:43:08 -0500 (EST)
> We've been using the LED in the current beta
> prototype for almost a year, include six demanding weeks in
> West Africa and travel all over the country. We've also had
> a model of the alpha prototype, which I think also uses the
> white 5W LED, on the floor of the Museum of Science since
> October, which counts as a kind of burn-out test:
> http://www.designthatmatters.org/news/archives/000129.html
>
> We haven't considered halogens before now. The original
> proof-of-concept, built in Spring 2002 with Saul's help,
> used a 1W red LED. I don't know if the 2.009 team ran a
> thorough evaluation of lighting options, or if they simply
> went with the 5W LED because it was a direct scaling of the
> proof-of-concept.
well, i just asked saul [griffith] about the 1W [LED he used in the proof-of-concept design] and he said "it was a different project, smaller scale" do you know the EE that decided on a 5W led? I still think its a poor idea. lots more lost power, a lot more cost. I have my 10W halogen bikelite setup (same formfactor as 5W bulb, i just need more light to scare cabs with ;) here on campus, we could just try it and see if it works. itll be here every day...attached to my bike! its got a dimmer switch so I can turn it down to 7W output.
> The urgent problem relates to completing a second-generation
> prototype design by the end of the semester, so that we can
> build them over the summer. Our hard target is that in
> September 2004, we have to deliver something like 50 units
> to World Education that will run for 500 hours over a six-
> month field test in Mali.
Sounds like theres two things: optics and bulb. but from last speaking to you it sounded like you were going to ahve to redesign the optics anyways since they cost $100. I really think a halogen is the right thing, it is lower power, and will last a thousand hours before bulb replacement. for a few dollars more per bulb we can get bayonet (like flaslight bulbs) instead of bi-pin, which is virtually foolproof for bulb-changing.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Limor Fried
Date: Fri, 5 Mar 2004 05:05:03 -0500 (EST)
> To carry on the LED vs halogen debate, I'd like to get your
> opinion about efficiency and heat. Will DelHagen pointed
> out that, although the LED may require a less efficient DC-
> DC converter to use the 12v power supply, the LED itself may
> be much more efficient in terms of converting electricity to
> light. Do you have any sense for how this works out?
well, its one of those things where you can read the numbers, or you could just try it. i think we'd have to just try it. i know that 'in theory' halogens and LEDs are about the same. but it sounded like you had more of a $ concern than efficiency concern.
> Second, heat. Right now, the LED is separated from the
> microfilm by the roughly 1.5" thick condensor lens. We
> haven't had any trouble with the microfilm getting
> overheated, but it isn't something we've thoroughly tested.
> Do you have any sense for the relative heat output of the
> halogen with respect to the LED?
that is a good question. im not sure. my halogen is in a metal casing and exposed to air so its quite cool. 1.5" is a pretty large distance.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
From: Tim McNerney
Date: Fri, 5 Mar 2004 02:23:39 -0500 (EST)
Limor Fried makes a pretty convincing case that a small halogen bulb would be better than a 5W LED + switching power supply + heatsink + fan, in terms of cost, reliability, and development time. Also, the user interface for "your battery is low" is simple: the bulb gets dimmer. Plus we are more likely to find optics to match. I like the idea already.
Posted by Timothy Prestero at March 6, 2004 06:12 PMOne thing a halogen design might add is the ability to scale the output, perhaps dramatically higher, by changing the bulb rating (and perhaps some of the focal distances/diffusers). This is simply not cost-effective with LEDs.
The following may not apply to a 'base' version of projector, which does not require any form of active cooling (it has just a bulb, reflector, and baffled air-cooling). Seems to me that Limor et al. are correct in thinking that 7W halogens shouldn't need sophisticated cooling or fans. This isn't more power dissipation than a large nightlight bulb...
A cheap version of a halogen power supply could use distinctive base-pin spacings for different operating voltages/demanded wattages, with one pin socket as 'ground' or common for all the bulbs. This would scale the output, and perhaps activate different controls or devices, depending upon the load plugged in. To go from bright to energy-saving, simply open up and switch bulbs. You might need a purpose-made high-temp socket (e.g. ceramic) but this doesn't seem to be a very expensive thing to make.
A simple water filter, in conjunction with a diffuser, would easily absorb the portion of a halogen bulb's steradial heat flux that the lens system might allow in contact with the film. A small pump, lines, and an appropriate heat-sinking modality should be sufficient for almost any Kinkajou configuration; design the pump to run off the fan shaft if you want full cheapness. Extending the water filter to act as a 'jacket' around the bulb area can help greatly with deleterious effects of heat from the bulb on surrounding structures...
Sufficiently-purified or partly-distilled water for the filter should not be difficult to obtain as needed. (IIRC, you don't want straight distilled water because it can have bad effects on ordinary glass over time when heated) It may be desirable to have some kind of soluble additive for the circulating water.
There is one problem with halogens that bears consideration: fingermarks and dirt can induce shattering at high envelope temperature. This would indicate some careful instructions for bulb-changers, and perhaps some air filters and elastomer seals for the bulb compartment.
My feeling about the LED is that some form of thermistor should be provided *in the device's leads,* perhaps physically near the capsule that holds the actual diode structure. Theoretically, shouldn't a PTC thermistor act rationally to clamp the heat dissipation at a particular level, which would regulate the output brightness and temperature 'automatically' for any environmental and maintenance condition?
An alternative would be to provide the thermistor with a 'third' output lead from the LED capsule, and use this as a reference output to appropriate circuitry in the power supply. This design could be easily placed across the back of an existing LED module. With some care, the back of the module could be opened up and the thermistor 'potted' relatively close to the active diode. I doubt this constitutes an expensive operation, or one which requires extensive skill or time. It should be an interesting exercise to design the external components in a power supply which respond to the thermistor signal...
Posted by: R.M.Ellsworth at March 6, 2004 10:25 PM(also i just noted, 5 watt luxeon is actually 2 leds in series, a 6.9V drop. so it cant be run off of a 6V battery...didnt know this before. i will henceforth assume a 12V battery)
lumens per watt:
5 watt white luxeon: 120 lumens -> 24 lumens/W (we cant use this regardless because they burn out in a few hundred hours of use)
5 watt green luxeon: same, 24 lumens/W
1 watt white luxeon: 22 lumens -> 22 lumens/W
MR-11 6W: 80 lumens -> 13.5 l/W
MR-11 10W: 190 lumens -> 19 l/W
MR-16 10W: 400 lumens -> 40 l/W
theres also a large temptation to overvolt them, thus reducing the life but increasing efficiency.
for example, overvolting the MR11-6W will give 16 lum/watt but lower the life. the problem is that if you're running off of a 12V battery and you're using the 12V bulbs. undervolting an MR-16 10W by 10% bulb will get you half the lumens (200 lumens) and take the wattage down to 9W, for ~ 22 lumens/watt, like the luxeons. still, its 9W.
(see http://nordicgroup.us/s78/#Watts%20Versus%20Lumens)
and http://www.ukrecscuba.org.uk/ukdiver/torch/ge%20halogens%20tech.pdf - the section on overvolting and undervolting lights) basically, a 5W luxeon not that much more efficient, especially when you take into account that you lose 10% of your efficiency when you DC/DC convert as you will have to with LED's.
the MR-11 is smaller than the MR-16.
the question is how many lumens do we need to project a good image? 100? 200? 300? we can always distribute various wattage bulbs - 6W/10W/15W for example and have the user choose the best one for the classroom and duration.
Posted by: limor at March 10, 2004 09:30 PMin regards to fingerprints: MR-11 and MR-16 halogen bulbs can be purchased in a waterproof metal/glass enclosure that you can wipe off, but is ~1" away from the actual bulb/filament so bulb changing is safe. thats what most people use. there is also a dichroic uv filter on the glass lens.
Posted by: limor at March 11, 2004 12:35 AM
