NYC Taxi LED rooflight project

An interesting project was dropped into my lap today. There is great interest in replacing the fluorescent light source currently used on taxi roof advertisements with something LED-based. We (meaning myself and the person I'm doing the project with on a consulting basis) want to make something which is better than the 4-foot fluorescent tube currently used.

First, a little background. Some but not all NYC taxis have a fairly large (~4 feet long) lighted rooftop advertisement holder. Here is a general idea of what they look like although the ones in NYC are somewhat different:



Current models use a single four-foot fluorescent tube for lighting. We're considering two options to replace this light source. One consists of two strips of 100 5mm white LEDs. The other would be Luxeon-based. Since the light from LEDs can be more easily directed where needed we obviously don't need to match the fluorescent tube's output. Point of fact, I believe the tubes were somewhat underdriven anyway. Regardless of which type of LED we use, the final device must be a simple bolt in and wire-up replacement. There is also a price constraint. One competitor is already selling 100 LED strips for $65 (I think per unit, although that might actually be for two units). In any case, we're aiming for a retail price of $65 total, which means these things must cost us less than ~$45 to make.

5mm LED strips represent one option, although I personally have my doubts if two 100 LED strips would provide sufficient lighting. Also, I have a lot of concerns over lifetime. Tests a few years ago showed 5mm LEDs degrading to 50% brightness after 6000 hours. Considering that the lights will be on whenever the taxi is in use, and taxis here run about 18 hours/day, this means either unacceptable dimming after a year, or the need to overlight to compensate for the expected dimming. Does anyone here know if the lifetime problem with 5mm white LEDs has been solved yet, especially with the lower cost HK LEDs which we'll be forced to use (at $0.43 each even in large quantity Nichias are out for this project)? In the quantities I envision needing (100,000 to 1,000,000 pieces), I've seen some HK suppliers offering high-brightness white LEDs for ~$0.11 in quantities of 100K+. I think for this project wide angle (40°) LEDs will be better suited. This gives an LED cost of $22 for two 100 LED strips. Add in the PC board, electronics, mounting system, and hopefully the cost should come in at less than $45. Assembly will cost significantly more here than with something Luxeon based, but the LEDs themselves probably cost less per lumen.

The second option is Luxeons. I'm thinking perhaps 10 L3's running at 700 mA (or less) should be sufficiently bright if the light is directed properly. Does anyone have an idea of how bright (in lumens/ft²) lighting advertising signs typically are? The problem here is obviously cost per lumen. I'll need to obtain the L3s at $2 or less each in order to remain competitive but we will be using large quantities of 10,000 to 100,000. Naturally, I'd prefer T bin but I couldn't care less about tint. I'll just make sure to match tints on each unit if I get a batch with different tint bins. I heard several times on these forums that L3s can be obtained at such prices straight from the manufacturers in China or wherever but only in huge quantities. Can someone here provide me with some contacts so I can inquire about this further?

Putting aside cost, another problem arises with L3s. They need to be heatsunk which means you can't mount emitters directly on a PC board. One option here is epoxying the emitters to a four-foot long, 1 inch wide aluminum strip which is turn gets bolted to the top of the ad box. This means the need to hardwire the emitters together which will obviously increase labor costs. Another option is to have a PC board with holes where the board connects the emitters together yet still allows them to be glued to a piece of metal. Any other ideas here? Think in terms of low cost and ease of assembly. Remember that the whole thing with 10 emitters, the driving circuit, and the PC board/heatsink must come in under $45 ($50 tops) assembled or it's a nonstarter. The taxi industry doesn't care if ours is 100% better if it costs $5 more. They'll buy the competitor's cheap garbage instead. Don't worry about the electronics end-I can probably design a halfway decent step-up driver for 10 L3s in series for less than $2 in parts.

Now about driving current and lifetime. I'm not sure exactly how long this thing needs to last but I'll assume at least as long as the car, which is typically about ~5 years. This means up to 50,000 operating hours. Unless 5mm white LEDs have improved they obviously won't cut it, nor will driving the L3 at 1A. At 700mA I should just barely make the lifetime requirement but should I cut down even further just for a safety factor? I'd rather even have 15 emitters running at lower current if it means 100,000 hours lifetime, or if needed to give 50,000 hours under the adverse conditions taxis operate under.

To take this all one step further, how are blue and true green 5mm LEDs for lifetime? Another option instead of 5m whites is red, green, and blue in the proper proportions. Red obviously has no trouble reaching a 100,000 hour life but what can be expected of the other two colors? I know they'll last longer han white and probably less than red, but can I expect at least 50,000 hours to 50% brightness? RGB will allow both increased efficiency and a little less cost than whites. Color mixing should be OK due to the large numbers.

I'll be finding out some more about the project in the next few days but any help/suggestions that anyone can give me will be appreciated. I'm hopeful that the competitor's version will turn out to be so bad that the cost constraints will be relaxed a bit. Naturally we can design things optimally given enough money, but it takes really clever engineering to get things out the door super cheap. And the taxi industry is about as penny-pinching as they come.

First of all, whats the point? I'm guessing that its going to be far more expensive than the fluorescent option. Is the only advantage going to be less downtime (no tubes to burn out)?

Mark

Sounds like a very cool project. I would be interested in learning more about it. Such as how you got involved, what they are looking to achieve, etc. It would seem that for a 5yr life, floresent wouldn't be a bad option so I'm struggling with the LED switch. I would definitely go with white LEDs because as they age, what if the colored ones dim or shift at different rates? This would affect your color. Also, I think the cost savings in material would be eaten up by increased manufacturing costs.

Very cool and I can't wait to hear more!

I just found out about this late this afternoon so I'm not sure exactly what the point of replacing the fluorescent tube with LEDs is. If I had to guess, I'd say that the taxi fleets are getting very short life with fluorescent tubes due to bad driver circuits and/or shock and vibration destroying the tubes themselves. We've already had similar vibration problems destroying filaments with some types of incandescent lamnps. Consider how bad NYC's roads are and the fact that taxis travel on them 18 hours a day. Also consider that I've seen some electronic devices designed for this industry which frankly make me cringe. I wouldn't be surprised if fluorescent ballasts are being replaced on a regular basis and burning up wiring when they go. This might be why LED is being looked into. At least you're not dealing with high voltages in a possibly wet/humid environment. I'll keep everyone posted as I find out more.

As to how I got involved, I worked in a taximeter shop for a few years repairing meters from 1988 through 1990. It was actually my first real job after graduating college in 1985. After the company closed the NY branch one of my bosses who is also my personal friend bought the business and has been running it since late 1990 as his own personal business. I've been doing consulting work for him since then. Most of this has consisted of repairing various devices such as meters, talking taxi boxes, and taxicam controllers. However, I've also done design projects for him for the various devices which the TLC mandates for taxicabs. Since my friend is well aware of my knowledge of lighting and LEDs this project seemed a natural for me. Assuming we can pull it off, they'll be a little bit of him and me in over 10,000 NYC taxicabs. /ubbthreads/images/graemlins/wink.gif

What about using electro - luminescent sheets? Or would they not be bright enough.

Electroluminescent sheets would definitely fall outside the cost constraints. Also, they don't last particularly long. I remember we tried EL for backlighting LCD taximeters in the late 1980s/early 1990s. In the field they needed replacing at least every year, and they required high voltages to operate. The EL panel and transformer cost over $20. We ultimately replaced them with 8 green LEDs costing about $2. Although these were abysmal in efficiency compared to those of today, they lasted a few years in service (without any sort of surge protection!), and were almost as bright as EL. Today we could probably get by with two LEDs and have much brighter backlighting, but since that meter is rarely seen I'm not bothering to do a redesign of that backlight board. Also, EL lighting is not particularly efficient (5 lm/W?). Given the current efficiency and simpler design requirements of LEDs, EL panels are not long from obsolescence.

OLED panels would be a natural for this is they were cheap enough and somewhat efficient but they're neither. I'm guessing that the actual amount of light which is ultimately used from the fluorescent tube is at best 1000 lumens and probably much less. It is entirely possible that the LED-based solution will actually last longer, cost less, and use no more power than the fluorescent. I'll be finding out more tomorrow.

I'd be concerned about using white LEDs to backlight transluscent advertising because of the spectrum deficiencies. They do OK with blue and green, but red often comes out a ghostly pink. White LEDs backlighting vinyl can do some other odd things to the colors depending on what sort of pigments & printing process was used.

I did some research on displacing a 20W/950lm tube along with some 6W & 8W tubes a few years ago and reached the conclusion that even with the labor and parts savings, the costs and color issues just didn't justify the switch. The warranty issues associated with the hot-running ballasts that seemed to be frying A/V gear in the same cabinet were another manner that only elevated the LEDs vs FL case to a crap-shoot (our customers could care less about energy efficiency and the florescent lights weren't really essential to operation they could deal with the occasional replacement).

I'd look at using clusters of RGB superflux "spider" LEDs or just go straight SMD on small modules. You can just plaster a half-dozen or so modules per side.

If you would prefer high-power LEDs, I think you'll have better luck with SMD power LEDs such as Osram's "Golden Dragon" - place it on a well-designed MCPCB and you'll be good to go for automated fabrication.

My understanding of RGB LEDs is that they have far fewer of the lifetime issues that white LEDs do - it's phosphor degradation that hurts their output. Blue LEDs can have some declining-output issues if they emit traces of UV that clouds the epoxy.

You might try contacting a specialty firm such as Lumilex - they might have a semi-standard product to ofer you like the 12V boards they're selling through led-supply.com

...or not. Many of these firms are niche manufacturers and charge a hefty premium for their wares.

Don't rule out some heatsinking for the 100 LED panels, maybe nothing major but that many LEDs will generate heat which if not gotten rid of will most likely decrease their life. Probably some simple fin based sinks would do fine.

About the only way to heatsink "commodity" LEDs (standard thru-hole LEDs) is to leave some exposed copper trace around the cathode side of the LED - there is no other good thermal path out of them.

Hello Jtr1962,

You might try contacting Mike at InReTech.

I believe he just finished some light displays and sculptures in the Portland OR area. He may have crossed similar bridges.

Tom

I saw a NYC cab while out to lunch today. The sign is one large plastic shell, perhaps thinner under the actual ads, with a rubber gasket around the base. Is it possible that the for-hire/off-duty lights might feed back into the ad lighting system?

The dome looks pretty weatherproof, and running fluorescents requires this. It's a sealed plastic cavity and bad for heat considerations. It has a large amount of space, so air circulation might be okay.

I like the "spider" LED idea personally. They're sort of a middle of the road between 5mm and Lux. More power/heat per unit, but not enough to require individual dedicated heatsinking.

How much work would it be to grab two cabs and have someone shine a LED flashlight or two inside the second one's sign to determine the required lighting. I know that my L4 will light up a large area.

It seems to me that the signs will be created to match the light source, so a more translucent panel with two or three luxeons may work.

I'd also look into adding a simple light sensor to shut off the thing in the daytime and when in brightly lit areas.

Daniel

I'm sure all of those of you who are in NY realize that many of the rooftop signs have gone with a combination amber LED graphic sign, next to an LCD screen. One displays sports scores, weather, etc. while the other displays ads.
There is also another amber LED sign that has three lines of text.

What would be nice to bring back in this fashion, and which used to be as much as a staple of the city, are the so-called "bat wings" (reminded me more of Batman ears) on the buses-- the similar flourescent tube backlit signs mounted above the windows, manufatured by Bus-o-rama. Those would be cool in full color, or sections of LED's and LCD like the cabs. The newest buses are the low-floor models and have the space above the window to place such ads.

OLED would be good for the interior electronic route strip maps in new subway cars, but that is still too new and not really out yet for the cheapo MTA to use. They are still trying to figure out a new, completely digital sign for that, so they don't have to replace plastic carsa with the route printed on when a train is moved to another route. In the meantime, I'm hoping the next batch of cars will use full color LED's for the interior and exterior route and destination markers. For the newst cars out now, they went with single color red or red-orange and amber for the outside, and 3 color RYG for the inside.

Just to chime in here, I agree with Gadget_Lover about an auto sensor, a cheap addition that will be a saving grace durring the summer. Also I think IdleProcess hit something with the SuperFlux (spider) LEDs. These are bright when compared to 5mm, and have a wide output angle. Also since they are through-hole, fabrication would be kept simple. I dont know what kind of ratio would be needed, but for this project, that is the way I would go.
Jeff

Your most best bet would just to be a LOT of 5mm LEDs, they are cheap - reliable and easy to get; use a piece of translucent plastic buffed up to scatter the light and get rid of the "spots"

Someone posted the article about the art we did, here is a link to it:

http://www.trimet.org/max/yellowline/stationart/rosequarterart.htm

And a webcam that sortof works:

http://www.katu.com/traffic/trafficcams.asp?id=24

The artwork uses a total of 800x 5mm SuperBright KingBright Green (and some Blue) LEDs - we run them at approx 10ma each

Someone also posted this:

----
About the only way to heatsink "commodity" LEDs (standard thru-hole LEDs) is to leave some exposed copper trace around the cathode side of the LED - there is no other good thermal path out of them
----

Don't bother, the incredibly small leads coming out of the typical 5mm LED are way too small to handle any amount of heat, its like trying to pull a car with a large chain that has a couple of very small links at the end

IMHO, The key is going to be to use 1 watt side emitters in series. In enough volume, you can get them in the price range you are seeking.

The Red for sure are essentially T bin, but you may need Lux IIIs to get T bin power from green. Blue is less power,but not as much is needed for color balance.

Look over the DCC at Lumileds. Some good guidance on how to get the color balance to work.

I would use RGB mix, not white. The lifetime is a non issue, the efficiency is often better (without binning).

Maybe put them in strings of nominal 10 - 11 Vf and a small resistor.

Thanks for all the suggestions/ideas so far. I didn't find out anything new today other than that my friend feels EL panels might be the best solution if the can be made in approximately 1'x4' sizes for under $40 and if the lifetime is longer than the ones I remember. I did research them a bit today and found that some have lifetimes up to 30,000 hours which might be OK, and their efficiency, while not spectacular at 5 lm/W, would be acceptable since 100% of the light would be going where we want it. Two questions then regarding EL panels. First, does anyone have some idea what custom made panels in white of about 1'x4' size would cost in large quantities (10,000)? Second, do white EL panels with a lifetime of 30,000 hours exist? I was unable to find the answers after doing some research on the subject today.

Assuming EL panels turn out to not be feasible, we're back to LEDs. I like the idea of using superflux LEDs. Have the lifetime issues on these been addressed, or will they have the same 6,000 to 10,000 lifetimes of 5mm whites? As for high power LEDs, my one and only reason for even considered Luxeons or SMD power LEDs like the Golden Dragon was because of the 5mm white lifetime issues. It is much easier to get even illumination with a few hundred 5 mm or super flux LEDs than with 10 or so L3s. There are also no heatsinking issues as with power LEDs. And as INRETECH said, 5mm and/or superflux are the easiest and cheapest to obtain.

I like gadget_lover's photocell suggestion and will probably incorporate it. This is a no brainer in that it more than doubles the lifetime. Few cabs operate between 1AM and 5AM, and daylight averages 12 hours/day over the year during the hours when they do operate. A photocell then should increase lifetime by a factor of roughly 2.5. Therefore, actual operational lifetime requirements of maybe 15,000 to 20,000 hours instead of 50,000 might be fine for this application.

What are the thoughts here about maybe overlighting by a factor of perhaps three? This would mean the LEDs would need to dim to 33% before light levels would be considered marginal. With 5mm LEDs dimming to 33% would probably take about 60% longer than to dim to 50% according to the lifetime charts I've seen. This would put the useful lifetime at ~10,000 hours (15,000 if I underdrove them at 10 mA). Combined with the photocell, operational life would be ~25,000 hours or more, or roughly 3.5 to 4 years. This might actually be adequate. If superflux whites have somewhat longer operational life than 5mm whites, I'll stick to those instead.

Color mixing with RGB is still on the table. My main concern here is color shift if the LEDs age at different rates. I'm assuming that the blue will be first to dim followed by the green. Therefore, if I start out on the Planckian locus the light will first start to turn yellow-green and then more and more yellow. If we're only talking a slight shift over maybe 10,000 hours we can live with it. However, we definitely don't want rotten cat urine green/yellow signs 3 years from now. /ubbthreads/images/graemlins/wink.gif And what color temp should I aim for when I calculate the RGB ratio? 6500K? 5000K? What is typically considered "white" for ad displays?

As for driving these, series groups of three with a resistor as HarryN said is one option I'm considered. I'll definitely put in a large filter cap and surge protection, and I'm seriously considered using a low dropout regulator to provide even brightness over the 11 to 14 volt range typically present in automobile power supplies. In the absence of that, I may opt for using one of the commercial series step-up LED drivers. Linear Technology makes one that can drive two series strings of 10 LEDs each. Therefore, I may consider making the 100 LED strips (if that's what we ultimately use) in 20-LED modules, and then connecting the modules together end to end. However, cost may force me to use the resistor, voltage regulator approach.

I might be able to obtain a sample in the next few days so that I can ascertain for myself the actual illuminaiton requirements in lumens. This will give me a better idea of exactly what my options are. I'll keep everyone posted as the project progresses.

I saw some pictures today so I have a better idea of what is needed. Here is the top light assembled:



The height is about 14" and it is roughly 14" wide at the base. The fluorescent tube is about one third of the way down from the top. We saw somebody else's samples using either one or two rows of 100 5mm LEDs. Neither really put out enough light according to my friend:




Before proceeding further I needed to estimate the light output of the fluorescent tube. Since they're using cheap tubes these probably put out no more than 2500 lumens at full power but most automotive ballasts underdrive the tubes by quite a bit. Therefore, let's say it puts out on average 1500 lumens. Based on the geometry here (there are no reflectors), at best two-thirds of the light, or 1000 lumens, goes where it is needed. The rest is wasted. Therefore, two rows of LEDs each putting out 400 to 500 lumens should be adequate. Obviously, 100 5mm LEDs are not up this task, at least not for another two or three years anyway. /ubbthreads/images/graemlins/wink.gif And I'm sure that at $65 per 100 LED strip our competition is using cheaper LEDs putting out perhaps 1 lumen each (and/or they're overdriving them severely). Therefore, I plan to use super flux LEDs. Power LEDs are likely out because of the heatsinking requirements and the fact that I can't find any source for obtaining large quantities of them cheap enough.

I'd like to bounce of few of my ideas off everyone here. First off, the 60° angle of most super flux LEDs works out perfect with a mounting system similar to the bottom picture. Nearly 100% of the light goes where it is needed. Second, the ones I'm currently looking at put out 5 lumens at 100 mA but I'll probably underdrive them at maybe 70 mA which should be good for perhaps 4 lumens. Any ideas if white super flux LEDs suffer from the same lifetime problems as 5mm? There is a better thermal path, so hopefully not, but any experiences with them? Also, are any readily available which maybe output around 6 lumens at 70 mA for perhaps $0.25 in quantity?

Color mixing is probably off the table. I came up with several mixes based on the relative flux of different colors from this manufacturer (5 lm green, 2.7 lm blue, 2.5 lm red). The best was a mix of 4 red, 3 blue, and 1 green which gives a CCT of 4200K and is very close to the Planckian locus. The downside is that it only gives about half the light of the white on average. While efficiency is not a major concern here, I want to limit the total power input to not much more than the tube used (perhaps 35 watts) for heat management reasons. Color mixing would require probably over 100 watts for adequate light output.

I'm of two minds about driving these things. I plan to do a modular approach and design the boards in 20 LED sections each connected to the car power in parallel. This way if there is a problem only 20% of the LEDs on a light bar will go out. I'm considering using the LT3466 to drive these but the cost ($2.24 each in 2500s) may make it prohibitive. Another alternative is a 9 volt low dropout regulator on each board, and running the LEDs in series strings of two each with a dropping resistor of about 30 ohms (for a drive current of around 65 to 70 mA). The first approach, assuming the LT3466 gives an overall efficiency of 85%, results in an input power at 12V of ~5.8 watts per 20 LED module. The series pass regulator approach means a constant input current of 700 mA, which is 8.4W at 12V. The board size of each module will be about 10"x1.25". Should I be terribly concerned about heat build up, and opt for the more efficient and more reliable LT3466 approach, and/or perhaps even reduce drive current to 50 mA? Or would 8.4W in a board that size not present a problem? We can always hopefully cut a few vent holes in the outer plastic diffuser such that air circulates but whatever water gets in doesn't find it's way to the LEDs. Any experiences managing heat in this sort of project? Obviously more efficient LEDs would help here, but it seems the super flux whites are not more than maybe 15 to 20 lm/W.

I also found out that a photocell is not needed. These are currently connected to the headlight switch, so they're only on at night.

That's about it so far. As it stands I need to hand make a sample within a week or so, and then start thinking hard about the remainder of the design. My main concern here is obviously light output and lifetime. It needs to last at least 5 years in service, which is 20,000 hours (12 hours/night), without dimming too much. I also want to make sure that we don't have any premature failures from surges.

I'd seriously look at tri-band CCFL tubes, they can definitely withstand the abuse if mounted right (I've taken them to 50G), low cost. www.lcdl.com http://www.jkllamps.com/index.cfm?action=fam&tid=2&fID=31 They are much cheaper from overseas, nearly free in comparision to prices for most companies here in the US.

There is alot of light that gets bounced back in there, I'd really consider painting everything but the surfaces you want light to escape out of, with a nice white the contains alot (+20% titanium dioxide) of pigment, skip the cheap zinc oxide and calcium fillers (if you can find a cheap rugged barium sulfate pigmented paint, that works too). Look for paints with a reflection of +93%.

You might consider either angling the LEDs to cast the light in a direction you want the audience to see (maximize the light) or using turning or focusing films such as is available from 3M under the TRAF and BEF name, or materials from Wavefront Technologies.

I understand you may figure the white plastic to act as a lambertian diffuser, but even getting that oriented in the right direction helps alot.

Astra Products makes some rather efficient diffusers that are made by Claire, http://www.astraproducts.com/

I'd second the concern on the white LEDs. The red looks very weak and pinkish like it was said. The blue is rather vivid. The green is even washed out a bit. It will depend alot on how narrow the "color filter" is the advertisement. RGB LED sources will make the colors surreal, more vivid than normal, and make the signs really stand out.

Any heatsinking you put in there for the LEDs will help out alot for lifetime, and as you know, underdriving also helps. You'll definitely need this to get the 20,000 hours you want.

Here is a constant current source that will cost you 0.22 ea, but can only take 30V (keep in mind automotive surges...) This one will take 200Vin but can only deliver 70mA, http://www.onsemi.com/pub/Collateral/NUD4011-D.PDF

Supertex makes a number of interesting chips, http://www.supertex.com/feature_flexfamily2.html

The MLX10801 is a multiple purpose LED driver for high power LEDs designed for automotive applications like interior lighting, puddle lights, turn signals, head lamps. The device can be calibrated for different powered emitters and optimal device parameters. The circuit is load dump protected for a 40V load dump pulse:
http://www.melexis.com

If you are going digital, there are a number of chips with contant current output such as, Toshiba's TB62701AN, or Maxim's http://pdfserv.maxim-ic.com/en/ds/MAX6964.pdf.

Another thing, if it is legal, with RGB, you could make the sign blink different colors or whatever, to indicate the taxi is free business.

Keep in mind, these LEDs are outside the "warm comfy" or "cool and comfy" interior of a taxi cab.

You could ask the LED maker for the MTBF of the LEDs over your temperature/vibration ranges. LOL, try and get that out of a HK/Taiwan LED supplier...

As far as your board layout, instead of skinny traces, use big wide copper planes to help pull the heat out of the LEDs. I see this big wide metal bottom that a fella could pick up too....

If you go high power, consider taking a look at the latest releases of the CREE XL7090 LEDs, they have much improved efficiencies, and can be all machine placed. CREE also uses the silicon gel inside that doesn't yellow like epoxy based LEDs do over time. There are some LEDs that have a better epoxy, but nothing that holds up to the light/heat like the silicone does.

The Golden Dragons made by OSRAM are lower cost when compared to Luxeons, but they are also amongst the lowest output of the high power LEDs.

Anyhow, that should be enough for the night, have fun.

[ QUOTE ]
jtr1962 said:
I saw some pictures today so I have a better idea of what is needed. Here is the top light assembled:

The height is about 14" and it is roughly 14" wide at the base. The fluorescent tube is about one third of the way down from the top. We saw somebody else's samples using either one or two rows of 100 5mm LEDs. Neither really put out enough light according to my friend:


Therefore, two rows of LEDs each putting out 400 to 500 lumens should be adequate. Obviously, 100 5mm LEDs are not up this task, at least not for another two or three years anyway. /ubbthreads/images/graemlins/wink.gif And I'm sure that at $65 per 100 LED strip our competition is using cheaper LEDs putting out perhaps 1 lumen each (and/or they're overdriving them severely). Therefore, I plan to use super flux LEDs. Power LEDs are likely out because of the heatsinking requirements and the fact that I can't find any source for obtaining large quantities of them cheap enough.

Second, the ones I'm currently looking at put out 5 lumens at 100 mA but I'll probably underdrive them at maybe 70 mA which should be good for perhaps 4 lumens. Any ideas if white super flux LEDs suffer from the same lifetime problems as 5mm? There is a better thermal path, so hopefully not, but any experiences with them? Also, are any readily available which maybe output around 6 lumens at 70 mA for perhaps $0.25 in quantity?

We can always hopefully cut a few vent holes in the outer plastic diffuser such that air circulates but whatever water gets in doesn't find it's way to the LEDs.

> too much of a hassle, does not fit with drop-in replacment/low cost model, water will get in as it has that sneaky capability /ubbthreads/images/graemlins/wink.gif


Any experiences managing heat in this sort of project? Obviously more efficient LEDs would help here, but it seems the super flux whites are not more than maybe 15 to 20 lm/W.

I also found out that a photocell is not needed. These are currently connected to the headlight switch, so they're only on at night.

> Excellent, low cost photocells are not reliable.

That's about it so far. As it stands I need to hand make a sample within a week or so, and then start thinking hard about the remainder of the design. My main concern here is obviously light output and lifetime. It needs to last at least 5 years in service, which is 20,000 hours (12 hours/night), without dimming too much. I also want to make sure that we don't have any premature failures from surges.

[/ QUOTE ]

Many of what I consider 'too many' unknowns here for an accurate assessment on feasibility of the business model.

Durability, in severe service conditions over 5yrs?

1st thing I would be considering; you say 10,000 cabs to be retrofit? Is this a city contract, open to private bidders. Sure it's somewhat OT from the questions being asked, but very, very relevant in the decision making process, as far as your personal friends financial risk/exposures. You get into politics when dealing with government approved projects, then you need a good attorney on retainer with knowledge of details/problems to 'work over'. Who is to say if you get the green light on a competitive bided retrofit, the project terms get changed before you get into the 1st year?

Perhaps jtr and his friend are more aware and informed about all of the dynamics involved, but I hope they have the big picture in mind.

That being said, if all of the 'other' issues are accounted for, what about those durability, operational condition parameters; seems we don't know enough about how the either LED's or fluoros are effected by what range of conditions?

It almost seems as though jtr is talking about his friends business would be fabricating 10,000 of these retrofits? Over what time period, how quickly could a cheap solution being manufactured. How is the competition, at least the one known competitor, getting the retrofit manufactured? It is one thing to design a working model; another to timely produce at the very lowest cost...which is what seems to be as a strong limiting constraint. Oh yeah, and it seems like jtr has to put this together in record time...happy birthday! /ubbthreads/images/graemlins/wink.gif.

Although it should not be a problem, typical auto alternators, when the (yeah like low cost ballasts, yet another electrical part prone to high failure rates in some instances) the alternators voltage regulator is working properly, you will always see higher than 14 during and shortly after starting an engine, typically around 15+v. Faulty voltage regulators go higher still (just ask any competent auto mechanic).

1st thing I would try to become informed about is what are the failures/weaknesses with the current T5 based fluoro tube systems? Then you have an idea of what the LED replacement is going to need to reproduce for greater reliability. NYC is subject to below 0F temperatures, not including wind chill, almost every winter, with high humidity and heat in the summer. While the lights may not be "on" during the daytime, inside that plastic housing, the temperatures may hit 180F or more. So I take it the plastic housing are 100% UV immune? In other words they never age or turn yellow? Just wondering, since we are being so critical about how white the LED light is, but not considering changes in the color of the plastic shell; or differences in the colors or opacity of the actual rotating advertisements, correct?

I suppose if the retrofit contract would take a year or less, then you are hopefully safe from budgetary constraints and other changes, because at sometime in the future, like EricB pointed out, many other forums of advertising technology maybe in more vogue. I don't pay much attention to advertising on MTA vehicles, but here in LA, many of the buses, in addition to traditional static poster images, also have newer large advertising screens of graphics/images using some type of technology, maybe LED's?

Typically the USA tends to be several years behind when it comes to trickle down advanced technology, which seems to be implemented sooner in Asia or Europe. I wonder if any of the CPF members in Asia have read this thread, what taxis in Seoul are like now? Seems to me earlier this year there was some excitement by djpark and a few others about the 8mm LED's. And there was someone here in the USA? that was selling them for about same prices as 5mm's, IIRC coming out of China/HK???. Do a search on the 8mm LED, maybe those have high enough output/low cost in volume numbers. But that was at the beginning of the year. Cree products are probably not cost effective for your application.

CCFL could be dirt cheap coming from China, but as with most technology involving electronics, you need to have a full appreciation of what is top quality and what is substandard or too unreliable for your own needs.

Not very useful, my opinions, unfortunately /ubbthreads/images/graemlins/frown.gif. See jtr, you are missing out on many things (vague OT reference, another forum) with that irrational fear of yours /ubbthreads/images/graemlins/smile.gif

If your competitors have the proper contacts in Asia, then perhaps their inferior solution will cost 1/10th your ideas, and they simply come up with a better version at a bid that is 1/2 what you could do and still make a profit. Recall Sun Tzu's philosophy on defeating the enemy.

Investigating, having the whole design manufactured in Asia as a turnkey option, would take time and money. But if there really is that 10,000 unit possibility, then it may be worth the investment...taking a trip to HK/China for business purposes wouldn't be the worst thing that could happen /ubbthreads/images/graemlins/smile.gif , unless you and your friend stand to 'lose the shirts off you backs'.

I would seem to me, that with the dropping cost of LCD's, improvements, higher contrast/output's, we should see moving images rather than static advertisements; much as is done now at the checkout counters in the major supermarket chains here in So. Cal. LCD's as TV advertising have come of age; a captive market, those standing in line at the checkout lines...though we know a certain number of highly educated individuals will ignore moving images, and still read the National Enquirer front page 'stories/gossip/scandals' /ubbthreads/images/graemlins/grin.gif.