Seeking advice about design/engineering issues when making an LED flashlight

[bpierce]

I mentioned NiCads only because that is what was in the satellite power supplies that I had been involved with. This was 20 years ago. Interestingly, a satellite that was launched with our power system came back to earth a couple of years ago. I think they got over 15 years of operation from those nicads. Not bad for going through a charge/discharge cycle every 90 minutes! It wasn't a full charge/discharge, but that's still a lot of use.

Supercaps are pretty big, but I used a nice supercap from Bussman-Cooper in a recent bicycle light design. 100F at 2.7v. Not too bad at all. Larger than a AA, but will run a LED at a decent level for 10 minutes using a little boost converter from Zetex (now Diodes Inc.). This gets back to a fundamental question: what is the spec for run time and for output level? Is this light supposed to let you go find a real flashlight that is in the next room, or is it supposed to run at 3 watts for a few hours?

See my reply below to Fireclaw18. Based on a lot of feedback I've gotten, I'm thinking I'd target an output of roughly 250 lumens and a run time of 45 mins to an hour for an initial release; also I like the idea of using an LiFePO4 battery. That amount of output is probably enough to help people get around in an emergency, so if I have room within a reasonable cost to ramp something up it would be the run time. I suspect many of my customers won't have a better flashlight in another room, so better if mine can last as long as possible (but if it's too expensive, they won't buy it in the first place!). I'd love to go tend toward premium specs, and maybe eventually can have a premium line, but that might be a tougher market since the people who would buy a premium Flashlamp probably don't need one . Maybe I should be targeting a much longer run time out of the gate given that luminosity?

 

[DIWdiver]

Personally, I love designing stuff. I don't love having my designs go into production because it means that I have to support the product. This means that I have to figure out why half the parts aren't passing the end-of-line test, or figure out what to do when that unique I.C. is no longer in production and no one makes a replacement, or what to do when you start getting a lot of failed parts back from the customers. I've had some fun research jobs where I was challenged and learning, and eventually it all got thrown away when funding disappeared. No production problems, no reliability issues.. it was great!

I say a HUGE +1 to that!

 

[DIWdiver]

Part of the reason I raised particular concern about the power dissipation is because, in the prototype, some of the internal parts were 3D printed. The team that built it for me warned that the material wouldn't stand up to the heat the way a prod unit would, so I should be careful to leave it on too long. I think they are being straightforward, but seeking reassurance that it probably is only a potential issue with the printable plastic. I feel like I'm already getting some of that reassurance from these responses.

If the heat isn't being managed well enough to protect the plastic, even printed plastic, then it almost certainly isn't being managed well enough to protect the LED. Going to better plastics won't help the LED. This would be a BIG red flag for me, and I would want to know what the LED die temperatures are. IMHO, a reasonable thermal analysis is almost a requirement before even going to prototype, much less any farther, and testing to back up that analysis would be required before going beyond prototype. Expecting to fix a thermal design problem with better materials is a recipe for disaster.

 

[DIWdiver]

If you want 250 lumens OTF, maybe 300 emitter lumens, you need about 2.3W. To run this for 45 minutes, even with moderate driver efficiency, you should be able to get this out of a single AA (14500 size) cell in any of the chemistries. LiCo is undoubtedly the best energy density, NiMH probably the cheapest. I was all ready to tell you that LiFe cells are much more expensive per watt/hour, but a quick fact check says that may not be true any more. I would do some serious research before settling on a battery, but I'm starting to think that LiFe may be the best choice, much to my surprise.

Keep in mind that the cell chemistry is at least a factor in determining the driver you can use. With LiCo, LiPo, LiNi (IMR), you use buck or linear regulators. Linear are definitely cheapest, linear or buck are most efficient depending on design specifics. One or two NiMH cells require a boost regulator. LiFe requires either boost or buck-boost, which is the most complex and least efficient, unless you use two cells which puts you solidly in buck territory. You will need both purchasing and engineering expertise to help you wade through the sea of choices to make the choice that is best for you.

 

[bpierce]

If you want 250 lumens OTF, maybe 300 emitter lumens, you need about 2.3W. To run this for 45 minutes, even with moderate driver efficiency, you should be able to get this out of a single AA (14500 size) cell in any of the chemistries. LiCo is undoubtedly the best energy density, NiMH probably the cheapest. I was all ready to tell you that LiFe cells are much more expensive per watt/hour, but a quick fact check says that may not be true any more. I would do some serious research before settling on a battery, but I'm starting to think that LiFe may be the best choice, much to my surprise.

Keep in mind that the cell chemistry is at least a factor in determining the driver you can use. With LiCo, LiPo, LiNi (IMR), you use buck or linear regulators. Linear are definitely cheapest, linear or buck are most efficient depending on design specifics. One or two NiMH cells require a boost regulator. LiFe requires either boost or buck-boost, which is the most complex and least efficient, unless you use two cells which puts you solidly in buck territory. You will need both purchasing and engineering expertise to help you wade through the sea of choices to make the choice that is best for you.

More great info, and points in a a few good directions for more research. Also thanks for the warning about the heat dissipation above -- I am going to press them on that issue, ensure we do a lot of testing, and also get a 2nd opinion since it's so critical. If you don't mind, I may ping you once the specs are finalized to get your thoughts (probably still a couple months away).

 

[Hoop]

I would take a close look at the new Nichia LED's such as the 219B's. They are as efficient as XP-G2's and operate at a slightly lower voltage while offering 80+ and 90+ CRI variants.

 

[bpierce]

I would take a close look at the new Nichia LED's such as the 219B's. They are as efficient as XP-G2's and operate at a slightly lower voltage while offering 80+ and 90+ CRI variants.

Thanks! I'll check them out.