[QUOTE=danlecompte, post:14, topic:390149]
It feels like this was designed in the early 90s when LED flashlights were first coming out and it has never been updated since. It’s a shame, with even a little reengineering of the LED unit this could be an amazing flashlight.
[/quote]
As a flashaholic and an electrical engineer, I must disagree. At the sort of currents these coin cells deliver, the difference between modern power LEDs and these 5mm LEDs is a factor of two or perhaps three. This means, with appropriate electronics, you could get twice the light output for the same runtime, or the same output for twice the runtime. And we’re talking replacing three LEDs that may be a dime each with 1 LED that may be a dollar each, so it doesn’t seem like it would add much to the cost.
But there’s two catches – first, the pattern of light the LEDs generate, and second, that modern LEDs gain their efficiency partly by having lower voltage.
For the pattern of light, it’s simple – 5mm LEDs have a built-in lens, focusing the output to perhaps 20 degrees or so (you can order them in various beam widths), so you slap them behind a clear window, and you’ve got a usable flashlight beam. Power LEDs generally have a lambertian or nearly lambertian distribution, which in layman’s terms means the distribute light through an entire hemisphere, with most of it in a 120 degree (or so) cone – spreading the light thinner makes it dimmer, so for most applications you need to add a reflector or focusing lens of some sort to concentrate it into a beam of 10 to 20 degrees. That means adding a component (optic of whatever sort) and usually pushing the LED further into the head to get it at the focal point, which might require a redesign of the housing.
To understand the second catch, know that a big part of what makes a button-cell/5mm LED flashlights so cheap and common is that the internal resistance of the batteries is enough to control the current to the LEDs. It’s not elegant, and you get quirks like the (relatively) high brightness each time you turn it on, fading out as the surface charge is used, as sdc100 noted, but it is dirt cheap.
Now the catch is, when you replace the LEDs with modern, lower voltage LEDs, the greater voltage difference between the LED and the battery stack causes more current to flow, making it more than twice as bright for less runtime. IMO, that’s not at all a good thing – you want one or both parameters to increase, and neither to decrease – and to avoid it you need to either add several more components (requiring a good deal more space, combining with the first catch this definitely results in a housing redesign) to regulate the current, or take the simple solution of using one fewer battery – leaving it about as dim as the original, with runtime a little longer, but it’s only using 3/4 the batteries to do it. While this is a win for efficiency, customers look at “brightness” and “battery life”, not “brightness * battery life / number of batteries”, so it barely looks any better as a product.
Really, in designs like this, the coin cells and LEDs have some nifty synergy, and you can’t really win without upgrading them both, which of course you can’t readily do without compromising the magnetic stickability.
At this point… I should point out there are other battery options, completely unpalatable to the normal customer, that a flashaholic might try to improve one of these with no housing changes. Going from four 357 cells to a 10180 rechargeable Li-ion (one 10180 costs as much as this whole light. Any of the few chargers that A. will charge a 10180, and B. will not destroy it in the process, cost even more.) leaves me 2-3mm in the battery compartment, so I may be able to cram a regulator in there… Swapping in an XP-G or even XM-L LED, with the shallowest optics I can find… I may or may not succeed, but there’s just a chance I’ll turn a $5 light into a slightly better $20 dollar plus hours of time light, which is what a hobby’s all about, yeah?  In for 3!