Burned out LED's

[kf0rt]

Once again, asking for the collective wisdom... :)

I've homebrewed a couple of clocks that use LED's for the hands (gets a WTF? from everyone who sees them). First one was built using red, green and yellow LED's. Bought the LED's from Mouser and it's been running 24/7 for about 4 years now without a hitch. These are the T-1 3/4 standard 5mm round LED's; about as common as they come.

Pics and description here:
Dot Clock

For the second clock, I wanted a "cooler" look for the living room, so I went with white and blue LED's. Once you get away from the three easy colors, LED's get kind of expensive, so I did some bargain shopping to find the white & blue. That clock has been running less than three years and already has several burned out LED's and quite a few that are now dim (both white and blue). It looked perfect when first built.

I didn't think LED's were supposed to burn out. I mean... both clocks have identical driver circuits and the older one has had no problems.

Did I just get a bad batch, or is there something about the odd colors that make these more sensitive to burn out?

I've seen some real "sweet deals" on blue and white LED's from Hong Kong and China on eBay. Tempting, but not if they burn out over time.

[N8YX]

What current (measured) is being put through them?

I have some experiences to relate depending on your reply...

[WØTKX]

High failure rates in the cheap ones. The biggest thing that kill 'em is heat. Wrong current limiting R too, but I trust ya on that one.

[n2ize]

The only types of LED's I have ever had go bad were the super bright bluish-white illuminating types commonly used in desk lamps, flashlights, etc. Sometimes they will go out completely, sometimes they go very dim and sometimes they go intermittent... on sometimes off sometimes. Never had a problem with standard indicator type LED's.

Oh... and laser LED's of the red and infrared variety. Just look at them cross eyed and they will crap out on you.

[KG4CGC]

I'm thinking the newer ones don't handle the full rating as the ones they are designed after but, and I'm only guessing, a 12% increase in resistor value may increase their lifespan. This however, does not take into account temperature.

[kf0rt]

Well, here's the thing (and I'm open to "bad design")...

I haven't measured the current -- would have to set up the scope and "get serious" for that. The display panel is multiplexed as a 12x12 array, so each LED is only powered ON for 1/12th of the time (if powered on at all -- depends on what time it is and where the LED sits in the array).

In my experiments, using any current limiting R caused the display to be too dim, so none were used. One side of the array is driven with ULN2803 line drivers, which reduces the voltage a bit, but the whole thing runs on 5V and with no limiting resistors (yeah, I know). My original PCB design had spaces for current limiting resistors, but those were removed in Rev B.

The "odd math" here is that the LED's closer to the :01 position see a lot more activity than those nearer the :59 position. Yet, the failures appear to be randomly distributed. One would think that if the design limits of the parts were being pushed, the failures would cluster around the 1 minute mark. This is what has me thinking that it's bad parts. Well, that and the fact that the other clock doesn't have this problem. For instance, the "noon position" LED is on all the time (sees the 1/12 duty cycle 100% of the time) and still works fine. The 6 O'clock LED just died this week, and it sees the 1/12 duty cycle 50% of the time (six hours, twice a day).

Wierd.

[N8YX]

I got hold of some "super bright" clear green 3mm LEDs from a Taiwanese outfit earlier in the year. These were used to replace the incandescent bulbs in my TS-940S' sub-display. I installed them on the sub-display board and connected each device to the board's 10v distribution bus with a 430-ohm, 1/4w resistor in series with the anodes. This value limited the forward current to well under the 20ma maximum spec...but I still had two of the friggin' things burn out within a week of installation.

Yes, I'm thinking pi$$-poor fabrication technique and QC here.

[W6EM]

I have used LEDs to replace incandescent dial lamps for years. Never had a problem. If I don't have specs, I assume about 2.5v for Vfd, and size my resistors to limit forward current to 10mA maximum. Not using a current limiting resistor, despite the voltage source, and running near maximum spec current is asking for trouble, IMO.

My FT-7 has two white and one yellow 1mm LEDs. In my TS-940, something similar. The only problem with my 940, if you want to call it a problem, is that the dimming feature no longer works since dropping the current to under 10mA makes no perceivable difference in LED intensity. Most references recommend PWM for dimming applications, but I've got enough noise generators to deal with. I don't need another inside my rigs. Ha.

Try the brightest LEDs you can find, but limit current to 10mA and you shouldn't have a problem.

73.

Lee

[kf0rt]

Ordered some new LED's from Mouser and I plan to build up a new clock display panel sometime over the next month or so. Still going "on the cheap" but at least these will be a known quantity with actual datasheets. And prices have come down quite a bit. Last round, I was paying about 50 cents apiece for both white and blue in the "Far East overstock" market (who knows what they really were). This time, LiteOn white were 29 cents and Kingbright blues were 18 cents (in x100 quantities). 100 of each for about $55 delivered. I figure if I have trouble with one color and not the other, it'll point to manufacturing issues, but I'm betting these will "just work."

Kind of agree on the PWM issues, Lee. The current vs. brightness curve on LED's aren't often linear so PWM gives the best shot at real control, but what a hassle. I suppose one could use a digipot and sort the linearity out in a table, but that (and PWM) sure seems like a lot of work for display lumination. I've actually got a 2 (or 3? I forget) stage dimmer built into the clock firmware where it will skip every other (or 2 out of 3?) mux cycle(s). In playing with it, I decided I like it turned up all the way, so that "feature" has never really been used. Easy to implement when all the LED's are driven by software anyway.

[WØTKX]

You're working too hard... http://led.linear1.org/led.wiz