Traffic Signals: Green LED Failures

Traffic Signal - dead green LEDs - 2012
Traffic Signal – dead green LEDs – 2012

In our (admittedly limited) travels around New York State during the last half decade or so, I’ve seen many (as in, dozens of) traffic signals with this failure:

Apparently the topmost LED string burns out first, leaving the other two (?) strings intact. The earliest picture I have dates back to 2008, so this is a problem of long standing that’s probably wiped out any projected maintenance cost reduction for the entire purchase. The most recent failure I spotted, a few weeks after taking this picture, has a flickering upper string that means it’s not long for this world.

Somewhere up around Albany, I recently saw a green signal with only that string lit up and the other two (?) strings dead, but that’s the sole exception to the pattern.

Of late, NYS DOT has been installing a different green lamp with the LEDs in each string scattered over the entire surface and no diffuser. That means a failed string, of which I’ve already seen several examples in the area, darkens a few spots without being particularly obvious; a less common failure has a few flickering “pixels” that will eventually go dark. While that’s a net win, I wonder why only green lamps have this problem: we very rarely see red or amber lamps with any failed LEDs.

One red LED lamp down the road did fail spectacularly: the whole thing flashed, slowly and somewhat irregularly. Not a flicker, but a flash: long off and short on.

It’s hard to get pictures of failed traffic signals…

While I suppose I should report them, previous attempts to do so have only led to requests for the ID number of the traffic control box, which generally can’t be seen from the traffic lane. I am not stopping at an intersection, getting out, finding the box (perhaps crossing the intersection to get there), finding the ID number, and taking a picture for later reference; you know what happens to people who take pictures of infrastructure. You’d think the signals could phone home on their own, but they’re likely not connected.

13 thoughts on “Traffic Signals: Green LED Failures

  1. I think I saw a “How it’s made” episode on the construction of these type of lights. If I remember correctly, I was a bit shocked to see the led leads simply twisted together, not soldered. I came as no surprise to me when I saw, in lights around town, the same thing you are seeing.

    1. led leads simply twisted together, not soldered

      Well, that would explain nearly everything, wouldn’t it? Ouch…

  2. I do agree that it is weird that the greens appear to be more failure prone.

    I like to think I notice the same kind of infrastructure failures, but I have not ever noticed a dead or dying LED turn signal.

    There is a lot more to driving LEDs for brightness and longevity than is commonly thought. I am not completely convinced that even the LED manufacturers know how to do it.

    1. Most of us know how to manufacture them. One problem is that manufacturing shortcuts reduce the price, and governments are often legally compelled to buy the lowest bid. China decided to convert most of their streetlights to LED’s and bought from a low bid Chinese company who did a shoddy job of getting the LED’s thermally bonded to the heatsinks, so they started dying in droves, leading to cancellation of many other streetlight projects: a market shock LED area lighting is still feeling.

      1. governments are often legally compelled to buy the lowest bid

        Alas, because the previous practice of buying from a higher bidder often wound up putting money in a relative’s / friend’s / gangster’s pocket; it seems we can only choose among bad alternatives.

        a shoddy job of getting the LED’s thermally bonded to the heatsinks

        I’d shocked to hear that… [grin]

  3. In our rural county, we have few lights. Haven’t seen any failures. When I lived in San Jose, they were early adopters of the LED signals (made sense–local manufacturers needing to get a start in a new market niche, plus a ton of signals to maintain) and they rolled out in the ’90s. For years, all you saw were red LEDs, with the old bulbs recycled for the other colors. It was about 5 years (by which time I was up in Oregon) that we started to see green and yellow LED signals (your mileage may vary–36 of 40 miles of our trip to town have no traffic signals…).

    A couple of thoughts: The reds were introduced first, presumably since in big cities a red light is activated 99.999% of the time </sarc>, but also because 2) a failed red signal is Big Trouble while a failed yellow or green is Lesser Trouble and 3) high intensity red LEDs were developed earlier. In short, Red signals were probably designed before the cheapification process got fully underway. It’s also possible that the red signals are made to a higher standard, as per reason 2.

    Early adoption note: in the 90s a freshly converted red signal stuck out intensely. Apparently the original circuit had a peaking pulse to get the red bulb up to intensity quickly. Once the LED came in, the signal showed a flash before calming down to a steady state. After a while, this got corrected. Never saw the behavior in the other colors.

    1. a failed red signal is Big Trouble

      That may account for the fact that repairing the flashing red signal took only two or three weeks, not the forever-and-a-day accorded to the green ones.

      1. One advantage of smaller population centers–a complaint to the Powers has to go through fewer layers of ineptitude. NY seems to be as capable (or is it culpable?) as California in that regard. Oregon DOT is fairly responsive, and our county very responsive. The county commissioners learned a lesson or two on responsiveness after a couple of adverse elections. (All three members got dumped in 2 years.)

        I suppose the better fail rate for yellow could be due to the lower duty cycle.

        1. a couple of adverse elections

          Around here, once you’re in, you’re in. Verily, despite dire legal proceedings and even being dead, you can be re-elected…

  4. One factor is that traffic lights in general all have the Green lamp at the bottom. This subjects the Green lamp to more stress as the whole lamp assembly swings from the opposite end like a pendulum. I have not seen a single Yellow or Red lamp failure in the last 10 Years here in Connecticut where most of the states traffic lights are LED now.

    A second factor is that the most common LED traffic light comes in one of two forms, passive voltage generator types and switching converter types. Virtually all Red and Yellow lamps use many, many LEDs (upwards of 200) to make up for the relatively lower efficiency of the Red and Yellow LEDs and they use a passive voltage circuit. The Green and some Yellow High Efficiency LED lamps use about 140 LEDs and a switching supply. The switcher provides more current to more strings of fewer LEDs.

    Another thing to note is that the LEDs are mounted at the end of their leads rather than being mounted flush to the PCB boards with the leads cut off behind the board. The PCB is usually a single sided PCB (copper on one side). This means the green LEDs have less support to prevent them from vibrating like little maracas. The solder joints on single sided PCBs are realitively weak.

    The only solution is to analyze the failure modes and either use ruggedized LEDs (if the internal LED to lead frame wires are breaking) or change the construction technique to minimize potential damage from vibration (if there are broken solder joints).

    1. the LEDs are mounted at the end of their leads

      That seems odd! Does the epoxy package (or whatever you call it) fit into the lens assembly, so the leads bridge the gap?

      I’d expect better design and construction for something that’s supposed to last forever, but … maybe I’m overoptimistic.

      Thanks for the update!

  5. Duty cycle seems to demand less of green. In the late evening until early morning, green is often not used … replaced by flashing or persistent red or amber. Maybe design or component specs over-compensate.

    1. I just spotted another blown green signal, so the design spec certainly lags reality…

Comments are closed.