The big lie about LED lighting
I’ve had it with LED lamps. The world has been told that LEDs are the future, in part because they are economically the right form of long-term lighting, and there are environmental benefits as a great aside. Well, maybe the environmental argument is true, but the economical one is not.
My wife has converted a substantial amount of our home lighting, as well as our holiday decoration lighting, to LED bulbs. Despite all this investment, I have yet to experience the primary benefit of long life. This made me sit down recently and ask myself why.
As it turns out, the answer is quite simple. The lifetime is not a function of the LED, but rather the total circuit solution.
Figure 1 shows the schematic for an incandescent light bulb. As I once read in a college textbook, the analysis of this circuit is left to the reader.
Contrast the Figure 1 schematic with the Figure 2 schematic that shows an offline LED lamp schematic minus the LEDs. As I’ll show soon, there is no need to analyze the Figure 2 circuit operation.
To estimate the failure probability of an electrical circuit, one can apply the product rule for probabilities. Essentially, the rule is that the overall probability of failure is a product of the individual, uncorrelated failures.
Notice that the Figure 1 circuit has one circuit element. Let’s assume the probability of failure for a circuit element is 100 ppm. After all, we’re discussing a 50-cent light bulb, not a $30,000 car. Applying the product rule, the ppm failure rate for Figure 1 is 0.9999 or 1 out of 10,000 light bulbs.
Next we look at Figure 2. There we have approximately 60 circuit elements. Using the same product rule and 100-ppm failure probability, the failure rate of Figure 2 is 1 out of 167 lamps. Now this is a big deal, but not yet a catastrophe.
After a quick look around my 2,200-sq.-ft. house, I counted about 50 light bulbs. Once again, applying the product rule, the incandescent lamp failure rate inside my house should be about 1 out of 200. And the LED lamp failure rate should be about 1 out of 4! A 25% failure rate is ridiculous for a product that costs $10 or more and does nothing more than a product that costs 50 cents.
Perhaps a wiser engineer more skilled than I in the area of reliability would take issue with my 100-ppm failure rate. But doesn’t it make sense that an offline power controller with nearly 100 components will have a higher failure rate than a system with 1 component? But no one ever talks about that. We are always told to compare the tungsten lifetime of 1,000 hours against the LED lifetime that is claimed equivalent to the lifetime of the family pet, or something close to that.
According to my wife, who is not an engineer, she purchased nine LED lamps for the kitchen and back porch over a period of one year, from three different manufacturers, and paid on average about $45 each. Out of the nine, and after two years, one by one they all started to fail. As of this blog, seven of the nine are dead.
I did a further reliability analysis on Amazon.com, looking at the ratings for LED lamps. Out of 400 reviews, 25 people gave a popular $10 LED lamp a rating of one. I started reading through the one-star reviews and virtually everyone who scored the lamp as a one claimed their score was because the lamps failed very quickly.
Even most of the 18 people who scored the lamp a two did so because of failure. I wonder how many of the five-star ratings would still be five stars after two years.
I’ve been designing LED circuits for years now. My first circuit was for a stadium sized LED TV and more recently battery powered lamps and backlighting for TVs and laptop computers. It would be disingenuous of me to claim that all LED lighting is useless. There are many great applications for LED lighting. But replacing a one-element off line circuit is not one of them.
Scott Elder, principal engineer with Linear Technology, is a 28-year veteran of analog IC design and the named inventors on 16 patents, all in the field of precision analog signal processing, power management, RF CMOS and LED drivers.
This article first appeared on EE Times’ Planet Analog website.
Related links and articles: