Open Source Software Conference October 2009

I'm presenting two classes at the Utah Open Source Conference this year. I've had some ties to that group (and other open-source projects) for quite some time now. The cost is quite low, and as far as training goes, there are some great presentations on everything from virtual machines to advanced image editing with GIMP.

If you can make it, I highly recommend it. I'll be talking about securing your data over the internet with OpenVPN and filtering your home or office web traffic with Dansguardian and transparent proxying. There are some big names in the open source community presenting as well, including one guy who literally wrote the book on open-source telephony.

Nice to see more solid data on LED efficacy

Interesting blog post on one of the industry publication websites. They have some nice data on efficacy (power to light conversion efficiency) of several white LEDs used in light fixtures. Remember the efficiency numbers stated don't include power loss in the power conversion circuitry (in fluorescents, this circuitry is called the ballast) nor the losses due to the optical system (lenses, reflectors, etc). It's interesting anyway.

Zigbee, EnOcean, and Energy Harvesting

I won't take a lot of space for this one, as interest in this topic is probably narrower than what I usually discuss. This article about Zigbee vs. EnOcean in energy harvesting sensors for building management and energy savings presents some interesting points. Interestingly enough, illumra makes systems with both technologies, and even uses them together in a hybrid (or bridged) configuration. We use Zigbee for the backhaul, and EnOcean for the energy harvesting controls. As always, a disclaimer - I do work for/with illumra, so I'm not a completely objective third party relating to those products.

Expensive LEDs - why?

One contributing factor to the high cost of LEDs - they have to be kept relatively cool to preserve their long lifetimes. This means the "bulbs" require expensive metal heat sinks and ways to get the heat from the LED into the heat sink. There are lots of great innovations in this area, but it still adds a lot of cost.
A separate but related issue is that the efficacy (luminous efficiency) of the LED drops off when you turn the brightness up. There's a great article in the latest edition of the IEEE Spectrum magazine that starts from a nice high-level overview, then digs down to some very technical information.
What it boils down to is, we can have very bright LED lights, but they'll be less efficient than they could be, or very efficient LED lights, but they will be less bright (or more costly, assuming we just put more LEDs in each bulb/fixture).
Engineering is always about making the right tradeoffs - there's a lot of research going into LEDs right now, and over time we'll see the benefits.

What did you do today?

I recently helped pour some concrete as part of a project with some extended family members, quite a change from the usual comfortable office environment where I do embedded control system development work. It's fun to do something different once in a while, it gives you a different perspective on things.

As I considered the permanence of the concrete we poured, worked, and finished that day, I realized that there is also a lot of permanence in my day-to-day work. While the constant flow of new products, new features, and new customers makes for interesting work, with new challenges, I realized that the firmware and hardware I (and my colleagues) develop (and the company sells) will be running continuously for many years. Will it run as long as the concrete lasts? Probably not in a device sold today -- after all, the microcontrollers that run the code only have a memory life of 100 years or so, and probably won't be around that long, on average. However, the same code I write today will probably be programmed, in one form or another, into devices over the next 20 or more years. Over the lifespan of the controller, it could save many millions of kWh of energy (and of course, lots of money and energy-related emissions). That's something to be proud of, I think.

So what did I do today? A lot more that you might think, just looking at the code.

Efficacy (luminous efficiency) is getting there...

At last, a production LED that approaches the efficacy of good fluorescents (100 lumens per watt).

I saw this announced at the Lightfair tradeshow this week. I expect to post more on other subjects later. Now admittedly, this doesn't include power loss in the power supply (ballast), but we're getting there. I'll be interested in how the light radiates and how much they cost as well, because that will make or break it in the long term.

Biggest Barrier to LED lighting

Okay, there are several barriers, and it could be easily argued that there are more serious ones than this. I believe the biggest problem with LED lights will be shortened life to to overheating in standard fixtures (I believe short CFL life is most often caused by overheating, and I'm not convinced that Energy Star or other standards address this). Most fixtures that have been designed for edison-base incandescent bulbs are not designed to remove the heat from the fixture or bulb. Instead, the fixture is designed to withstand the heat (ever see a label on a light fixture that requires 105 C rated wiring?). Some of the worst offenders, as far as keeping the heat in, are recessed lights that are rated for insulation contact. This means that the fiberglass (or blown-in) insulation can completely cover the light fixture inside the ceiling. We did some testing with Illumra wireless relays in the ceiling box and found that even with a 40 or 60 watt incandescent bulb, the wiring box could reach 80 or 90 degrees C.

Okay, enough background, why is this such a big deal for LED lights? Put simply, like any electronic device, LEDs do not like to be hot. It shortens their life a lot. For the physics behind this (if you care) the Arrhenius equation gives you some idea - doubling the temperature doesn't just cut it's life in half, it cuts it much shorter. You see, most light fixtures have been designed to take light from a (fairly focused) point source, the filament in the bulb, and diffuse it evenly over a space. Putting lots of LEDs at one point concentrates the heat, making it harder to keep the LEDs cool. Spreading out the LEDs is a good solution, but then the fixture isn't designed to take light from many points and distribute it.

I've seen a number of LED lights in a fluorescent tube form factor, but with most (or all?), the LEDs radiate only out one side of the bulb, defeating the reflectors in the fixture, preventing the light from shining where it should. In an office environment, you would probably get very uneven lighting with this solution (until they start mounting the LEDs 360 degrees around the tube).

The real solution is to design the fixture from the start with LEDs in mind. This will require a "ballast" (really a power supply) that provides a constant current to run the LED string, along with either failed LED bypass or failed LED indication (if each LED is removable). I suggest that, if a standard small form plug-in LED module included a small parallel LED with a large value resistor, then when that LED fails (open) the small cheap LED would illuminate to show which module(s) need to be replaced. Otherwise we'll be in the same situation as with a string of holiday string lights, when one bulb is slightly loose, the whole string fails to light. The whole string should be driven with a constant current (for most even illumination). Typical drive currents are 350 mA, and if the correct number are arranged in series, the voltage of the string could be set not far below the voltage provided by a rectified AC line voltage, which will help with efficiency (using a switching constant-current power supply with AC line power factor correction is even better - see my earlier posts about that).

More to the point of overheating, spreading the LEDs around a large surface helps solve the problem. Imagine a standard office ceiling fixture with 4 T8 fluorescent tubes. Remove the tubes, then cover the surface of the internal reflector with LEDs. The heat is spread out, the whole string can be in series, and when one cartridge fails, the small red LED shows you which one to replace. This solves most (if not all) of the issues with LED lights. Old fixtures could even be retrofitted with this solution without requiring complete replacement of the fixture. With a more expensive "ballast," dimming could be supported for daylighting or load shedding. As I've mentioned before, however, phase-cut dimming is not a good idea. Either powerline-based data communication or wireless control (insert shameless plug for Illumra products here-remember I work for them :-) is the only good option for dimming the lights. The extra cost (in the bulb/ballast/fixture) to support phase cut dimming (just so you can use your old "standard" wall box dimmer) is not worth it.

Two unrelated notes:

One, I've been adding new features recently to the firmware in many of the Illumra dimmers and load controllers. From day one, we've always tried to select default operating modes that are the most convenient for the users of the products. At the same time, while they save energy as much as possible, we don't want their operation to be intrusive. For example, if a receiver is set up to turn off the lights when the space has been unoccupied for a while, for a time, after the lights turn off, if occupancy is detected, they turn back on. However, California Title 24 (which includes regulations updated in 2005 for how these devices must operate) doesn't allow automatic turn-on when you enter a room. We're making Title 24 compliance the standard operating mode for our receivers, even though it's potentially more intrusive (meaning you must turn on the lights manually when you first enter the room). It's been a tough decision, but since an automatic-on mode is available, we'll still support modes that require less intervention. Perhaps if you are ever irritated by a light that could turn on automatically, but doesn't, you can remember that you may be saving a little more energy (and money) that way.

Two, while I won't be there personally, if you're in New York next week, drop by our booth at the Lightfair tradeshow. Illumra products will be shown in one part of the EnOcean Alliance booth, as well as integrated into a number of products under various brand names throughout the show.