(Originally posted on October 6, 2010)
In a previous post, I wondered aloud:
“The implication of Moore’s law, along with implicit corollaries for energy storage technologies (batteries, capacitors, etc) – is there a law yet in Wikipedia for this??”
Turns out, there’s been some buzz about this question recently. As pointed out by Gigaom, Thomas Friedman, in an otherwise excellent piece in the New York Times (September 25, 2010) on the need for the United States to ‘drive’ an electric car program as aggressively as it did its own Moon Shot program in the 1960s, repeated an assertion that there is indeed a kind of Moore’s Law already in effect for batteries: “the cost per mile of the electric car battery will be cut in half every 18 months.” Gigaom correctly pointed out that there is no such “law” currently in effect.
Techies have been held in awe of Moore’s Law and its consistent returns for so long that it’s natural for them (us) to assume that every other hard challenge of science and technology would eventually be tamed and mastered in a similar manner. So far, though, battery technology appears to be immune to this romantic notion. According to Bill Gates, “There are deep physical limits” (also reported by Gigaom) when it comes to batteries.
Perhaps Moore’s Law is really an “assertion”. As originally stated, it referred to circuit density (the “number of transistors”), but some have shown that circuit performance has actually hewed the line as well, also doubling every 18 months. The difference between density and performance is significant, even though as things get more complicated, ‘performance’ may be hard to universally measure. As circuits have shrunk (as densities have increased), clock rates have also increased: more transistors to do more work per clock cycle, leveraged by more available cycles per second.
Why does Moore’s Law apply only here (and not to, say, battery capacity)? Is it because increasing transistor density “just” comes down to figuring out how to print circuit patterns onto silicon using increasingly shorter wavelengths of light, while “simply” managing to deal with the various weird physical effects that come into play when you’re working at nanometer scale, while designing the circuits for testability and robustness, etc? Can any of the physical techniques be applied to battery technology, to increase surface area, etc? I’d better stop here since I’m just guessing about this .
Note that in the Friedman article, the metric was not battery capacity, volume, weight, or energy density but “cost per mile of the electric car battery. That’s surely an important metric to focus on now and not a bad place to start; my guess is that getting the un-subsidized cost of a 100 mile range electric car to drop below, say, $20k, would be an interesting milestone. The Nissan Leaf, an all-battery vehicle with a 100 mile range, carries an MSRP of nearly $33k; after subsidies, Nissan estimates that the take-home price starts at $25k. At some point however, longer range will also become a distinguishing factor, and so battery energy density will become an important metric to track.
But at a higher level, the real metric is about the cost, density, or capacity of “energy storage”… there are other ways, besides batteries, for storing energy, such as: capacitors, flywheels, fuel cells, compressed air (yup), and spit (kidding!).
What does this have to do with ‘Bots? A lot. As energy storage technology (admittedly, mostly in the form of batteries) improves, resulting in smaller batteries that pack more of a punch (in terms of total energy stored and/or in terms of amount of energy that can be delivered in a given time period, or “power”), and/or can store energy for longer, then the set of scenarios you can envision with a self-contained connected device gets that much richer. Couple that trend with CPUs that can do more with less power, and with sophisticated ‘sleep’ modes, and you get even more leverage. Batteries that last 10 years are now commonplace… imagine a self-contained wireless device packing a battery that can power it, for, say 20 or 30 years… you’d start to think differently about the scenarios it would enable. You could, for instance, build them into semi-permanent structures: boot ‘em and ‘forget ‘em’… for security, maintenance, building health, and other kinds of monitoring (such as managing wilderness areas, tracking geologic events, and so on).
Those kinds of scenarios imply interesting requirements for the software that would power those devices, and the systems that would manage and monitor them. Unless you’re a programmer for a Mars rover, accustomed to not being able to reach over and hit the reset button at a moment’s notice, my guess is that achieving a world where devices go for 30 productive years possibly untouched by a human may take some work. It’s an interesting problem.
PS: Not to mention the environmental issues associated with the production of so many batteries and the challenge of recycling them when their useful life is over, or related health issues – have you ever seen a ‘leaking’ dry cell and the damage done to its immediate environment? (Added 6 October 2010:) And then there are the social / policy / privacy issues. More on that later.
PS: Watch any of the videos on this site (this one is my favorite) and ponder the technology advances that have made this form of ‘connected device’ commonplace, and a platform for frenzied experimentation and innovation: small, cheap, lightweight, powerful batteries, sensors (accelerometers, gyros, pressure sensors), cameras, compute devices, motors and associated electronic controllers, servos, GPS modules… all integrated and leveraged by sophisticated software.