Sunday, July 13, 2008

Power Electronics

For this entry, I'm going to take a departure from the usual microelectronics and HPC topics and focus on power electronics topics on which I have been keeping .. current (forgive me).


According to Wikipedia, Lithium Ion batteries can store 160 Wh/kg or 576 kJ/kg with maximum power output of 1800 W/kg. The wiki page on ultracapacitors indicates 120 Wh/kg as a comparison point for Lithium Ion batteries, but Li-ion is obviously somewhere in the 100-200 Wh/kg range...

Solid-state Ultracapacitors support substantially higher power than Li batteries (6 kW/kg from wiki), though the energy per weight efficiency is often an order of magnitude lower. This means they can be charged and discharged much more rapidly than a battery, but contain much less energy. Since there are no chemical processes, the number of charge cycles for the system is many times greater than a Li battery.

MIT Professor Joel Schindall uses aligned nanotubes to increase the surface area for charge storage in an ultracap and aims for 30-60 Wh/kg. Read his article from the IEEE spectrum.

EEStor makes ultracapacitors in an ultrasecretive fashion, but the claimed results beat Li batteries in terms of weight to energy stored by a factor of 2 (200-300 Wh/kg or 1 MJ/kg). Zenn Motors is apparently using this in their electric vehicle offering.

A few months back I was reading about Graphene transistors. Capacitors using Graphene Monolayers seems like a good idea. An Indian team reports 31.9 Wh/kg.

Solar Power

The next several years will be the early adoption period for solar energy. Clever mayors will propel political careers on their ability to successfully deploy solar energy in their towns.

Making Solar-Tiles-On-Every-Roof happen is like making One-Laptop-Per-Child happen: the problem isn't just finding the cheapest solar tile you can make so much as having a scalable deployment plan (if you spend too much time making really cheap laptops, you forget that the real problem is the per-child part). A scalable solar plan should put grid-tied panels on roofs while appreciating accelerating returns from volume and adapting to a variety of local financial incentives.

Israeli Bank Hapoalim offers a loan program and Israel Electric offers grid-tie incentives to encourage the adoption of Solar technology.

Without established best practices in these loan and incentive programs, there's a lot of difficult ROI math to convince people to pay for a panel. It seems like the best solution to get solar electricity to market is to provide free solar panels and grid tie installation in exchange for a mortgage redeemable in sub-market-price electricity. Such bonds can tie together with "weather risk bonds" to form a new type financial products allowing you to invest solar energy by essentially renting someone's roof to profit off the power generation (roof-renting actually makes sense in more than one way in a foreclosure market). States and cities could provide tax incentives on this type of investment vehicle to make it more attractive against other financial products and to encourage people to invest in regional solar power.

An interesting HPC side note related to weather bonds: I wonder if our superb supercomputer models can be used to predict total solar panel output on a given day...


A superconducting electric car has been built in Japan. The concept seems sound: higher current density means you can use a lighter conductor for the same magnetic field, not to mention the efficiency boost from 0 electrical resistance. There's obviously the added liquid nitrogen tank too, which is also useful when you're trying to get away from a T1000. They claim 10% more range. I wonder if their suspension systems uses the Meissner effect...

It seems plausible that the superconducting electric motor could also be used to efficiently compress liquid nitrogen while the thing is recharging too (I have the same inclination to insist on FPGA acceleration for FPGA accelerator compilation).

Nuclear Fusion

Google Tech Talk about Bussard's Polywell Fusion Reactor. The majority of Fusion research is in toroidal Tokamak magnetic containment structures. Bussard's comment is that we have had all these problems generating sustainable fusion with Tokamaks, and yet we look up and see thousands of fusion examples, none of which are toroidal. Instead they are all held together using gravity: a radial 1/r^2 law.

To achieve a similar 1/r^2 radial fields, we can create an electron ball in the center of a reactor and use this field to draw fusuable ions close to eachother. This idea, called inertial electrostatic confinement, had been explored by Bussard's colleagues in the development of the fusor. The fusor used electrode cage to create the electric field for the electric confinement, but collisions with the cage resulted in a net energy loss and a burnt up electrode cage.

The insight in the Polywell reactor is that trapping electrons in a magnetic field is a lot easier than trapping fusable ions as in a Tokamak (because an electron doesn't weigh a lot). The Polywell creates an electron "wiffle ball" using a tetrahedra of coils to avoid the heating problem with cage based designs. In his Google Tech Talk video, Bussard describes the engineering challenges and insights discovered during the various design iterations. He died shortly after receiving funding to continue the program, but the work is still continuing.

When you see him present the history of Fusion research and the reason things are the way they are with the massive Tokamaks dominating research, you can tell he just knows that controlling ion momentum in magnetic fields "won't produce fusion power, but it will produce great physics." And so the funding continues.

He knew he was on to something with inertial electrostatic confinement using magnetic fields so he patented the design. Then he improved his design by using circular rings instead of square rings, and then ideally spacing his non-ideal rings. I wonder if they could modulate the current in the rings to produce dynamic stabilization and try to narrow the cusps.

You almost want to build a really large one with superconducting electromagnets and use it to hold a small star in place.

No comments: