MEMS Switches for Low-Power Logic

[AntiProtonBoy]

Interesting article about the potential of using MEMS technology to replace transistors in an IC:

http://spectrum.ieee.org/semiconductors/devices/mems-switches-for-low-power-logic/0

[slateraptor]

I dare to say this will never be mainstream.

[ejeffrey]

I certainly wouldn't expect this to replace transistors in your desktop CPU, but there are a lot of nice applications.  The likely starting place is going to be harsh environments.  MEMS circuits are essentially completely impervious to radiation, and can run at higher temperatures than any semiconductor.  In principle I think they can operate down to lower voltages because there is no intrinsic energy scale like the silicon band gap energy.  This article is about digital logic, but it is also possible to build analog circuits like amplifiers from MEMS switches.

[alm]

I imagine manufacturing something complex like a modern uC with MEMS would require a fairly hefty cash investment. Will the market for radiation-hard parts and extended temperature range parts be large enough to pay for this? It might also cause export issues, since the US is somewhat touchy about exporting rad-hard parts.

[slateraptor]

I imagine manufacturing something complex like a modern uC with MEMS would require a fairly hefty cash investment.

I'm imagining painful ripple propagation delays and micro-mechanical failure. Modest-scaled niche applications, definitely, but I just don't see mainstream logic at all.

[SeanB]

Who would have thought 20 years ago that an automatic gearbox would be controlled by electronics, operating hydraulics that have almost no moving parts other than actuators? That was in the days when an auto box had a massive mechanical assembly that had hundreds of moving parts to do the gear selection. Then came fluidic boxes ( and a more stringent oil viscocity and filtration requirement) and more reliable electronics, and you have the current generation of auto boxes with 6 plus speeds.

this is good for automotive or industrial, where you can have a sealed assembly that has to last for years in harsh environments. Speed and low power tradeoffs yes, but will be more rugged as well.

[AntiProtonBoy]

As the article suggests, not everything has to scream at 1 GHz. Massively parallel architectures could make up for the lack of switching speed. At any rate, I love the concept of having mechanical components on the nano scale. It's very cool.

[slateraptor]

As the article suggests, not everything has to scream at 1 GHz. Massively parallel architectures could make up for the lack of switching speed.

I suspect achieving 32kHz in an architecture with non-trivial dependencies like a modern uC will be challenging enough...1GHz is simply a pipe dream.

Exploiting parallelism seems natural enough, but what's gained in reduced propagation delays is lost in real estate...and these relays aren't exactly small. Furthermore, there's a very, very large subset of problems that can't be solved by 9 women in 1 month, so to speak.

I don't think it's possible to speak of mainstream logic without the flexibility of being general purpose. For niche applications, definitely, but no way mainstream. As alm noted, short of complete ozone collapse and/or nuclear winter, the demand for rad-hard switches with extended operating temperature range just isn't there.