An accelerometer so good that only quantum mechanics limits it

The Ars Technica article oversells the shot-noise limit. What you really want to see in this business is a thermally-limited oscillator; the Brownian motion in the spring driving the mass. For a quantum mechanically limited oscillator, check out work like this (which shares an author with the paper linked by Ars Technica): <a href="http://www.nature.com/nature/journal/v478/n7367/full/nature10461.html" rel="nofollow">http://www.nature.com/nature/journal/v478/n7367/full/nature1...</a><p>A shot noise limit is not an inherent reason for kudos. In particular, this sensor is shot-noise limited at frequencies above a few kHz. In this context, the shot-noise limit may only represent the intrinsic noise of the optical readout, not the intrinsic thermal noise of the oscillator. Their noise figure of 10 ug/rtHz is interesting, but not unprecedented.<p>The Micro-G FG-5X represents the state of the art at low frequency and can do 15 ug/rtHz at sub-Hz frequencies.<p>For a more-fair comparison, in a standard MEMS form factor, the Analog Devices ADXL 103 and 203 do 110 ug/rtHz at 100s of Hz and below and cost &#60;$10 each.<p>It'll be way cool to see what their oscillator will do with improvements. Optical readout has less influence on the detector mass and comparable precision to the best capacitive readout.<p>Link to the paper on the arXiv: <a href="http://arxiv.org/abs/1203.5730" rel="nofollow">http://arxiv.org/abs/1203.5730</a>

FYI, the shot noise limit is one which exists when experimentalists are restricted to "classical" techniques, i.e. they don't use unusual quantum input states or measurements. That's probably a reasonable assumption for anything that could make it into a consumer product, but "quantum-enhanced measurements" can do better. Giovannetti et al.'s article is pretty good for those with the background:<p><a href="http://www.sciencemag.org/content/306/5700/1330" rel="nofollow">http://www.sciencemag.org/content/306/5700/1330</a> <a href="http://arxiv.org/abs/quant-ph/0412078" rel="nofollow">http://arxiv.org/abs/quant-ph/0412078</a>

The researchers are from Cal Tech and Univ. of Rochester (in case anyone was curious, since the article for some reason fails to mention the institution responsible for the research)<p><a href="http://www.nature.com/nphoton/journal/v6/n11/full/nphoton.2012.245.html#/author-information" rel="nofollow">http://www.nature.com/nphoton/journal/v6/n11/full/nphoton.20...</a>

The actual paper[1] states that their prototype has similar performance to the best commercial sensors. However it is much better than any previous optical sensor, and they claim that if they scale the device they can reduce the thermal NEA to 150 ng/rtHz with a 25KHz bandwidth, which would be commercially interesting.<p>1 <a href="http://arxiv.org/abs/1203.5730" rel="nofollow">http://arxiv.org/abs/1203.5730</a>

Their whole opening about the delay in rotating a phone seems way off base; I thought developers added that to the OS intentionally to avoid over-sensitive sensors triggering rotations too frequently. While the sensitivity of the sensor plays a part, this seems like an entirely software implementation that has little to do with physics.

Given its potential use in inertial navigation system of guided missiles, I would expect the state of the art in realizeable accelerometers to be classified. Can anyone shed any light on how this would compare with what is known about "weapon grade" accelerometers, even if the old ones ?

I would like to link to a post I made about a week ago: <a href="http://news.ycombinator.com/item?id=4856642" rel="nofollow">http://news.ycombinator.com/item?id=4856642</a><p>There was a discussion in the comments about whether enough cumulative integrated acceleration errors would prevent some sort of system like this from replacing GPS. (Granted, we were discussing gyroscopes, but I think this still relates).