posted by Bryon Moyer
A while back Micralyne announced a MEMS reference process. There are a few of these running around: attempts to achieve – or at least grasp at – a standard process that can address a wide range of MEMS devices.
Most of Micralyne’s processes are confidential, per their customer relationships, in typical MEMS style. What they did here was to take a “neutral” improved version of what they do well and open it up. They’re not sure that customers will simply line up and use that process in high-volume production outright, but at the very least it’s a conversation starter and a way for them to show their capabilities without divulging secrets.
Micralyne’s strength is primarily optical MEMS: mirrors and comb drives and such. Those feature large in their process, but, in order to be a bit more general, they added some inertial devices, like a 2-axis (but not a 3-axis) accelerometer and a gyro, as well as some biomedical devices.
It’s a two-wafer process (plus handles); cavities are etched into the base wafer and the top wafer; the top wafer is inverted and fusion-bonded to the bottom wafer, after which the top-wafer handle is removed. From the top, release is performed and then metal is laid down. This metal step pertains in particularly to giving mirrors a nice reflective surface.
As a complete aside, in the discussion of their optical capabilities, there was repeated mention of “hitless” functionality with respect to the mirrors. I actually had a hard time finding out what this meant, and a conversation with Micralyne helped clarify. For any of you who are, like me, not steeped in optical, this is a way of changing optical routing in an optical switch without interfering with other channels.
It’s actually a pretty simple concept. Below I show a scenario with various fibers being routed to various other fibers via the gold mirrors. In particular, fiber 3 routes to fiber 1 (moving bottom to top). Let’s say we want to reroute that so that fiber 3 now routes to fiber 6. If we just move the mirror across, then the light stream from fiber three will interfere with all of the other receiving channels as it scans across (which I’ve tried to illustrate on the right, with the stars indicating interference as the beam moves; at the particular moment shown, it’s made it as far as target fiber 4 on its way to 6).
So the hitless idea is that you simple tilt the mirror in the orthogonal direction first so that it’s no longer targeting the receiving fibers. You can then sweep it across to the new target; the light is now moving under the other beams and doesn’t interfere. Once over in the new position, you then bring the beam back up to its normal working position and the connection is made (with no disturbance to the others).
You can find more info in their release and whitepapers.
posted by Bryon Moyer
Today’s note comes from the Department of Not What It Sounds Like. It’s about a company called Paper Battery, which doesn’t make batteries, and, what it does make isn’t made out of paper. The burdens of old names that stick…
What they do make is a supercapacitor. We’ve talked about supercaps before; there’s nothing revolutionary in concept. The story with PBC (much less confusing company name as an acronym… but don’t confuse it with PCB) is about their form factor: thin (0.3 mm) and flat. They can be, for instance, added as a “layer” over a battery to provide a combined battery/supercap “bundle” that has much more stable voltage.
Image courtesy Paper Battery Co.
The idea here is that such a flat component can be spread out in ways that use less space than explicit components that need their own plot of PCB. But the spatial efficiencies come with specific extra benefits, both electrical and thermal. As a layer or “wrap,” these supercaps can provide electromagnetic shielding; they can also act as head spreaders, helping wick away heat from hot spots that they might cover or underlie.
In the future, you could see multiple caps and voltages, paving the way for better DC/DC converters.
They say they have the technology sorted, but they’re optimizing for higher-volume manufacturing. In a few months they’ll make the decision to use either a contract manufacturer or to build a fab. If they roll their own, they say that it would be 12-16 months from now when they went live; they could get a contract guy going 3-4 months faster.
You can find more on their site.
posted by Bryon Moyer
We’ve looked at QuickLogic’s sensor hub solution in quite some detail in the past. It’s programmable logic at its heart, but is sold as a function-specific part (as contrasted with Lattice, who sells a general-purpose low-power part into similar applications). QuickLogic recently announced a wearables offering, which got me wondering how different this was from their prior sensor hub offering.
After all, it’s really kind of the same thing, only for a very specific implementation: gadgets that are intended to be worn. Which are battery-powered and require the utmost in power-miserliness to be successful.
You may recall that QuickLogic’s approach is an engine implemented in their programmable fabric. They’ve then put together both a library of pre-written algorithms and a C-like language that allows implementation of custom algorithms; in both cases, the algorithms run on that engine. So the question here is, did the engine change for the wearable market, or is it just a change in the algorithms?
Image courtesy QuickLogic
I checked in, and they confirmed that the engine has not changed – it’s the same as for the general sensor hub. What they have done is focus the libraries on context and gesture algorithms most applicable to the wearables market.
Sometime back, we looked at how different sensor fusion guys approach the problem of figuring out where your phone is on you. A similar situation exists for wearables in terms both of classifying what the wearer is doing and the gadget’s relationship to the wearer. QuickLogic’s approach supports 6 different states (or contexts): walking, running, cycling, in-vehicle, on-person, and not-on-person.
They’ve also added two wearable-specific gestures for waking the device up either by tapping it or by rotating the wrist.
Critically, they do this with under 250 µW when active.
You can read more in their announcement.