Last Wednesday I attended the seminar on “Detection Systems for Biological threats” in London, organised by the Sensors & Instrumentation KTN. There was a good range of speakers and attendees from industry, academia and government, and some interesting and thought-provoking talks.

Dr. Ian Lawston - chief scientist at the Dstl detection department - discussed the performance requirements for detection technologies. Apart from the obvious needs for sensitive detection of as many threats as possible, he emphasised that currently improvements are most needed not in the detection technology itself, but in the area of sample collection and processing. Real-life samples are inhomogenous, messy and full of contaminants - and in many cases sample prep will take far more time and require far greater skill then doing the actual detection.

Supt. Steve Doel of the Police National CRBN Centre stated that at present the police does not have any approved means of detecting biological threats - so there is a very strong need for robust, cost-effective solutions which can give the level of confidence required for typical police scenarios (e.g. evacuation or quarantine of airports).

There were a number of presentation on various detection technologies. What struck me is that these tend to be stand-alone systems - the raw sensor data is interpreted in an integrated device and processed into a simple recommendation without taking other factors into account. It could give significant benefits if instead all of the different detection modalities would feed their raw sensing data into one intelligent processing unit that would base it’s output on the combination of sensing results. This would require a common language to express the raw sensing data, and the need to build up combination profiles for certain threats. Any comments?

For the upcoming Lent term I’ll once again be mentoring a group of students doing an i-Teams project. This term we’ll be looking at which applications for carbon nanotubes would benefit most from a production process that slashes the cost of producing nanotubes by an order of magnitude - a production process developed by a group lead by Prof. Derek Fray of the Department of Materials Science at the University of Cambridge

Unlike most current production technologies the new process creates carbon nanotubes from the solid phase - from graphite. This means that production is far more efficient and production capacity is greatly increased.

Prof. Fray’s team has also shown that they can modify the production process to produce carbon nanotubes that are filled with metals such as tin. The use of these filled nanotubes in Li-ion batteries has shown to greatly increase their capacity per unit weight - a very important consideration for future generations of hybrid or all-electric vehicles.

In the course of doing client- or internal projects we sometimes come up with ideas for spin-off products. Some of these we’ll bring to market ourselves under the mju-tec brand.

The first mju-tec product will be the PAM401 4-channel pico-ammeter. PAM401 is a USB device so a standard PC is used to control the instrument and view the results. The most sensitive range will measure currents up to 15pA with 20-bit resolution - in this case measurement bandwidth is limited to 1 sample per second. Maximum sampling rate is 3kHz for full-scale ranges of 50nA and upward to 7000nA.

PAM401 pico-ammeter

We’re still fleshing out the application software and are open to suggestions from potential clients - let us know what features would be useful.

My client Inscentinel and a prototype explosives detector I designed and built for them (see case study) were the subject of a page 3 article in the Daily Telegraph by Richard Savill - the full text follows:

Bees, latest weapon in the war on terrorism

Honeybees trained to sniff out explosives could soon be used at airports in the fight against terrorism.

Researchers have trained bees to extend their proboscis when smelling a particular explosive and have also developed a “sniffer box” to indicate when the bees show signs of detecting explosives. A spokesman for Inscentinel, of Harpenden, Herts, said teams of sniffer bees could one day be part of the screening process at airports and other venues, including museums and major sporting events.

The bees are know to use their sense of smell in the wild when they are gathering nectar to make honey and extending their tongue or proboscis indicates that they have found their target. Inscentinel thought this trait could be harnessed to help the fight against terrorism. Before being placed inside the detector, the bees are conditioned by giving them a reward of sugared water when exposed to the smell of explosives.

The prototype “sniffer box” model holds 36 bees in small containers. Air is sucked by a fan into the box through plastic tubes and passes over the bees. If explosives are in the air, each trained bee will stick out its proboscis. An optical system is embedded inside the container so that whenever the bee extends its tongue it breaks a beam of light, which then triggers a signal through the computer.

Unlike sniffer dogs which require months of training, it takes only a few hours to train the bees. “The advantage of bees over other animals is that they are really sensitive, cheap and are everywhere in the world”, said Mathilde Briens, the head of research and development at Inscentinel. “The training all revolves around response and reward, a classical Pavlovian conditioning of the honeybees. We expose the bees to the odour, say the smell of TNT explosive, for a few seconds and simultaneously give the bees a reward of sugar syrup.”

It is not just explosives the insects can sniff out. In many cases they can out-detect even the most sophisticated of electronic sensors, picking up tiny concentrations of substances from drugs to food products, and even dry rot.

The machine is not yet commercially available, but has been partially funded by the Home Office OSCT (Office for Security and Counter Terrorism). The company said it had carried out successful tests with the Government. Inscentinel, a small company of three staff and two directors, said the bees could be trained to sniff out anything from home-made fertiliser bombs, to demolition dynamite and C-4 plastic explosives. The company said the device could also be used for medical research, detecting chemicals in a patient’s breath, urine or blood.

The story in pictures can be found here.

Another interesting read was Critical Mass by Philip Ball, in which he explores how the complex behaviour of crowds - people, grains of sand or molecules - can be governed by each participating “agent” adhering to very simple rules. He uses insights from the field of statistical physics to model situations where human crowds interact, such as markets, audiences, electorates etc.

His point is that in order to explain the behaviour of crowds there is no need to look at the individual agents in too much detail. He acknowledges that his models are extreme simplifications of reality, but show that even with this gross simplification the rules of the agents’ interaction dictate the macroscopic behaviour of the group.

I’ve just read “Small things considered” by Henry Petroski. The book is subtitled “why there is no perfect design”, and explains how product design is shaped by constraints that are often conflicting and change over time. He uses everyday, seemingly trivial objects as examples - like the paperclip. Although the idea of the paperclip itself (i.e. using a springy bent wire to hold together sheets of paper) is clever but hardly a quantum leap what made it’s widespread use possible was the design of a machine that could make them cheaply in high volumes. The same drive by production technology has governed innovation in the design of the paper (later plastic) disposable drinking cup.

Using examples ranging from kitchen sinks to car cupholders and staircases to supermarket layout Petroski explores other common constraints and drivers. Especially interesting are constraints dictated by conventions: Ever wondered why on phones the top row of numbers is 1, 2, 3 whereas on calculators it is 7, 8, 9?

A year or so ago I was preparing for a medical imaging project looking at using OmniVision’s OV6920 CMOS imager chip. Not only is it small (at 2.3 x 2.1mm), it also requires very little external components and is cheap enough to be used as a disposable item. We’ve built a few evaluation boards if you’re interested…

In Vision Systems Design of January 2009 I came across an article mentioning an even smaller chip: 500 x 540 microns - at 140 x 140 pixels not a great image but it does mean it can sit at the tip of an endoscope of less then 1mm diameter! It is manufactured by Awaiba Lda of Madeira, Portugal. They specialise in custom CMOS imager chips, but there’s no mention of the tiny chip on their website.

I’d certainly be interested to do a project with that chip.

Marc

For three years the small Chester Conquest mini-mill in my workshop has served me well. It enabled me to quickly make up parts for prototypes or adapt ones that were bought but needed a bit of fiddling with. It is small but most of the parts I need to make are small too. I’ve been looking at retrofitting it for CNC operation for a while now…

But now I’m  shopping for a new mill.  I feel it’s better to start with a more substantial mill - more power, more travel and more rigidity if I’m going to spend time and money to move to CNC. For now I’m keeping the mini-mill as well, as it will take some time to adapt a new mill for CNC use, and I don’t want to be without one. Having a mill available - even a small one - when converting another one to CNC use is also a major advantage.

I’ve basically looked at the following three options:

There are some very nice small CNC mills available from Syil, based on the Sieg X2 and X3 mini-mills or a slightly larger mill. But they’re quite expensive, and still a bit too small for my liking. I have discarded the “adult-size” mill as it just takes up too much space in my workshop for the relatively infrequent use I will make of it.

Then I came across the Chester Century mill, also being sold by Amadeal as the XJ20. Even more attractive is it’s slightly larger cousin, the XJ25. Travel of 550mm x 200mm, and true variable speed from 50rpm up to 3,000rpm from a 1200W motor. A lot of mill for the money, and from the looks of it very suitable for CNC conversion. I had a look at it at Amadeal, and I’ve ordered one for delivery next week…