This is for for a scanning surface potential measurement tool, used in contamination detection in semiconductors: 40 attocoulomb sensitivity in a 4 MHz bandwidth
This one is in cooperation with the Kuno biochemistry group at Notre Dame—A front end for picosecond spectroscopy with a filtered supercontinuum source. The interesting part about this one is that it has to have good cancellation performance down to about 10 nA, about 1000 times less than the original noise canceller or the New Focus Nirvana (which is closely modelled on my original design, except that mine is 40 times faster).
This was a prototype of a transcutaneous (through-the-skin) sensor for blood glucose and blood alcohol.
Optical and optomechanical design, detection and control electronics and software, prototype construction: first spectra were taken 5/16/2013, and technology transfer to a contract engineering firm was essentially completed 6/14/2013.
Most notable was the schedule requirement: from a standing start, in less than 6 weeks' work, I did a complete photon budget, designed and built all of the optics and electronics, wrote all the software, integrated and shipped the system. It worked great.
Because of the schedule, the prototype was built mostly out of stuff I had in my drawer. That meant that all the coatings were mistuned, which cost a lot of light, but it nevertheless works very well. There are a few amusing features.
For instance, the grating is mounted on a servomotor intended for radio-controlled airplanes. While rather unorthodox, this is actually a pretty sweet solution for a lowish-resolution spectrometer—you get a rare-earth magnet motor, titanium gear train, magnetic position encoder, and servo control electronics in a 2-cubic-inch device that costs $139 in quantity one. It has to be designed out of the actual product, because it isn't quite good enough at making very small moves repeatably, and besides, it wouldn't do to show the Food & Drug Administration a device built with toy parts!
If you want to try this, I suggest putting an optical position sensor on the grating, e.g. by putting a barefoot diode laser next to the slit and detecting the specular reflection with a lateral-effect photodiode or a webcam sensor. That way, minor nonrepeatability of the tuning doesn't turn the slopes into absorption error. Human tissue has a huge absorption slope in the SWIR—about 2 AU in 100 nanometres—so this is a real issue.
A bit of a departure from our usual fare: a low cost, high speed sensor for detecting blood spots in eggs using spectral differencing. I'm going to be doing the firmware as well as the optics and electronics, and this will be the first actual client work for our newly qualified PCB designer, Magdalen.
Competing devices use xenon flashtubes and very expensive photodiodes, but still need a lot of calibration and tweaking. That makes it a good candidate for our signature technique: A really careful photon budget followed by a design that actually reaches the theoretical optimum performance.
Egg grading is a good example of a measurement dominated by nuisance data: normal variations between samples that have to be distinguished from the thing you're looking for. Big eggs, small eggs, white eggs, brown eggs, thick and thin shells, double yolks, mottling, you name it.
High speed egg-grading machines can inspect and pack as many as 250,000 eggs per hour; running two shifts a day, that's enough to keep up with about 4 million hens!
Update: Design and prototyping completed, 12/2014; productizing underway at a contract engineering firm in the Czech Republic.
Latest: Work began in early October; progress to date has been encouraging but (due to resource constraints) not quite as fast as we'd like.
Building on my In Situ Coherent Lidar (ISICL) particle detection technology, this new and more advanced ISICL will be capable of detecting individual 0.18 μm metal particles moving at up to 3 km/s (Mach 9) anywhere in the vacuum chamber and mapping them in 8 dimensions [position (x, y, z), velocity (vx, vy, vz), time, and particle size]. The Doppler frequencies are as high as 7 GHz, which is pretty different from the original ISICL's 2 MHz, but somewhat surprisingly the basic design is the same; the only major changes are a somewhat more powerful laser, a longer working distance, and a much higher bandwidth back end.