A few years ago, I tried to start a small hardware startup called opaq. The idea was to build a device that could detect black ice in front of vehicles using spectroscopy.
The basic principle was simple: water, ice, and dry pavement reflect infrared light differently. By shining several infrared wavelengths onto the road and comparing the reflected intensities, the system should be able to tell whether the surface is dry, wet, or covered by ice. In principle, it should even be possible to estimate the thickness of the water or ice layer.
The idea was to illuminate the road in front of a vehicle with several near-infrared laser wavelengths and use the relative intensities of the reflected light to detect black ice. Conveniently, the fiber-optic telecommunications industry operates in the wavelength range needed for this, making suitable laser diodes not to hard to find. I got mine from ebay, at 860 nm, 1310 nm, and 1550 nm, plus a rather exotic 1410 nm diode in a C-mount package.
Along the way, I also got my hands on an old vidicon-based SWIR camera that can image wavelengths up to about 1800 nm. Apart from these old camera tubes, the only practical imaging technology that reaches this spectral range (to my knowledge) is based on InGaAs sensors, which are prohibitively expensive.
The camera made for a fun demonstration of the underlying physics. I placed a small puddle of water on a thermoelectric cooler (TEC), illuminated it with the 1410 nm laser diode, and froze it. As the water turned into ice, it became almost transparent at this wavelength.
Below are images of the water before and after freezing, and a video of the experiment can be found here.
With the basic principle confirmed and the laser diodes in hand, I started building the first prototype.
The first prototype was large and rough, with home-etched PCBs and a housing that I machined by hand. It already contained all the essential ingredients, though. The four laser diodes illuminated the road in front of the vehicle, while a single lens collected the reflected light onto one photodiode. The photodiode signal was amplified by a transimpedance amplifier and digitized by a Red Pitaya, which also communicated with a laptop over Ethernet. Each laser was modulated at a different frequency and I implemented four digital lock-in amplifiers on the Red Pitaya to separate the reflected signals from the individual wavelengths. Comparing their intensities allowed the system to distinguish dry pavement, liquid water, and ice, and even estimate the thickness of the water or ice layer.
The second version was much smaller and more serious: proper PCBs from JLCPCB, soldered with hot air, more sophisticated laser drivers, and a transimpedance amplifier with automatic gain control. The automatic gain control is important because the amount of reflected light can vary enormously with distance and pavement. The housing for the second prototype was machined on my CNC mill, which was the reason I bought the CNC machine in the first place.
To make the second prototype more compact, I designed the whole package in CAD first. A nice side effect was that I could also turn the model into a 3D animation of the prototype. That animation ended up taking way too long to make, but it looks cool.
Designing the photodiode preamplifier was also when I learned that the root of all evil in photodiode circuits is the capacitance of the photodiode, just as warned in The Art of Electronics.
In the end, I spent a lot of time and money on this project. I bought scientific equipment, designed electronics, machined parts, wrote signal-processing code, and learned an enormous amount.
But I did not pursue it further. The main reasons were lack of funding, and honestly also my lack of experience with the business and organizational side of turning a technical prototype into a company. Another major reason was that I could not get a patent on the basic idea since it had already been described in the 1990s in several patents, i. e. here and here.
It never became a company, but it was probably the project that taught me the most about taking an idea all the way from physics to electronics, mechanics, software, and an actual working prototype.
The old project website is quite fun and is now archived here.