How to make flat, thin (<1mm) optics for a very wide field of view camera
Above: Printed pinhole being mounted on a prototype 125mg TinyTam 16×16 sensor. The focal plane is at the top of the chip and is just 0.4mm wide!
Sometimes an invention is so simple you don’t immediately recognize it’s value. This was our experience with our flat printed optics technology. This technique allows the construction of optics that are extremely light and support a wide field of view (approaching 180 degrees). This technique can be applied to practically any image sensor with exposed silicon.
To understand this technique, let us start with Snell’s Law. The picture below shows a ray of light originating in a first medium, let’s say air, and traveling through a slab of a second medium, let’s say glass or plastic. Let n1 and n2 be the respective indices of refraction for the first and second mediums. n1 for air is barely above 1.0. n2 depends on the second medium and may be 1.4 to 1.9 or more depending on the type of plastic or glass used.
As the ray of light crosses the boundary between the two mediums, it will refract. This is basic optics. Snell’s Law states that n1 * sin(theta1) = n2 * sin(theta2). The result is that as the ray enters plastic or glass from air, it’s “angle of incidence” with respect to “normal” will be less inside the plastic or glass than outside. More on Snell’s law can be found in the Wikipedia entry on Snell’s Law. One implication of Snell’s law is that the value theta1 in this example can vary from 0 degrees to (just under) 90 degrees, but theta2 will be constrained to a smaller angle called the critical angle, which is arcsin(n1/n2).
We can apply the Snell’s Law principle to make a wide field of view pinhole camera, the cross section of which is shown below. Essentially a thin piece of clear plastic (or glass) is placed directly onto an image sensor. One side of the plastic makes direct contact with the image sensor. The other side of the plastic has an opaque mask printed on it that covers the whole surface except for a pinhole opening. This forms a classic pinhole camera, except that Snell’s Law reduces the angle of incidence of the light rays entering the plastic to a value less than the critical angle. The result is that an almost 180 degree field of view may be projected onto the image sensor.
Such printed apertures are actually quite easy to manufacture. Basically, you can print them up using the same techniques used to make photoplots, which are often used to construct circuit boards or optical encoders. We’ve had whole arrays of these pinholes printed up on a single sheet of plastic for as little as $20 per sheet. Typically these sheets are about 7 mils thick, and are easily cut with scissors or a utility knife. The picture below shows a mask printed on one such strip, next to a penny and a Tamalpais image sensor chip mounted on the 125mg sensor board.
How to prototype: To mount the optics, first we have to cut it out of the sheet. The picture below shows a pair of pinholes. You’ll notice little “2” and “5” patterns next to the mask- this is how we determine if the mask is right side up! (The top one, with the “5” is right side up with the opaque mask on top.) Then we use a clear optical adhesive to mount the mask onto the chip. We place a drop of this adhesive directly onto the image sensor, and then place the pinhole on top. Typically we use a UV-curable optical adhesive, so we then place the entire assembly under a UV lamp (outdoor sunlight also works) until the adhesive cures. We use Norland optical adhesive, which may be purchased from Edmund Optics. The next picture below shows the printed pinhole attached to the image sensor chip. If you look hard you can see the pinhole (actually a square) just right of the center of the mask. The final step is to apply black paint around the edges of the mask so that only light going through the pinhole will reach the image sensor.
We’ve made many other variations of this basic device- slit openings rather than pinholes can be used if one wants to obtain primarily one dimensional information, for example to measure one dimensional optical flow. Arrays of pinholes can be used to grab multiple images of the same scene. We’ve used resolution enhancement techniques to reconstruct a higher resolution image from these multiple images, effectively making a poor man’s TOMBO camera. One can even construct a light field image from these multiple images, albeit at a small scale.
Aside from robotics, I can envision this technique being used to make all sorts of low profile “cameras” or “embedded eyes”- cameras this thin can be embedded in practically any electronic gadget, artwork, or even clothing! We have already flown our helicopter, indoors, using one and two cameras made with this type of sensor so we know that this technique will work for both outdoor and indoor robotics. This technology is perfect for integrating onto a toy.
This technology is “patent pending” and is the subject of two US Patent applications and two international PCT applications. The first application covers the above material while the second covers techniques for implementing the above within a chip.