I’ve seen several articles re “in the near future” stuff about bug sized drones used as spy warfare robots. How small a lens assembly can you have and still have it be useful? Modern smart camera & lens assemblies are probably close to state of the art and give remarkable performance, but if you scale down to the size of a beetle it will have to be much smaller still.
How small can a lens get and still have useful real world resolving power for spy camera type scenarios?
There are two effects you have to worry about. One is the Raleigh limit: As your aperture size approaches the wavelength of light you’re using, you lose resolution. Light has a wavelength of about half a micron, so with a one-millimeter lens you could get resolution of about 1/2000 radian, or about 1/40 of a degree. That’s not ideal, but certainly still usable.
Second, the smaller your lens, the shorter its focal length will have to be, which will in turn result in a smaller image, which means that you need a smaller receptor. But I don’t think we’re close to any fundamental limits on receptor size, yet, and we’re making them smaller all the time.
With cells on CCDs/CMOSes, smaller isn’t better. The smaller the cell, the less light collecting ability it has and the more noisy it will be.
It all depends on what we mean by “useful real world resolving power.” As Chronos said, a 1mm aperture lens will have a theoretical limit of 1/40 degree. That’s roughly equivalent to FHD (1080p) resolution for a 28mm-equivalent view angle (typical of smartphone cameras).
Pixel size on modern CMOS sensors are approaching 1 micron. At that pixel size, an FHD sensor is about 1.9x1.0mm size, and needs to sit about 2.5mm away from the lens.
If you can settle for broadcast TV resolution, you can shrink down the aperture to 1/3 mm or so, and the sensor & distance will be proportionally smaller.
I think the limit will be the same as the resolution limit for microscopes - any smaller, and the wave nature of light will limit the resolution. It says the practical limit is about 200nm. So we have a factor of 4 or 5 improvement in sensor size before we hit that limit.
I’d think that a bumblebee could handle a decent sized lens. Or maybe a swarm of bees with small lens could be ganged together to form a bigger picture. Drone birds or bats seem more likely at first. And who says they need to be precise? A fly has a enough information from it’s eyes to harass a target.
I say again, smaller cells in image sensors is not an improvement.
This thread is about how much cameras can be miniaturized. In that context, smaller pixels are an improvement. Also, the disadvantages of smaller sensors can be partly offset by improvements in the sensor & readout electronics (i.e. lower noise) and optics (faster lens).
If all you care about is image quality, then of course you want the biggest sensor you can afford & carry. The sensor array for the Kepler satellite is about the size of a dinner plate. But in most cases, the choice of sensor size is a compromise between size and image quality. And with improvements in technology, the camera can be made smaller and still provide the same (or better) quality.
The thread is about how small they can be and still have useful resolution. A good lens in front of a garbage sensor will give a garbage image. And “faster lens” means “bigger lens.” So unless you are only interested in still lifes in bright sunlight, you do not want a tiny lens and tiny sensor. Tiny lens and tiny sensor equals blurry motion and noisy photos.
The camera lens on the front of my iPhone 6 is pretty damn small. If that tech is available for general commercial use, I would expect that high tech stuff that the military would pay for, is likely much better.
So what? The OP asked about bug-sized spy robots. So “useful enough” resolution in this case means sufficient for navigation (which can be very low resolution indeed), plus some capability for higher-resolution surveillance photos.
A robo-bug could conceivably navigate on just a few hundred useful pixels (real bugs do, after all), and then once it reached the target take a long exposure at higher resolution. Wouldn’t work for moving targets, but still useful for a lot of things.
Sensors are close to a fundamental limit. You have a limit based upon the noise at the sensor. This noise includes the dynamic range of the number of photons hitting the sensor. Currently the conversion efficiency of a sensor is very hard to improve upon - if you are using a backlit sensor you are seeing numbers like 30%. You can’t improve this with electronics - that just amplifies the noise equally with the signal. Indeed all of the electronics is just adding even more noise.
There is a fundamental link between spatial resolution and noise. Noise acts for many intents in a manner similar to grain in film. As the noise goes up the ability of the system to resolve spatial features goes down. This is part of the reason many small sensor cameras can take fantastic pictures in broad daylight, but produce vastly inferior results when the light levels drop.
In the end it comes down to Shannon. The information bandwidth the camera system is capable of is intrinsically limited. It is limited by the lens - its diameter determines the effective bandwidth (via the Rayleigh limit) and the brightness of the scene along with the f-number determines the signal to noise. You can’t get past this.
“Useful resolution” depends on the application. There will be many applications where a 200nm pixels will be useful.
No, faster lens means a smaller ratio between focal length and aperture. If you shrink the pixel size and focal length proportionally (i.e. keeping the same field-of-view), and keep the same aperture, you end up with a more compact system that collects just as much light. Image quality will not be compromised. The only caveat is that this faster lens (same aperture, shorter focal length = faster F ratio) is more difficult to design and build. And have shallower depth of field.
Pinhole camera suffers from the same limitations as a lens, but adds one more, the resolution is limited to the ratio of the focal length and the hole diameter. This means very large pin-hole cameras work, and you can take some quite remarkable pictures with them. But as the size drops things go bad. Pin-hole cameras start to fail when the hole-size starts to create noticeable diffraction patterns in the image. This is an issue because although the camera shrank, the wavelength of light didn’t. The lovely advantage of a pin-hole is that everything is in focus. The down side is that that focus is equally blurry.
A 1mm pin hole over a 3mm sensor would have a resolution of roughly 9 pixels. 0.1mm hole, and 900 pixels, but already starting to show diffraction related degradation. Long way from the 2 million pixels of HDTV.
Right, but it’s one less thing to have to align on your bug-sized bug. Also saves on weight.
If alignment is an issue, one could design a cemented lens assembly that can be directly bonded to the sensor. It would be more rugged than a pinhole.
About the only advantage of a pinhole is that it doesn’t need any focus adjustment. But that’s just a way of saying it’s equally blurry at all distances.
I think the camera on a cell phone could be easily fit into something the size of a hornet or even a mid-size spider.
A really interesting approach would be to develop a transmitter and electrode that could be strapped onto a live bug and intercept its optic nerve/brain (or whatever processor).
I once saw an ancient electron microscope with a tiny wire into the head of a live grasshopper (the San Francisco Exploratorium is a fantastic place).
Now THAT is a project worthy of DARPA.
They don’t see very well.
Well, you still need room for a motor, a battery pack, a transceiver, a micro processor and flight controls. plus, maybe, a 2nd visual system for navigation.