Monday, November 22, 2010

Bigger pixels or smaller?

Yesterday we reminded ourselves that bigger pixels can gather more light, so does that mean we should buy the camera with the biggest pixels? Well I’ll tell you up front that I can’t answer for your specific application, but I do have a couple of observations to share.

Start by realizing that the sensor size is limited by the size of the lens that’s going to go on the front of the camera. Most of us use C-mount lenses, but that limits how big the sensor can get. So if you want more resolution (meaning more megapixels,) each pixel has to get smaller. On the latest 5Mp cameras using the Sony ICX625 sensor the pixels are just 3.45µm square. That means they don’t capture much light.

Options? Well go to a camera that takes a bigger lens, something like an F-mount. You will of course pay a lot more money, which could perhaps be spent on additional light. It all depends on your application.

A second point to ponder is resolution, in terms of line pairs per millimeter. Smaller pixels will give you better resolution, which can be important if you want to do highly repeatable gauging. The downside though is that you can bump up against the diffraction limit of the lens. A lot of folks don’t seem to have grasped that it’s no good splurging on a 5Mp camera if the cheapo lens on the front can’t deliver better than 10 line pairs per millimeter resolution.

So big pixels mean greater light sensitivity but low resolution. Small pixel scan give superior resolution, if you pay for a quality lens, but you’ll need lots of light.

The choice is yours.


Vladimír Držík said...


I'm not sure that the statement "smaller pixel=less light" is exactly true. In absolute terms, definitely. But think about this situation. You have two cameras with the same chip resolution, but different pixel sizes. You are observing the same FOV. Naturally, you must use different focal lengths of the objective lenses. The result is that in both cameras, the lens projects the same amount of light from the FOV to the area of the chip. In other words, in the camera with the smaller chip, the light is more concentrated to the smaller chip area. As a result, each single pixel gets the same amount of light, in both cameras. This is of course true only if their resolutions are the same.

The other issue is the fill factor, which is really usually lower for smaller pixels. But that's a different technical issue.

Spencer Luster said...

Hi Vladimir,

Actually, one must not forget to take into account f-number and object distance in this discussion.

Let's say you start with a 50mm fl. lens and you are imaging your object at 1:1 onto a detector that is exactly the size of your object. Object distance will be 100mm in this case.

If the lens is set at F/8, the aperture size will be 50mm/8 = 6.25mm diameter. A point at the center of the object will send a cone of light rays to the lens, with a half angle of about 1.79 degrees. [ atan((6.25mm/2)/100mm ) ]

On the other hand, if you wanted to image the same object onto a half-sized detector (so mag. = 0.5X now), the same lens would put the object at a distance of 150mm. For the same f-number, the half-angle of the cone would be 1.19 degrees -- less light would be collected.

If you wanted to keep the same object distance, then you would use a lens with a focal length of 33.3mm. In this case the object distance is held at 100mm, but the 33.3mm fl. lens *at F/8*, would have an aperture size of only 33.3mm/8 = 4.16mm diameter. Again, the lens would collect a smaller cone of light to be projected onto the detector.

The way to have the same light collection for either case is to use a smaller f-number.

Spencer Luster