Canon has just developed a new full-frame sensor designed specifically for capturing videos in low light. Why the hype? The sensor is 24x26mm but only has a pixel array of 1920x1080 (just over 2Mp), the standard for HD video. Implication: huge sensor + low pixel count = huge pixels, which are great for recording in low light.
So, why is this?
Megapixels are a good thing, but only to a point. While camera manufacturers often emphasize pixel counts as a selling point, a high pixel count does not always equate to a good image. More than anything else, it is the size of the pixel, not the size of the pixel count, that makes good images.
The general rule is that bigger pixels perform better. As seen in the picture above, if each of the sensor formats were to have, for example, 16 megapixels on the sensor, it it easy to realize that the 1/2.5” point and shoot sensor will have pixels that are much smaller than those on the full frame SLR sensor. This relationship between sensor size and pixels contained (number of pixels divided by sensor area) is referred to as pixel density, which basically states how many pixels are crammed into a given area. In 2012, many digital SLRs have pixel densities around 5MP per square centimeter while many point and shoots will have a density of around 50MP per sq. cm. This, more than anything else, shows why digital SLRs perform so much better than point and shoots in high ISO/low light situations. Want proof? Go here for an in-depth comparison.
The reason for this greatly differing performance is what is called the signal to noise ratio. Really, it is not a numerical ratio, but a relationship between the amount of noise and the amount of signal (light) the pixel captures. All electronic sensors, including camera sensors, inherently produce noise. The problem with the small pixel is the amount of signal (light) they are able to capture. Small pixels of point and shoots have a small surface area and are able to capture far less signal than a large pixel of a digital SLR. Since the large pixel can capture a lot more signal than a small one, that signal can effectively drown out the noise, which is something small pixels cannot do nearly as well because of their small size.
To quantify, let's say that both a point and shoot's and SLR's pixels have an inherent noise level of 5 units. However, the point and shoot captures only 20 units of light while the SLR captures 100. In the end, the signal/noise relationship for the SLR is 100/5 while the point and shoot is 20/5. The 100 units of light will be far better than 20 units for overcoming 5 units of noise.
A further complicating factor is the camera's sensitivity (ISO) setting. Higher ISOs are more sensitive than lower ones. While the higher ISO will boost the camera's sensitivity to light, it will also amplify the noise that is inherent in the sensor. The end result is that at high ISO, the same amount of light will be captured (although more quickly) but the amount of background noise will increase. For example, say a pixel captures 100 units of light with 5 units of noise at ISO 100. That sensor will capture the same 100 units of light (although more quickly) but 50 units of noise at ISO 3200. The closer the amounts of signal and noise, the less the noise can be drowned out by the signal.
Back to Canon's new sensor.
By going for such few pixels, Canon is bucking the trend of squeezing more and more pixels into a given space in order to boost high sensitivity and low-light performance. As for what Canon will use it for, current bets are on astronomical and medical research, not consumer-grade products like dSLRs.
For more info:
Digital Photography Review
Hit the 'subscribe' button for automatic email updates when I write something new!