Digital Camera Image Sensors


There seems to be much confusion amongst the public at large about the image sensors used in digital cameras. The figures the makers quote are basically meaningless unless you are "in the know" about what they mean when they say "it has a 1/1.8 inch sensor".
The example of 1/1.8 inch type of measurement relates to a standard used starting in the 1950s about Vidicon tubes used in TV cameras. That example of 1/1.8" works out to 0.5555 inches (1 divided by 1.8 is what it means) and that is the diameter of the original glass Vidicon tube that had an image size similar to the image size in the digital camera they are talking about. Due to the internal workings of the Vidicon tubes the available image area was less than the diameter of the tube being roughly 2/3 of the tube diameter. So in the case of the 1/1.8" tube the image area was 7.176mm x 5.329mm, the diagonal being 8.933mm which is regarded as the "normal" lens focal length to use with that image size.


Image courtesy of Molecular Expressions

It doesn't matter that the digital device is CMOS or CCD and that it does not have any glass tube around it, the makers still insist on using the archaic and misleading glass tube diameter measurement of that old Vidicon tube. Bizarre.
They should be using the diagonal measurement of the true image area, but none will do that as people would then realise how tiny those sensors really are in the majority of cases.


KAF-5101CE 4/3" system 5 megapixel chip image courtesy of Kodak here

This is the chip that is in the Olympus E-1 system announced on 24th June 2003.
Later Kodak chips are in the 8 megapixel E-300 and E-500, and 10 megapixels in the E-400.
They vary a lot in size and form in various cameras, but the basic shape is a flat chip of some type that the lens system focuses onto.

The table below is rounded to one decimal place and hopes to show which camera uses which sensor size.
Knowing the sensor size helps you realize that smaller is worse because as the image sensor size shrinks, so do the individual pixel sites. As more megapixels are packed into a smaller sensor then each individual pixel gets smaller again. This is a real problem as electrical "noise" due to the very normal random movement of electrons in any material that is warmer than absolute zero (-273 degrees on the centigrade scale) is added to the electrical signal generated in each pixel site by the quantity of light hitting that pixel. The bigger the pixel site, the better the real signal and that helps swamp the background noise. Astronomers cool their telescope image sensors with liquid nitrogen to keep the noise lower, but that is slightly impractical in a hand-held camera.
 
 

Image Sensor Type
Used in...
(some examples, not meant to be a comprehensive list)
Width
mm
Height
mm

Diagonal mm
(equals 'normal' lens focal length)

Focal length
Factor
1/4"
 
3.2
2.4
4.0
11.3
1/3.6"
 
4.0
3.0
5.0
9.0
1/3.2"
Canon A100, A200
4.5
3.4
5.7
7.9
1/3"
Nikon Coolpix 100, 300
Casio QV-8000SX
4.8
3.6
6.0
7.5
1/2.7"
Nikon Coolpix 2500, 3500
Sony DSC-P31
Pentax Optio 230, 330GS
Canon A40
Olympus C-730
Minolta Dimage X, Xi
5.3
4.0
6.6
6.8
1/2.5"
Olympus C-740, C-750
Ricoh R3, R4, R5
Canon SD700, SD800
1/2"
Nikon Coolpix 950
6.4
4.8
8.0
5.6
1/1.8"
Nikon Coolpix 995, 4300, 4500, 5400
Canon Powershot G2, G3, G5, S30, S40, S45 , SD900
Kodak DX3900, 4900
Pentax Optio 330RS, 430RS
Olympus C-5050, C-5060, C-8080
Ricoh GR-D, GX8
7.2
5.3
8.9
5.0
1/1.76"
Leica Digilux 1
Panasonic DMC-LC5
?
?
9.0
4.7
2/3"
Nikon Coolpix 5000, 5700
Sony DSC-F717
Minolta 7i, 7Hi
8.8
6.6
11.0
4.1
1"
 
12.8
9.6
16.0
2.8
4/3"
Kodak-Olympus-Fuji-Panasonic-Sigma new digital standard. This system uses the same standard lens mount for any brand camera and lens.
Maybe this is the digital system for the future.
First model is the Olympus E-1 with the 5 megapixel chip as displayed above. Second model is the Olympus E-300 with an 8 megapixel chip also made by Kodak. Now also E-500, E-330 and E-400.
Visit the 4/3 site for more information.
18.0
13.5
22.5
2.0
  Canon D30, D60, 10D
22.7
15.1
27.3
1.6
  Nikon D1, D1H, D1X, D100
23.7
15.6
28.4
1.5
  Pentax *ist D
23.5
15.7
28.3
1.5
  Canon 1D
28.7
17.8
33.8
1.4
APS
Advanced photo system film cameras (various crops in camera) APS-H here at 16:9 ratio
30.2
16.7
34.5
1.2
35mm
35mm film cameras.
Full frame digital SLRs such as Contax N, Canon 1Ds, Kodak DCS14n
36.0
24.0
43.3
1.0
645
Medium format 120 roll film
56.0
41.5
69.7
0.6

The normal lens above shows what the default should be for the focal length that gives a generally accepted normal perspective to a photo, the smaller the image sensor, the shorter the focal length of the normal lens. Of course the normal lens is a rough guide only, typically the normal for 35mm is regarded as 50mm which is sort of close to the true 'normal' diagonal dimension of 43.3mm.

The focal length factor is based on the width of the sensor in question related to the width of the usual 35mm film frame. I chose to use the width as that is usually the most important dimension of a frame. Most charts use the diagonal measurement which can be a little misleading at times due to differing aspect ratios involved. Use that factor to multiply the given lens focal length with that sensor size to find the 35mm equivalent lens. For instance, the Nikon D100 uses a 23.7 x 15.6 mm sensor and has a factor of 1.5, if you use a 28mm lens on the D100 then it will behave as a 42mm lens (28x1.5=42).

Please realise that only 35mm film, some digital SLRs and APS film have the aspect ratio of 3:2, all the rest have the aspect ratio of 4:3 just as in TV sets, although some of the 4:3 ratio cameras can also provide 3:2 images as an option. This confuses the comparison of diagonals a bit so that is why I chose to compare the longest frame size for the lens multiplier factor. Added to that aspect ratio confusion is the fact that some very common print sizes all have different aspect ratios. Notes about prints on another page (being written). 

Be aware that due to the way they operate, CMOS sensors are inherently "noisier" than CCD sensors, but if the individual pixel size can be increased by making the whole sensor larger or better layout on the sensor chip, then at a certain point, and with better mathematical fiddling, the CMOS chip can be made to be almost as good as an identically sized CCD chip. The advantages of the CMOS chip are easier manufacture, more logical elements can be installed on the same substrate (thus making possible one-chip cameras), and lower power consumption. For really critical low noise applications like astronomy then the CCD device will probably still be used.
CCD = Charge Coupled Device
CMOS = Complementary Metal Oxide Semiconductor, don't worry, it's just the description of the basic chip construction method.

Of course the smaller sensors and smaller focal lengths used do have an effect on depth of field. The depth of field is generally much bigger so making it harder to isolate subjects by blurring the background. More on this subject here

More on sensor size:- the reviews of the Canon 1Ds camera that has an 11 megapixel full frame size sensor (so that a 28mm lens retains a 28mm lens angle of view) indicate that the resulting image is so noise free that the result is more pleasing than medium format film cameras. OK the resolution is not as good as medium format but the overall effect is better and is preferable to film. Many think that 35mm film is doomed, but I see that more likely the 120 film is doomed as there is now no point to have a medium format camera when a smaller and (in the future) cheaper 35mm based digital outfit will do a better job. This should be even better with the slightly more pixels of the Kodak-Nikon SLR and of course future developments like the possibility of the Foveon type of chip evolving to full frame size and many more pixels, there is great hope for the future.

Now all we need is an adventurous chip and camera maker to come out with a B&W only digital camera. That could be made possible by getting rid of the Bayer pattern of RGBG filters on the pixels and would dramatically increase the resolution of the B&W image due to the lack of interpolation that is needed with the Bayer pattern to arrive at the individual RGB values for all pixel sites. Put in a removable IR filter and the camera would appeal to many current very traditional B&W workers. Pipe dreams, but maybe one day....

Probably the best link for more real world information about digital imaging ....
http://www.luminous-landscape.com/tutorials/dq.shtml


Back to Index