Page 1 of 1

Checks of sensor resolution

Posted: Mon Aug 01, 2016 1:08 am
by forum_admin
There have been a number of recent cameras built where the specified sensor resolution was not quite right. Most notable were the FLIR E4 - E8 series which all used the same 320x240 Tau core with varying degrees of crippling in resolution and noise. See EEVBlog for the gory details, as the performance reduction was not irreversible !

The 'smoking gun' as it were was that the same lens was used on all the models and this proves that the same sensor was used throughout. This can be checked for any sensor / camera / lens combination and as long as 3 of the variables are known or can be guessed the 4th can be calculated.

The lens design focal length / pixel pitch / pixel count and field of view are inextricably linked and the most common question is to work out the sensor native pixel count.

The horizontal width of the sensor is px * pitch
px = horizontal pixels
pitch = pitch of pixels

The centre to edge of the image from a lens is FL* TAN(HFOV/2)
For this you can assume that it is a 'simple' lens and when focused at infinity the 'simple' lens must be at its' focal length from the sensor image plane.
FL = focal length
HFOV = horizontal field of view

So say we know the horizontal field of view is 50° from the datasheet, and the lens has 'FL = 9mm' written on it.
Therefore when focused at infinity the lens must be 9mm from the sensor.
Therefore centre to left/right side of the sensor is 9*TAN(25°) = 4.2mm, and the whole sensor active area is 8.4mm horizontally.

If we know that this is a 25um pitch sensor from the datasheet or sensor family then:
pixels = 8.4/0.025 = 336. So it is most likely a 320 x 240 sensor given the accuracy to which the lens focal length was known.

Alternatively if this was advertised as a 160 resolution camera:
pixel pitch = 8.4/160 = 0.0525mm, 52 um
As no-one still makes sensors at that pitch we would conclude it is really a 320 sensor at 25um pitch being downsized, or possibly a 640 @12um.


Bill