
IBIS4-1300
Document Number: 38-05707 Rev. *B
Page 20 of 37
curve is tolerant for external pixel offset error correction. This
means that pixel offset variations can be corrected by
changing the offset after the non-linear AD conversion. This is
so because the non-linear transfer function is
H(s) = 1-exp(-a*s)
by design, and neglecting the offset, the relation between the
non-linear output (y) and the linear output (x) is exactly:
Y = 1024 * (1 - exp(-x/713)) / (1 - exp(-1024/713))
This law yields an increased accuracy of about a factor 2 near
the zero end of the scale. It is thus possible to obtain an
effective 11 bit accuracy on a linear scale after post processing
by applying the reverse law to the non-linear output:
Z = -2 * 713 * ln(1 - y/(1024/(1-exp(-1024/713)))) = -1426 * ln(1-y/1343.5)
Then Z is an 11-bit linear output in the range 0...2047.
Operation of the image sensor
Set configuration & pulse timing
Figure 15. typical operation mode (readout of a frame)
Set configuration
-
output offset
-
amplifier gain
-
viewfinder on/off
-
determine integration
time of next frame
Set flag to restart YR shift
register with SYNC_YR
Execute row initialization
sequence
Acquire pixel (X,Y)
For Y = 0 till Ymax do
Y=
1026 -
integration
time ?
For X = 0 to Xmax do
Setflagtorestart YL shift
register with SYNC_YL
no
X
Yes
StartX shiftregister
SYNC_X
Next Y
[+] Feedback
[+] Feedback