Applications Information
(Continued)
While CIS devices do not usually have a reference level with
which to perform correlated double sampling, many have a
very repeatable reset level which can be used as a black ref-
erence allowing the LM9811 to perform pseudo CDS on the
signal. For more information on CIS applications, see Sec-
tion 9.0. When the Signal Polarity bit is set to a zero, the
LM9811 expects a positive going signal, typically from a CIS
device. When the Signal Polarity bit is set to a one, the
LM9811 expects a negative going signal, typically from a
CCD sensor.
4.6 Offset DAC
The 4 bit plus sign offset DAC is used to compensate for DC
offsets due to the correlated double sampling stage. The off-
set can be corrected in 31 steps of 1.7 ADC LSB size be-
tween 25.2 LSB and +25.2 LSB. Note that the DAC comes
before the PGA, so any offset errors at this stage are multi-
plied by the gain of the PGA. The calibration procedure de-
scribed in Section 5.0 demonstrates how to use the DAC to
eliminate offset errors before scanning begins.
Note
that
this
DAC
is
programmed
calibration/configuration and is not meant to compensate for
pixel-to-pixel CCD offset errors. CDS cancels the pixel-rate
offset errors.
during
LM9811
4.7 Programmable Gain Amplifier (PGA)
The PGAprovides 8 bits of pixel-to-pixel gain correction over
a 0 dB to 9 dB (x1 to x3) range. After the input signal is
sampled and held by the CDS stage, it is amplified by the
gain indicated by the data (“PGA Code”) on the CD0–CD7
databus using the formula:
4.8 Offset Add Bit
In addition to the Offset DAC, there is a bit in the configura-
tion register which, when set, adds a positive 8 LSB offset at
the output of the PGA. This offset ensures that any offset be-
tween the output of the PGA and the ADC is positive, so that
no dark level information is lost due to negative offsets. The
calibration procedure described in Section 5.0 demonstrates
how to set this bit.
4.9 ADC
The ADC converts the normalized analog output signal to a
10-bit digital code. The EOC output goes from high to low to
indicate that a new conversion is ready. ADC data can be
latched by external memory on the rising edge of EOC. The
RD input takes the ADC’s output buffer in and out of
TRI-STATE. RD may be tied to EOC in many applications,
putting the data on the bus only when EOC is low, and allow-
ing other data on the bus (such as CD0–CD7 correction
data) at other times. In this way the output data and correc-
tion coefficient data can share the same databus (see Figure
12).
5.0 CALIBRATION
Calibration of a CCD scanner is done to normalize the pixels
of a linear CCD so that each pixel produces the same digital
output code at the output of the scanner when presented
with the same image light intensity. This intensity ranges
from black (no light) to white (maximum light intensity). The
CCD’s analog output may have large pixel-to-pixel DC off-
sets (corresponding to errors on black signals) and
pixel-to-pixel variations in their output voltage given the
same white image (corresponding to errors on brighter sig-
nals). If these offsets are subtracted from each pixel, and
each pixel is given its own gain setting to correct for different
efficiencies, then these errors can be eliminated.
Ideally the digital output code for any pixel would be zero for
a black image, and some code near fullscale for an image
with maximum brightness. For a 10-bit system like the
LM9811, that code might be 1000. This code will be called
the Target Code.
The LM9811 eliminates these global and pixel-to-pixel offset
and gain errors with its Correlated Double Sampling (CDS),
Offset DACs, Variable Gain Amplifier, and pixel-rate Pro-
grammable Gain Amplifier. This section describes how to
program the LM9811 and the coefficient RAM being used
with it to eliminate these errors.
Calibration of a LM9811-based system requires 3 steps. The
first, described in Section 5.1, Offset Calibration takes a
black image and normalizes the digital output code for each
pixel to a code at or near 0.
The second step, Section 5.2, Coarse Gain (VGA) Calibra-
tion finds the optimum gain setting that places the output
voltage of all the pixels within the 9 dB adjustment range of
the PGA.
The final step, described in Section 5.3, PGA Correction Co-
efficients (Shading Calibration) describes how to calculate
the gain required to normalize the output of each pixel to the
desired output code (the Target code).
5.1 Offset Calibration
This procedure corrects for static offsets generated by the
CCD and the LM9811. Because the LM9811 uses CDS to
eliminate the pixel-to-pixel offset errors of the CCD, no
pixel-rate offset correction is required.
DS012813-32
FIGURE 24. CIS vs CCD Output Signals
www.national.com
25