Dynamic MTF improvement scheme and its Validation for CCD

operating in TDI mode for earth imaging applications

Neeraj Dubey, Arup Banerjee

Space Applications Centre, Indian Space Research Organization, Jodhpur Tekra, Ahmedabad,
Gujarat, India, 380015

ABSTRACT
The paper presents the scheme for improving the image contrast in the remote sensing images and highlights the novelty
in hardware & software design in the test system developed for measuring image contrast function. Modulation transfer
function (MTF) is the most critical quality element of the high-resolution imaging payloads for earth observation consisting
of TDI-CCD (Time Delayed Integration Charge Coupled Device) image. From the mathematical model for MTF Smear
MTF of 65% (35% degradation) is observed. Then a operating method for TDI-CCD is developed, using which 96% of
Motion Smear MTF will occur within the imaging operation. As a major part of the validation, indigenously designed and
developed a test system for measuring the dynamic MTF of TDI Sensors which consists of the optical scanning system,
TDI-CCD camera drive & video processing electronics, thermal control system and telecentric uniform illumination
system. The experimental results confirm that image quality improvement can be achieved by this method. This method
is now implemented in the flight model hardware of the remote sensing payload

Keywords: MTF, Slant edge, Autofocus, non-uniformity correction.

1. INTRODUCTION
This paper introduces two related contents, the first being the development of novel charge transfer methods for improving
the smear MTF and second is the development of a test system for measurement of “Dynamic MTF” of TDI image sensors.
The charge transfer method for improving the smear MTF is being discussed in [14] by Wang for “multi-transfer gate”
CCD architecture. The dynamic MTF test system is discussed in [4] by Schwarzer using point spread function approach
and in [6] by Bardouxi using continuous rotary based edge scanning.

TDI-CCD planned to be used in the mission has single transfer gate architecture in order to increase the CCD
operating speed. Due to the presence of the single transfer gate in CCD, there is less charge isolation in between the
horizontal and the vertical CCD. Thus, the CCD manufacturer recommends operating the CCD in “ripple clocking” where
in the smear MTF of 65% is achieved. We developed a novel charge transfer method for a “single transfer gate” CCD,
which achieves a smear MTF of 96% and also maintain a good charge transfer efficiency. The paper also discusses the
effect of this new method on other aspects of imaging.

Specific hardware & software system is built in order to measure dynamic MTF of TDI-CCD detectors. The method is
based on a moving edge, synchronized with the movement of charges inside the TDI detector. Unlike other methods the
moving part is a fast steering mirror which has high accuracy, low temperature drift and absence of mechanical stage errors
like eccentricity error, wobble, etc. The measurement methodology is based on the knife edge one: but instead of keeping
it stationary during detector voltage sampling, the knife edge is synchronized with the charge transfer inside the CCD
programmed externally through clock timings, simulating image movement on the detector. In order to extract the contrast
modulation function difference between two sets of timings requires isolation of the CCD pixel data from electrical noise,
electrical coupling, thermal drift, inter-component vibrations, scanning uniformity, illumination uniformity & collimation.

The paper uniquely addresses the novelty in design & development of low noise camera electronics for TDI CCD including
biasing and video processing, thermal control system using refrigerant circulators & TEC (Thermo-Electric Cooler) for
vibration free maintenance of CCD temperature within ±1˚C, vibration removal scheme (using custom software developed

Earth Observing Missions and Sensors: Development, Implementation, and Characterization IV,
edited by Xiaoxiong J. Xiong, Saji Abraham Kuriakose, Toshiyoshi Kimura, Proc. of SPIE
Vol. 9881, 988113 · © 2016 SPIE · CCC code: 0277-786X/16/$18 · doi: 10.1117/12.2228105

Proc. of SPIE Vol. 9881 988113-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 to remove the frame with edge noise of more than defined percentage), intelligent scene based non-uniformity correction
scheme for source and detector correction, fast optical auto-focus scheme using MTFA method ( Maximizing the area
under MTF curve instead of Standard deviation as reported in other papers) and innovative active alignment methodology
using point source instead of passive alignment strategy using theodolite’s and auto-collimators.

In section (2), the paper discuss smear, its causes, effect of smear on imaging followed by mathematical models for the
same. Section (3) talks about the smear MTF improvement scheme by designing of clock timings to the single transfer
gate TDI CCD. Section (4) discuses dynamic MTF test system, (5) highlights the system alignment followed by results &
discussion in (6) and conclusion in (7).

2. IMAGE SMEAR
This imaging sensor implements time delay integration (TDI) technique for improving SNR. It uses multiple detectors
(called as stages) in along scan direction to collect multiple images of the same ground area as the image moves across the
detectors. The multiple images are combined in the detector to increase the effective integration time, thus improving the
SNR. In the ideal imaging system, the rate at which the sensors scan the ground (scan rate) is perfectly synchronized with
the rate at which charge is clocked out of each TDI stage (line rate). This synchronization minimizes the relative motion
between detector and ground, while it captures the image. Ideally, it would require an analog continuous sampling of the
image point or an array of thousands of infinitesimal point detector, both of which have technical difficulty in realization.
Hence, generally there are 1-4 sampling points within each pixel to sample the continuous image plane movement within
a pixel pitch. Due to this the charge collected is not the “exact” representation of the image point in the ground, but an
integration of multiple discrete points images of different time stamps.

2.1 Causes of Smear

𝐷𝑒𝑡𝑒𝑐𝑡𝑜𝑟 𝑃𝑖𝑡𝑐ℎ (𝑏)

Pixel clock rate is defined as 𝑉𝑝𝑖𝑥𝑒𝑙 𝑐𝑙𝑜𝑐𝑘 = --- (1)
𝑇𝐷𝐼 𝐶𝐶𝐷 𝐿𝑖𝑛𝑒 𝑃𝑒𝑟𝑖𝑜𝑑 (𝑡)

𝑉𝑆𝑎𝑡𝑒𝑙𝑙𝑖𝑡𝑒 ∗𝑓
The image scan velocity is expressed as 𝑉𝑖𝑚𝑎𝑔𝑒 𝑠𝑐𝑎𝑛 =
𝐻(𝐴𝑙𝑡𝑖𝑡𝑢𝑑𝑒) --- (2)

If, Vpixel clock ≠ Vimage scan then image smear arises, which leads to that the same area on the earth's surface will not be
imaged by the same TDI stage. The pixel clock rate in one line period could be expressed as the summation of the 2Φ
𝑏∗(𝑡−𝑡(𝑖))
individual charge transfer velocities 𝑉𝑝𝑖𝑥𝑒𝑙 𝑐𝑙𝑜𝑐𝑘 = ∑2∅
𝑖=1 --- (3)
2∅∗∆𝑡

ysica ' eaning o c oc iming w.r. image Physical Meaning of clock timing w.r.t image
motion vector : Disslacement motion vector Velocity
:

Photo sensitive Pixel - - Photo sensitive Pixel - -

Image point -
Image point - 437 cm/sec
v

Velocity
8 cmJsec

Time T 85.75 2us~

US
Pixel Time

Figure 1: Image transfer and scan velocity

Where ∅ 𝑎𝑛𝑑 ∆t represent the number of phase steps inside the CCD and duration of charge transfer respectively. The
term t(i) represents the instant that charge transfer occurs. From Eq. (2) and Eq. (3), the velocity of photosensitive pixel

Proc. of SPIE Vol. 9881 988113-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 and image point in shown, as well as the relative motion between photosensitive pixel and image point are shown in Fig.1.
The displacement difference between the red real line (photosensitive pixel) and blue dashed line (the image point) is just
the cause of clocking smear.

2.2 Effect of Smear on Imaging

Figure 2 shows different cases of smear. An imaging system with (a) no smear, (b) 1/2 pixel along scan smear,
(c) 1/2 pixel cross scans smear, and (d) with a 1/2 pixel along scan smear component and a 1/2 pixel cross scan smear
component. In (a), the same point on the ground (x) is imaged by all n TDIs; in (b), (c), and (d), the same area on the

TDI TDI TDI TDI

Elements Elements Elements Elements
#1

*2 x x

1161 Velocity
Motion

*(n-1)

a(n)

a) b) c) d)

Figure 2 Types of Smear with TDI CCD Stages

ground is not imaged by all n TDI.

When TDI sensor is used for imaging, it needs very tight synchronization between ground scan velocity (GSV) and line
velocity (integration time). A slight change in height may lead to change in GSV which may cause loss of synchronization.
This causes along scan smear in the image generated by the TDI sensor.

MTF for along scan direction smear is given by equation (4)

1 𝑆 𝑠𝑖𝑛(𝜋𝑆𝑓𝑥 )
𝑀𝑇𝐹𝑠𝑚𝑒𝑎𝑟 (𝑓𝑥 ) = 𝑠𝑖𝑛𝑐 [𝜋 ( + ) 𝑓𝑥 ] --- (4)
𝜙 𝑁𝜙 𝑁𝜙 𝑠𝑖𝑛[𝜋(𝑆/𝑁𝜙)𝑓𝑥 ]

Where
Φ is the number of phases used to transfer charge from one TDI stage to the next stage.
N is the total number of TDI stages in sensor
S is a smear
When S=0, MTFsmear is unity.
Smear ‘S’ is calculated using various parameters. Smear ‘S’ is the amount of variation in an area seen by the first stage to

Figure 3: Effect of Smear on images. The right image is with

Smear of 5% and the left image with smear of 1%
the last stage in terms of pixel size. This calculation needs some intermediate terms like ground scan velocity, ground scan
distance, and detector scan distance. Spatial frequency fx used in this formula should be normalized w.r.t. detector size as

Proc. of SPIE Vol. 9881 988113-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 Present Timings

at

VerticaIPlxel 2
..71117171717'r'71

Phases, Páels&Sub -Páel

Vertical Pixel 1
98%

Smear MTF =0.55

Readout Rate= 6.67 MHz
Motion Ph asetransfers = 1

Gate
TX
Figure 4: Initial Ripple Clock timings suggested by Figure 5: Ripple clocking explained with reference to charge
Manufacturer transfer in sub phases of the pixel

Horizonte
Gate
smear S is calculated w.r.t. detector size. The ground scan

%
velocity for the platform orbiting around the earth at orbital height h above the earth's surface is given by equation 5.
𝑅 𝐺𝑀
𝑉𝑔𝑠 (ℎ) =
𝑅+ℎ
√
𝑅+ℎ
--- (5)

Where
R is the average radius of the earth (6371e3 m)
h is the height of the platform above the earth's surface and
GM is earth constant (3.986005e14 m/ Sec)
Knowing the value of Vgs: the ground scan distance Digs: which is the amount of distance scanned by imaging beam on
the ground during its exposure time, can be calculated by equation 6
𝐷𝑔𝑠 (ℎ) = 𝑉𝑔𝑠 (ℎ) ∗ 𝑇𝑠
--- (6)

Where Ts is the exposure time of the detector. Translating this ground scan distance from an object plane to image plane
gives detector scan distance which is obtained using equation 7.

𝑓𝑙 --- (7)
𝐷𝑑𝑠 (ℎ) = 𝐷𝑔𝑠 (ℎ)
ℎ

Where fl is the effective focal length of the system. Knowing the amount of distance traveled in the detector plane for given
integration time, the amount of total displacement S in number of pixels can be calculated by equation 8.

𝐷𝑑𝑠 (ℎ)
𝑆 = 𝑁𝜙 [ − 1] --- (8)
𝑑

Where N is a number of stages in the TDI sensor and ϕ is the number of phases used for charge transfer in TDI, d is the
detector pixel width. The smear is zero when Dds is equal to the detector size d, which means that when the system is
perfectly synchronized image formed on detector moves by the distance equal to the detector size in one integration time.

Proc. of SPIE Vol. 9881 988113-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 Best Case Timings
99%

ntPg,at on
ne

Readout
99 ti
ne

Smear MTF =0.992

Readout R ate= 8.37 MHz
Motion Ph asetransfers =8

Figure 7 Architectural difference in Panchromatic &

Figure 6: Best Case timings
Multispectral TDI
3. ENHANCEMENT OF MOTION MTF DUE TO CLOCK SMEAR BY REDESIGNING
NOVEL TIMINGS
Improving the most critical payload performance parameter, i.e. modulation transfer function by ~ 48% is the major goal
of this activity. Improvement in the smear MTF is achieved by carrying out following changes in the Detector driving
conditions as explained below:

TDI-CCD requires 3 different groups of timing totaling to 16 clock timing signals for charge shifting. As shown in figure
4 the first group of timing is called the “integration time clocks” which are basically vertical transfer clocks, the second
group is called “Readout time clocks” which are the horizontal shift register clocks and the third group is called “Transfer
time clocks” which are the clocks for dumping the charge from vertical CCD to horizontal CCD. Timing Clocks are
generated by the FPGA in camera electronics. An improvement of 48% in the MTF is achieved by designing special clock
timings for the TDI-CCD.

The manufacturer of TDI-CCD proposes a “ripple clocking” timing as explained in figure 5. Here in the charge transfer in
the 4 phases of the pixel occurs in the very fast ripple mode. During the rest of the integration time, 2 phases are integrating
and the rest two are in blocking mode. Figure 4 shows that only 2% of the time is used in Charge transfer process (8 steps
for 4 phases CCD) while the rest 98% of the time is used for integration & readout process. This causes the TDI Sensor
to behave virtually like a linear CCD with huge smear MTF loss (65%). Figure 6 shows that best possible timing with all
the charge transfer steps divided uniformly w.r.t integration time, causing the highest (99%) of smear MTF.

In order to implement “best case timings” requires CCD gate architecture with multiple transfer gates as carried out in [14]
and also in Our MX CCD as shown in figure 6. Due to the presence of multiple storage gates, vertical & horizontal CCD
has good isolation in between them, which reduces the charge leakage in between the two CCD.

3.1 Novel Timing for a single transfer gate structure CCD:

The MTF of the clocking smear can be expressed as

𝑀𝑇𝐹 = ∗ ∑2∅
1 −𝑗2𝜋(∑𝑖𝑗=1 𝑙𝑗 −
(𝑖−1)𝑏𝑓𝑥
--- (9)
𝑖=1(𝑙𝑖 ∗ sin 𝑐(𝜋𝑙𝑖 𝑓𝑥 ) ∗ 𝑒 )
2∅
𝑏

From the equation 9, more the number of sub- transfer steps in a single pixel the higher is the smear MTF. It can also be
seen from the figure 8 that the smear MTF improves drastically from single phase to ‘3 phase’ timing and then tends to
saturate from ‘4 phase’ to ‘8 phase’ timing. Hence 4 phases transfer (8 sub transfer steps) seems to be the best method of
achieving the smear MTF.

Proc. of SPIE Vol. 9881 988113-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 However, the incorporation of such “best transfer scheme” required multiple transfer gates in the vertical to
horizontal transfer circuitry (as shown in figure 6) so that the vertical integration and horizontal shifting (of the previous
frame) are isolated with reference to each other. This feature is absent in most of the high speed TDI CCD, which are
designed with a single transfer gate as shown in the figure 7.

Due to this limitation the best timing option shown in the above figure is not possible to implement. In this work we
develop a novel new timing as shown in the below figure which improves the smear MTF, as well as have a good isolation
between horizontal & vertical stages thereby achieving CTE values better than the ripple clocking method.

As shown in figure 9 & 10, we implement an 11/2(One & Half Phase) phase timing which has 3 transfer steps
during the integration phase as well as good CTE (Charge transfer efficiency). The 3 sub-step transfer phase shown in
figure 9 achieves an improvement in smear MTF by 48%. The Eight transfer step “best case” timing approach would have
increased the smear MTF by 52%. In this timing the transition (negative) of first vertical phase and the transfer gate clock
are maintained same to have good CTE performance and good isolation for no charge leakage.

4. DYNAMIC MTF TEST BENCH

A highly repeatable test bench was developed to characterize the “Dynamic Motion MTF” of the TDI imaging sensors.

As shown in figure 11 &12, following are the main blocks of the test setup:

1.2
0.988 0.994
Proposed Timings for FM
0.962 0.967
1
96%
Transfer
0.8 Time 1
a
0.6
2
0.4
Readout 96 %
0.2 Time

Smear MTF =0.962

0
Readout Rate= 667 MHz
1 Phase 3 Phase 4 Phase 5 Phase 8 Phase Motion Phasetransfers =3ó1g 0 3 Small

Figure 8 MTF Values of 1/3/4/5/6 phase vertical CCD Figure 9: Proposed timing for Flight Model Implementation:
transfers 1 1/2 phase timing
Six axisA l ign ment
Slow Vertical Timing ( New)
Fast
Scanning
Collimated
Mirror
Target
Project) on

Focusing
Lens

Auto Lens
Focusing
Mechanism

DAQ, &Stitching
Software, Data
r Analysa
N
__' l /
\ / 8K TDI Detector Thermal

8K TDI Camera
Electronics
MI Control
System

Figure 10: New One & Half Phase timing explained with Figure 11 Optical test setup Diagram
pixel sub phase CI1, CI2, and CI3& CI4. Colored well shows
charge collection location with reference to the timings

Proc. of SPIE Vol. 9881 988113-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 a) TDI detector drive & processing electronics mounted with six axis alignment options

b) Custom lens with autofocus system

c) Scan mechanism

d) Data acquisition system with synchronization hardware

e) Thermal control system

f) Target projection system

g) Novel software with data processing, display, vibration rejection, 3D PRNU correction, auto focus & auto velocity
tuning.

4.1 TDI Detector Drive & Processing Electronics Mounted with Six axis alignment options

Apart from a host of off-the-shelf precision electro-optical instruments, the sensor array requires precision low noise,
programmable bias signals. The charge storage and transfer mechanism of CCD demands various clocks with precisely
programmable levels and timing synchronization. The detector requires 34 clocks for complete operation and to cater to
these precise demands required hardware is designed, developed and tested in-house. These programmable clocks / stage
selection logic is locally generated using an FPGA and level shifted independently using Intersil Elantec clock driver
EL7457. Appropriate damping resistors are used to control rise/fall times, undershoots and overshoots. Detector requires
a total of 14 bias voltages (7 for each array). These bias voltages are locally generated with appropriate limits using a low
dropout regulator, with over voltage and over current protection features. The bias voltages, clock timing relations are
programmable using PC based USB Interface and there is no need to reprogram the FPGA or change the resistors for bias
tuning.

The sensor array is read out using 16 analog video output amplifiers. These video signals are appropriately terminated and
simultaneously digitized to 14/16-bit levels using AFE. There is option for independent initializing and programming each
of these AFEs for optimization of gain, offset, timing, etc. The digitized video data are acquired to the PC hard disk using
the serial data link. Optionally, the sensor data can be read out using the National Instruments analog acquisition system
with NI PXI 5122 card interface.

4.2 Thermal Control System

The high Voltage swing & high Speed capacitive drive conditions leads to very high power dissipation inside the CCD.
Typically, the power dissipation is around 4W. High power dissipation inside the detector leads to a very large temperature
rise, which ultimately leads to a high dark current, high thermal noise, higher quantum efficiency. Hence, efficient
prediction and computation of electro-optical parameters like MTF demand’s that TDI CCD temperature be controlled.

Figure13: TDI Drive & Processing Electronics & thermal

Figure 12 Test Setup Architecture &Components control system

Proc. of SPIE Vol. 9881 988113-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 The CCD /TDI devices are packaged in a ceramic (Aluminum Nitride package) in order to remove the heat from the Silicon
dye. The packaged image sensor is mounted over an Invar cold finger to match the CTE of both the materials. The contact
conductance of Invar and CCD/TDI assembly is improved by stacking two thin Cu foils (70 um thick) in between them.

The invar cold finger is mounted on a thick copper block through flexible PGS (Pyrolytic Graphite sheet). The copper
cold finger is a thick copper block, supported on to the frame with spring mechanism for insertion / removal of the sensor
in the socket. Copper block has a circulator block on one side with attached heat sink and Thermo Electric Cooler (TEC)
on the other side. TEC also has a circulator water block attached on the other side. Since the assembly works in lab
continuously for hours for characterizing and testing the TDI/CCD image sensor, the heat removal from the TEC is done
by refrigerating Chiller with circulator blocks. Figure 13 shows the front and back view of the custom developed hardware
for the thermal control system of the TDI CCD.

4.3 Scan Mechanism

There are lots of methods of scanning an image along with the TDI Vector:

1. Drum Scanning
2. Convergent Beam Scanning
3. Parallel Beam scanning
4. Rotary mounted polygon scanning
5. Linear Stage scanning
Here, in this project parallel beam scanning method is employed due to the inherent stability in other axes and high accuracy
in the scan axis, low temperature drift and high scanning frequency.

The scanning is achieved by using Newport FSM-300. It's a 2 axis tip/tilt mirror with a maximum scan angle of ±26
married with a step size of 1 μrad and closed loop Bandwidth of 800 Hz. The FSM comes with its own controller, which
can be controlled by an analog input Signal of ±10V range. Since the scanning velocity of the FSM has to be synchronous
with the TDI charge motion velocity, following strategy is adopted:

Displaced
Un- Displaced
Mirror
Mirror

Figure 14 SCAN Geometry

 TDI charge motion is controlled by a set of timing signals which is generated by an FPGA in timing generator
card.
 The FPGA requires a master clock for its operation which is provided by an external function generator instrument
 The function generator is Tektronix make AFG-3102 with dual channel output. One channel is used to generate
an 80 MHz Square wave clock to the FPGA timing card, and the other channel is used to generate a ramp signal
to the FSM controller. Both the clock’s generated are synchronous and phase locked.
 Continuous waveform is given to the scan mirror and the TDI Sensor. The FSM is given a “ramp up” waveform
in the first phase with ramp rate matching to the charge motion velocity. Following are the formulas used to
calculate the velocity in both the cases:
𝑝𝑖𝑥𝑒𝑙 𝑝𝑖𝑡𝑐ℎ 7 𝑢𝑚
𝑇𝐷𝐼 𝐶ℎ𝑎𝑟𝑔𝑒 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 = = = 0.07𝑚/𝑠 --- (10)
𝐿𝑖𝑛𝑒 𝑡𝑖𝑚𝑒 90 𝜇𝑠

Proc. of SPIE Vol. 9881 988113-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 𝑏 sin 2𝛾
𝜃 = tan−1 --- (11)
𝑎 + 𝑏 cos 2𝛾

For Small angular shifts (figure 14)

2𝑏𝛾
𝜃 = 𝑎+𝑏 --- (12)

𝜃 2 (𝑎 + 𝑏)2
∈= --- (13)
2𝑏

∈ = 𝐷𝑒𝑓𝑜𝑐𝑢𝑠 𝐸𝑟𝑟𝑜𝑟

𝛾 = 𝑚𝑖𝑟𝑟𝑜𝑟 𝑡𝑖𝑙𝑡 𝑎𝑛𝑔𝑙𝑒

𝜃 = 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑠ℎ𝑖𝑓𝑡 𝑖𝑛 𝑡ℎ𝑒 𝑐ℎ𝑖𝑒𝑓 𝑟𝑎𝑦

𝑎 = 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑚𝑖𝑟𝑟𝑜𝑟 𝑎𝑛𝑑 𝑙𝑒𝑛𝑠

𝑏 = 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑏𝑒𝑡. 𝑚𝑖𝑟𝑟𝑜𝑟 𝑎𝑛𝑑 𝑑𝑒𝑡𝑒𝑐𝑡𝑜𝑟

4.4 Custom Lens with Autofocus system

Since we are using a parallel beam scanner, it requires a collimated beam of light from the target projection. But due to a
collimating lens, there is a decrease in the intensity many folds. So, in this setup we use a “Bi-telecentric” lens which is
telecentric in image and object space. In order to get the smaller spot size and higher Optical MTF at 70 lp/mm (equivalent
to the pixel Pitch of the detector), we have procured a custom built telecentric Lens of F#2. The lens has a low distortion
(10) and r-theta compensation due to the scan projection being an arc and not a line. The defocus error at the edges will have
an impact on overall charge integration, because as the charge gets integrated over the stages, the higher stages will have
a bigger projection pitch. The F#2 telecentric lens is mounted over a motorized Liner Stage from Zaber controlled through
a PC based Labveiw interface. The various focus position is scanned in the linear direction and the data is acquired. For
each set of data contrast is measured along with the area under MTF Curve. Best focus position is measured by maximizing
the image frame contrast.

A new, novel method is found from this experimentation which maximizing the area under the MTF curve, which gives
more accurate focus position than the contrast maximization function.

4.5 Target Projection System

The Target projection system consists of a uniform integration sphere with a QTH lamp, a spectral band pass filter
illuminating a special target as defined in figure 15.

To obtain the data for computing MTF of an FPA, a sharp edge (sharp razor blade with black coating in this case) and
a 5um pitch pin hole is prepared as a target element. A small pinhole was put in the cavity area of a razor blade to make it
a planar structure. The Slant angle of the edge was measured from the acquired image using trigonometry and was adjusted
to ~60 with respect to FPA column direction. The Slanted edge based measurement has very less dependence to the slant
angle below 10. (Slanted Edge with precision hole) is made to carry out the target alignment & imaging with the following
steps:

• Step 1: Point source based active alignment of optical elements & velocity tuning
• Step 2: Spatial frequency response / MTF measurement using Slanted edge.
• Step 2 is additive to step 1, hence Setup should not change in between. Hence the Slanted edge and the pin hole
have to be in the same plane.

Proc. of SPIE Vol. 9881 988113-9

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 The Specially prepared target is mounted over tip-tilt goniometer which has a provision to tilt for tilt angle adjustments (2
to 10degreese) and is aligned with other optical components as shown in fig. 16.

Tilting Mechanism Knife Edge Dynamic MTF Setup

Target: Blade Integrating sphere with 8K TDI with Drive and video Thermal System
target \ processing Electronics
I
with Black
particle
deposited

Pin Hole : 5 um
Target

Alignrdenf
Stages
/
Fast Scan Mirror mounted on
tip/ tilt platform
1

Lens : F n 2 Sync ronizing

Electronics

Figure 15: Special Target with slanted edge & pin hole Figure 16: Dynamic MTF Test Bench
4.6 Novel Software

Novel Software with data processing, display, vibration rejection, 3D PRNU correction, auto focus & auto velocity tuning
and MTF Analysis is developed in house.

Parameters to optimize for the software

• Focus position (Lens)

• Tilt angle (Scan mirror & detector)
• Velocity mismatch (Scan mirror & detector)
• Azimuth & elevation mismatch (Scan mirror & detector)
• Lens tilt
Challenges in the test setup

 PRNU correction algorithm

 Auto focusing & auto velocity mismatch algorithm to reduce the alignment time, accuracy
 Alignment of optical components
 Vibration proof system
4.7 3D Non-Uniformity Correction

Here in this project we have implemented a scene based non-uniformity correction instead of a standard single matrix non-
uniformity correction. Initially the sensor is characterized with the same setup for different illumination values (From Dark
to saturation signal around 30 sets of illumination profile are saved). For each set of illumination, non-uniformity correction
matrix is prepared and stored in the memory. Hence now we have a collection of NUC matrix in a 3D manner. The
“intelligent” software then checks the pixel values one by one for their intensity and based on their intensity appropriate
non-uniformity correction is applied.

Proc. of SPIE Vol. 9881 988113-10

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 4.8 Vibration Rejection scheme

During imaging of the slanted edge target, in order to reduce the photon shot noise, multiple frames are acquired and
averaged so as to reduce the photon shot noise. However, there are differential vibrations introduced amongst the Opto-
electronic components due to cooling fan of the light source, instruments, air turbulence, movement of a person in
proximity to this setup, etc. Due to this vibration introduced, frame averaging is not possible. Hence, software is prepared
(fig 17) to sort out the “good frames” and reject the “bad frames”. Once the good frames are sorted out, they are averaged
and given to the next chain of image processing. The “good frames” are typically selected by measuring the deviation of
the edge transition point in the subsequent frames and rejecting those who are deviating beyond the acceptable limit (which
is user programmable).

Figure 17: Vibration rejection feature in software

Following are the changes in the frame averaging policy (Sorting algorithm):

 Acquire more than 1000 frames

 Plot the transition edge point on the 1000 frames.
 Carry out 1000 frames averaging of this point.
 Remove the frame which is deviating by 0.1% of the average number
 Carry out frame averaging of the remaining frame. (typically around 50)

4.9 Auto focus & auto velocity mismatch program (Fine)

• Motorized linear Stage can be commanded to travel through PC interface.

• Velocity can be tuned througha PC based USB interfacee of FG.

• The main optimizing parameter was to see the “Standard Deviation” while changing any of the above 2
parameters.

• Better, accurate and more repeatable method is to see “Area Under the curve” of MTF Curve.

Proc. of SPIE Vol. 9881 988113-11

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 5. SYSTEM ALIGNMENT
Previous work carried out in similar setup development used auto-collimators for tuning the optical angles of various
optical components. Here in this setup we adopt a new active alignment methodology based on Dynamic PSF measurement
as reported in [4]. Alignment algorithm is written in figure 18 and figure 19 shows the PSF based alignment methodology.
Alignment of the optical & electro-optical components were carried out dynamically by following optimization

• Focus position (Lens)

• Tilt Angle (Scan mirror & detector)

• Velocity mismatch (Scan mirror & detector)

• Azimuth & elevation mismatch (Scan mirror & detector)

• Lens tilt

Initial Setup Preparation : Two Scan

axis possible X -X' and Z -Z'.
Choose axis X -X' such that the 45
degree angle is in the direction of
mirror scan
e= Along Y (upper Do Manual Focus
gonio) Check if the
tuning within 10 intensity of the
A Along Z ( rotory) steps movement
0 = Along X ( lower light source is
gonio) uniform for the
X axis is parallel to Do velocity 7 whole array
detector TDI direction tuning manual
Y axis is Perpendicular No
to Detector Face
Adjust Tilt angle O Spot like a +
along Y axis sign and not a
tail
Try to move a
spot in a line, vos
If the spot have Adjust A
variable intensity
, then varry A
Adjust m
To remove a
fixed blur in the 1
spot above Adjust e again ( tilt angle)
head adjust 0

Adjust Velocity again

Adjust Lens angles to remove
any stray residue light causing
blur Page 1

Figure 18: System Alignment Flow Chart

5.1 Alignment (Coarse)

• Try to image a point source and see for distortions i.e. if the point is imaged like a tilted line.

Proc. of SPIE Vol. 9881 988113-12

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 • Based the tilted Line slope & length determine the mismatch factor and then carry out corrections
𝑦
Tilt Angle = ε = tan−1 𝑝𝑖𝑥⁄𝑁 --- (14)
N = number of TDI stages
Speed Ratio = SR = 1 + xro / N --- (15)
Where Xro and Ypix are the co-ordinates of the upper point of the Spot image.

5.2 Importance of alignment angles (∅_𝒙, ∅_𝒚 & ∅_𝒛)

• ∅𝑥 is taken care by tilt adjustment

• If ∅𝑦 is not matched, as the scan mechanism scan from top row to bottom row, the pixel projection changes i.e
Spot become diffused and you can see an intensity variation across the line.

• ∅𝑦 The mismatch is identified by scanning the target in y direction and adjusting it till the same intensity is
observed in all y points.

• ∅𝑍 mismatch along with lens tilt will cause a constant horizontal defocus
'Frame Sortirg &1
Slantedtarget Image ROI Selection - Averaging for
ErrorAngle J, p'xe: 1 p:xe: 2 p:xe: vibration
Dark Correction
1, correction ,
I
VelocdyRatio 4,
Scene Based Non -
Uniformity
0.9 Correction
I
alifter 1%4 +
Super Sampled ESF LSF Creation by
Creation Differentiation
1.0

O V
Sl idirg Window 4(-(r

averagefor Noise Hamming Window Gaussian Window

reduction
1.1

00°

i.
FFT
>(
4,
Curve Plotting
Figure 19: PSF system alignment based on [4]
Figure 20: SFR/MTF Analysis Algorithm
WellOSOINSI WW1

Dynamic MTF Curves ( Measured with timing 2)

11.9010101111R110.4 4111112sn

MTF

E
ESF PSF -Slow&
MTF
-Fast*

o
5 15 25 35 45 55 65 75

a4944m4Y Iblmml

Figure 21 ESF, PSF & MTF Analysis tool Figure 22: MTF (Measured) for Slow (New) timings &
FAST (Old) Timings

Proc. of SPIE Vol. 9881 988113-13

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 5.3 SFR analysis software (Modified ISO 12233)

Acquired image is first corrected using two point non-uniformity correction algorithms and calibration data obtained using
uniform illumination of FPA at various intensity levels. Region of the image covering dark and bright portion around the
edge is chosen for analysis. An edge spread function (ESF) is generated by rearranging pixel data in the selected region.
A numerical derivative of the ESF is computed to obtain the line spread function (LSF). A Fourier transform of the LSF
is computed to generate the MTF and it is scaled to the appropriate spatial frequencies knowing the FPA pixel size and
over sampling factor. The correction of the numerical errors of derivative computation is applied to the computed system
MTF. FPA MTF is computed from the measured system MTF knowing optics and target MTF.

Changes in the basic MTF algorithm are

• Added Gaussian window with removing data having SD more than “X”

• Added Moving Average Window to reduce the noise effect and smoothen the edge

Modified SFR/MTF algorithm is shown in figure 20 and figure 21 shows the ESF (Edge Spread function), PSF (Point
Spread Function) and MTF Extraction in the tool.

5.4 Electro-optical criticalities addressed with the new timing of the TDI CCD:

 Charge transfer inefficiency (Vertical transfer & horizontal): Charge transfer is mainly due to the drift
diffusion mechanisms and the fringing field effect happening inside the CCD which require critical delay margins.
Failure to meet the different clock delay requirements leads to high VCTI, HCTI.

 Dark Signal Couplings: Due to the vertical transfer now taking place during the integration, it leads to high
vertical clock couplings in the pixel values. The transitions are adjusted to cause minimum coupling.

 Back End Antiblooming: In order to take care of the charge left on the other side of the Stage control gates, and
to drain them properly and avoiding any spillage in the neighboring pixels requires proper design of the timing
signal.

 The sequencing of the gates in the phases is changed to meet the maximum MTF requirement. Earlier Integration
was starting from phase 2-3 and so on, whereas now it is 1-2-3 and so on. This required careful creation of the
potential profiles of the TDI CCD.

 Selective saturation & leakage: Group of pixels having charge leakage & different saturation values.

Figure 23 Imaging in TDI Mode

Proc. of SPIE Vol. 9881 988113-14

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 6. RESULTS & DISCUSSION:

As shown in figure 22, 23 & 24 improvements in the image contrast was noticed. Improvement in Payload MTF
by 22% was experimentally measured. The results are highly repeatable and the same was checked by carrying out
repeated measurements. The off axis misalignment, tip & tilt corrections, auto focusing, auto velocity mismatch, were
done before the actual measurement and ensured to be same for both the new and old timings. Any leftover misalignment
error is same for both the new & old timings. The reason for the steep fall in MTF Curves is due to the trapezoidal aperture
response of the pixel. The deviation in theoretical & experimental values was due to the Charge transfer inefficiency (CTI).
The CTI in the new MTF timings was 0.27 as compared to 0.003 for old timings. This CTI will cause an MTFcti of 58%
against 99% of the newer timings. This 70% improvement of the MTF cti of 58% will lead to improvement in the Total
MTF from 22% to 37%. The rest 11% is attributed to random vibrations & turbulence MTF loss.

-2 -1
._. ill III-i lII
C-

-
__
_= 111 = 111
111 =c <
3 .3 III =
111 Ell. Ill -1? 111=6
1/1_s
/1
2
6 JII
Figure 24 Edmund Optics USAF Target 3" imaged by TDI Array in TDI Mode: Top Shows the image
taken by the TDI array, the Bottom image shows the actual USAF Target

Due to implementation of the new timings there is a coupling during the clock transitions which typically affects
a swing of 2mV higher signal in 1 pixel / port. Hence, 16 pixels in each 8K array will have a different calibration coefficient.
This effect should not cause any problem since the calibration coefficients are separately defined for each pixel.

7. CONCLUSION
New timing is developed which improves the payload MTF by 48%. The TDI-CCD manufacturer has recommended
the fast transfer timings which have a smear MTF of about 65%. In, this TDI detector, slow timing is difficult to implement
using single transfer gate architecture. Slow timing is developed & tested which will give a MTF improvement of 48%. In
the due course We have also developed a test setup to characterize the motion MTF of the TDI Image Sensors.

ACKNOWLEDGEMENTS
Author wish to thank Shri. P N Babu (Group Director, Electro Optical Systems Group)), Shri Saji Kuriakose (Past Deputy
Director, Sensors Development Area), Shri R M Parmar (Deputy Director, SEDA) for their support and guidance. I am
thankful to Mr. Anukaran Topno and Nikunj Zala for providing support to measure the MTF, Yogesh Shinde for his
Labveiw based SFR software development and Payal Sharma for FSM based activities. I am also thankful to Shri S B
Chakraborty, Shri D R Goswami and Shri Nilesh Desai for reviewing this paper critically.

REFERENCE
1. High precision MTF measurement instrument for focal plane arrays with on-chip TDI: H. Gumbel, S. L. Golub,
F. W. Adams, Jr., K. Morimoto, P. 0. Wolford,R. B. Jones Lockheed Palo Alto Research Laboratory
2. DYNAMIC MTF, AN INNOVATIVE TEST BENCH FOR DETECTOR CHARACTERIZATION Rossi
Emmanuel, Lardière Raphaël, Delmonte Stéphane, Alcatel Alenia Space
3. Synchronous Control Analysis of TDI-CCD Imaging System L. Zhanga,b, Y. X. Lia, X.W. Xub , Nanjing University
of Science & Technology, China

Proc. of SPIE Vol. 9881 988113-15

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 4. Dynamic PSF and MTF measurements on a 9k TDI CCD Horst Schwarzer*, Anko Boerner, Karl-Heinz Degen,
Andreas Eckardt, Patrick Scherbaum German Aerospace Center (DLR).
5. Modeling and measurements of MTF and quantum efficiency in CCD and CMOS image sensors Ibrahima
Djit´eabc. Pierre Magnana EADS Astrium, France
6. DYNAMIC MTF MEASUREMENT Alain BARDOUX1 CNES - Centre National d’Etudes Spatiales
7. Imaging artifacts due to pixel spatial sampling smear and amplitude quantization in two-dimensional visible
imaging arrays Terrence S. Lomheim Aerospace Corporation.
8. Fast MTF measurement of CMOS imagers using ISO 12233 slanted edge methodology M.Estribeau, P.Magnan
SUPAERO – Integrated Image Sensors Laboratory.
9. P. D. Burns and D. Williams, Refined Slanted-Edge Measurement for Practical Camera and Scanner Testing,
Proc. IS&T 2002 PICS Conference, IS&T, 191-195, 2002.
10. Modelling the imaging chain of the Digital Cameras, Robert Fiete , SPIE Ebook
11. Tracy Dutton, Terrence Lomheim and Mark D. Nelson, Survey and comparison of focal plane MTF measurement
techniques, Proceedings of the SPIE, Vol. 4486 #25, August 2001
12. Aperture Response and Optical Performance of Patterned-Electrode Virtual-Phase Imagers, H Hosack , Senior
member IEEE.
13. G. Hochman, Y. Yitzhaky, N. S. Kopeika, Y. Lauber, M. Citroen, and A. Stern, “Restoration of images captured by
a staggered time delay and integration camera in the presence of mechanical vibrations,” Appl. Opt. 43(22),
4345–4354 (2004)
14. Clocking smear analysis and reduction for multiphase TDI CCD in remote sensing system Dejiang Wang, Tao
Zhang,1 Changchun Institute of Optics,
15. G. C. Holst, Electro-optical imaging system performance (SPIE Optical Engineering Press, 2008)
16. S. L. Smith, J. Mooney, T. A. Tantalo, and R. D. Fiete, “Understanding image quality losses due to smear in high
resolution remote sensing imaging system,” Opt. Eng. 38(5), 821–826 (1999)
17. On-orbit spatial resolution estimation of IRS: CARTOSAT-1 Cameras with images of artificial and man-made
targets – Preliminary Results, A. Senthil Kumar, A.S. Kiran Kumar, ISRO

Proc. of SPIE Vol. 9881 988113-16

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/16/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx