3.1

Quantum efficiency

It is of prime importance to reach the requested signal level, for the radiometric requirements, with a minimum number of TDI lines, for MTF requirements. Backthinning technology offers detection efficiency around three times higher than what is achievable with standard photoMOS detectors arrays.

The TDI detection architecture coupled with the backthinning technology was the key to enter into the restricted domain of the high resolution imagers.

3.2

Readout frequency and conversion gain

Without any satellite slowing, the mission parameters enforce a readout period of 103μs for the complete 30,000 pixels panchromatic line, that is to say about 300MHz readout frequency for the line.

Specific attention must be devoted to the architecture and design of the output amplifiers in order to maintain low consumption while ensuring short video signal settling times which are necessary to guarantee smooth and flat useful signal levels for correlated double sampling. The design of the output amplifier must also take into account the need of a high electron to tension conversion gain in order to reduce the weight of the electronic noise of the video chain in the overall SNR.

3.3

MTF

The MTF satellite requirement before restoration has been split into allocations to TDI registration discrepancy, telescope, stray light and detector.

The MTF allocated to the detector itself – MTF = 0.50 over the Pan spectral band, excluding the already identified contribution from the number of stages – gave rise to several constraints:

3.4

Antiblooming structure

Instruments which do not integrate anti-blooming structure in the detection sensitive area of the sensor show a specific signature of charge overflow along the columns that degrades the image interpretation.

The occurrence of such figures increases as the aimed ground pitch is reduced and increases for urban areas where the probability of surfaces with high specular reflections is high: roof windows, metal chimneys, car windshields.

Using the relative Sun/Earth/Satellite positions in the programming center allows to avoid specular reflections over horizontal surfaces like seas, rivers or lakes. But for an agile satellite like Pleiades, small human-made glossy surfaces cannot be avoided because they are randomly distributed both in elevation and azimut. One can compute the elementary probability of a given “specular” landscape to be viewed at satellite level, as a function of the elevation range of its normal vector [θ_min, θ_max]:

where Ω_sun is the solid angle of the sun viewed from the Earth (6.8e-5 sr). Let d be the spatial density of bright objects, we can compute the expected number of specular objects in a given urban Pleiades scene of A=20x20 km² with Eq. 3:

For example, roof windows oriented in the range [15°, 30°] lead to approximately 11 specular spots per scene, with a density hypothesis d=1000 windows/km². This is far to be negligible, because large saturated artifacts may hide useful information and lead to systematic poor image quality over urban areas. Moreover, system products are pan-sharpened coloured scenes in which any specific Pan default is magnified: it is worth avoiding Pan overblooming where XS detectors do not saturate.

Fig. 3 shows the advantage of the antiblooming structure compared to sensors only equipped with antiblooming structure along the readout register. These simulations have been performed for two types of specular objects:

Fig. 3.

Comparison of antiblooming in the imaging area versus antiblooming along register.

Each picture is an extract of 66 columns x 66 rows. The circular spot due to the camera point spread function is partially covered by a thin vertical segment without antiblooming.

Each surface has a reflectivity equal to 0.08, which leads to an equivalent radiance level of 900000 W/m2/sr/μm for standard viewing conditions in the panchromatic band.

The TDI hypothesis are: 16 useful TDI lines, 9 hidden lines, charge capacity Qsat=120ke.

The typical length of an overblooming artifact is 40 rows and can therefore hide many useful details.

An architecture with an antiblooming structure in each elementary detection site of the TDI array cancels this drawback. This need was an important design driver since it was requested to integrate this structure in each elementary detecting pixel without any loss of quantum efficiency.

4.1

TDI mode optimisation

The design of CCDs for TDI mode operation requires some specific architectural considerations over and above those for linear or area imaging arrays. In order to achieve the good control necessary over the parallel charge transfer to accurately match the optical image motion, and to keep the phase consistent along the length of the detector, a specific transfer electrode bus structure was used. This has interconnection of the electrodes of each phase between lines within the image area, allowing the parallel transfer electrode bus structure to be sited along one long side of the detector, hugely reducing the time constant and phase delays associated with driving such a device from the short sides.

For accurate matching of the optical and charge images on the detector the pixel was designed with four parallel transfer phases. Various clocking schemes are then available, but for the electro-optical characterisation, charge is moved from phase to phase within the pixels at appropriate periods within the line time (intra-pixel matching), and the serial readout registers are operated near continuously to minimise the pixel readout rate from the output amplifiers. This makes it necessary to have a separate transfer gate electrode between the parallel and serial registers, allowing the charge from the parallel register to be transferred very rapidly into the serial register, and decoupling the needs of the TDI operation for charge positioning and timing from the readout operation.

4.2

TDI length selection

The average device illumination is anticipated to vary during the mission, and therefore to optimise the tradeoff between the integration time, the platform stability and the maximum charge capacity of the pixels the device has been designed with the ability to select the number of active TDI lines in the image area. The maximum number of lines is 20, with selectable reduction to 16, 13, 10 or 7 lines electronically during flight. Unwanted photocharge generated in the unused part of the image area is removed via the antiblooming structure to prevent any potential corruption of the image in the operational part of the device.

4.3

Design for Back-illumination

In common with most e2v imagers, the PLEIADES TDI sensor is designed for back-illumination. The active device wafer is mounted onto a silicon support wafer, and the substrate material removed to leave the required active thickness of epitaxial silicon. The back surface (now the optical input surface) is then passivated by ion implantation and laser annealing, and an antireflection coating optimised for the panchromatic wavelength range is added. Finally, an opaque light shield is added to the devices to prevent optical input to all but the imaging TDI lines of pixels. The wafers are then sawn and packaged.

The final thickness of the active silicon was selected after a trade-off analysis between the MTF, the red quantum efficiency, and the spectral photoresponse dispersion. Several wafers were produced with a range of active silicon thicknesses, and the parameters measured to confirm the outcome of the trade-off. An example of these results is shown in Fig. 6.

Fig. 5.

The packaged CCD

Fig. 6.

MTF and QE trade-off vs thickness

The antiblooming structure used was a ‘shielded’ drain design. This structure is ideal for back-illuminated devices since the surface drain is ‘shielded’ by a deeper diffused doping region, which produces a potential that prevents photogenerated charge from being attracted to the drain from the bulk of the active silicon. This means that there is virtually no loss of quantum efficiency with the addition of the antiblooming functionality.

4.4

Readout register and output circuit design

The constraints on the pixel rate per output, the line time and the number of pixels per line have resulted in a device design with 10 readout registers, each ending in a single output amplifier. The readout registers are of standard two phase design, with a single separately connected last electrode at the output end of each to allow optimisation of the output timing without the load of the whole register phase. The output circuits are of conventional two stage source follower design, DC coupled with common drain connections, loads for the first source follower transistors are on chip, and the second stage loads are off-chip.

The nominal output pixel rate per output is 6.6MHz, but the bandwidth of the outputs has been designed to accommodate significantly faster settling than would normally be required for this rate. This allows margin in the electronic drive and readout electronics timing for drifts due to ageing and radiation effects in the mission without adversely affecting the instrument performance.

4.5

Package design

The package is a co-fired Aluminium Nitride (AlN) construction, with bottom brazed Pin Grid Array(PGA) interface. Fig. 5 shows a picture of the device in its package, with the window removed for clarity.

In order to achieve the demanding geometrical and flatness tolerances required by the focal plane construction, the reference and mounting surface of the packages have been precision ground. Image plane flatness of a few microns is routinely achieved.

5.1

MTF and Quantum Efficiency

Measurements to support the trade-off between the long wavelength quantum efficiency and the short wavelength MTF were performed on a series of devices produced with varying active Silicon thicknesses. The results are shown in Fig. 6.

5.2

Charge to voltage conversion

The charge to voltage conversion factor of the 10 output amplifiers on the device average approximately 2.5μV/electron, with a standard deviation of ~1%.

5.3

Video signal waveform

A typical video signal waveform from the output amplifier operating at 6.6MHz with moderate illumination is shown in Fig. 7. It can be seen that the waveform is well settled in the reference and signal portions, and that there are significant margins for electronic sampling, suggesting that the device could be operated with good performance at higher frequencies.

Fig. 7.

Typical output waveform

5.4

Noise in darkness

The noise in darkness for operation with 20 TDI lines (maximum length), at a temperature of +15°C, with a line time of 106μs is ~130μV rms, with a dispersion of ~5% across the 10 registers.

5.5

Dark signal and non-uniformity

The dark signal for operation with 20 TDI lines (maximum length), at a temperature of +15°C, with a line time of 106μs is ~140μV, with a dispersion of ~3.5% across the 10 registers. The uniformity of the dark signal within each register is ~+/- 35μV, well below the noise level.

5.6

Charge capacity

The charge capacity of the pixels in the image area is a key parameter for the signal to noise performance of the instrument.

It is specified in terms of the lowest capacity pixel addressed by a given readout register (Qsat min), and the highest capacity pixel of the same register (Qsat max) giving therefore the lowest performance that can be achieved by the device, and the highest signal that will need to be handled by the output circuit and the video chain electronics. Fig. 8. shows the result for a prototype device.

Fig. 8.

Charge capacity