FR2660821A1 - Photosensitive charge-coupled array suitable for progressive scanning, and its method of manufacture - Google Patents

Abstract

The photosensitive array with n lines and p columns of photosites is such that CCD-type column transfer channels are covered by line electrodes, widened vertically in the region of the transfer channels, and in which the charges from the associated photosites are transferred by line addressing, line by line. For charge transfer from the photosensitive site to the output register, a voltage wave, extending over several lines, is shifted by means of a delay line, the output connections of which are linked to the line electrodes. The invention applies especially to image capture in high-definition television, in which the lines of the photosensitive arrays are read successively, with progressive scanning.

Description

Adaptive multi-sensitive charge transfer matrix
progressive scanning and its manufacturing process
The invention relates to photo sensors with matrix charge transfer, and more particularly relates to a photosensitive matrix with charge transfer which is particularly suitable for progressive scanning.

 Conventionally, photo matrices sensitive to charge transfer are organized to operate either according to a so-called "frame" transfer mode according to which the charge transfer to the photosites is carried out simultaneously for the entire frame, or according to a so-called transfer mode "line" according to which the loads are transferred line by line. In both cases, the transfer uses the fact that the successive lines constituting an image are interlaced in successive frames, and that there therefore exists, between two lines of the same frame read successively, a line of photo-elements not used for this frame. Thus, for example for a line transfer, the sensitive photosites for two successive lines are separated, but the size of a stage of the register necessary for the transfer is two photosites.

 For high definition television image taking, according to which the matrix of photo-elements comprises for example 1080 lines and 1920 columns of photo-elements, the scanning must no longer be interlaced but progressive that is to say that the successive lines of photo-elements constituting the matrix must all be read successively. For this, the currently known matrix structures are not suitable.

Indeed for the definition of the vertical charge transfer registers, intended for the transfer of the charges from the photo-elements to the horizontal output register, the pitch of the vertical registers for a progressive scanning must be equal to the pitch of the photo-element lines , unlike interlaced scanning where the step is double as indicated above, that is to say where the same transfer stage is used for two lines of photo-elements.

 When access to the charges formed at the pixel level, that is to say of the image element, is not obtained by a transfer of charges but in another way, charging or discharging of a connection for example, or reading in voltage, progressive analysis of the image is possible; this is the case, for example, in matrices organized for reading a single selected line, called line transfer or CPD and XY.

However, this possibility is obtained at the expense of performance because the capacity of the necessary intermediate connections, of high value, notably increases the noise level, and this all the more since the length of the connection and the number of photoelement lines are important.

 The subject of the invention is a photosensitive charge transfer matrix, suitable for progressive scanning mainly with a view to obtaining high definition images with low noise level, which charge transfer matrices would not allow. classics.

 The invention uses the property of charge displacement of charge transfer circuits, that is to say of transfer of photosites to the output register by total depletion and uses the selective line addressing of line transfer organizations or XY addressing; but, to allow the transfer of the charges of the successive lines of photo-elements, the invention uses the principle of transfer by a voltage wave.

 According to the invention, a photosensitive charge transfer matrix, suitable for progressive scanning, comprising a set of photosites organized in n rows and p columns, is characterized in that it comprises p vertical charge transfer channels in which the charge transfer is controlled by line electrodes separating the photosite lines and widened at the level of the vertical transfer channels to cover them, and transfer control means comprising a delay line provided with n equidistant outputs coupled to the line electrodes and the input of which receives a transfer wave extending at a given instant over several stages of the delay line.

 The invention also relates to the method of manufacturing such a matrix.

 The invention will be better understood and other characteristics will appear from the following description with reference to the appended figures.

Figure 1 illustrates the principle of reading charges stored under grids by a command by means of a voltage wave
FIG. 2 illustrates the general structure of the charge transfer matrix suitable for progressive scanning according to the invention;
Figures 3a, 3b and 3c show the evolution of the voltages under the line electrodes as a function of time and their position.
Figures 4a, 4b and 4c are explanatory diagrams of the operation of the matrix according to the invention
FIG. 5 illustrates an example of a validation circuit for line addressing by the delay line
FIGS. 6a and 6b show in detail, respectively in plan and in section, the structure of the photosensitive matrix at the level of a photosensitive element.

 According to the invention, the transfer of charges from a photosensitive site to the horizontal output register is carried out in the following manner: a voltage wave is transmitted to the surface of the silicon and drives the electric charges located in a vertical channel of the type "DTC" ("CCD" in English) that is to say charge transfer circuit. The length 1 of this wave in the direction of transfer is several tens of microns, that is to say that it extends over several lines, and its speed in the case of the application envisaged for a photosensitive matrix at high definition is high, on the order of 104 meters per second.

 The propagation of this wave could be controlled using a single grid elongated in the direction of the vertical transfer YY ', this grid forming a line RC with distributed resistance and capacity. However, since the attenuation of the longitudinal electric field which controls the transfer is far too great given the length of the grid, this single elongated grid is in practice replaced by a series of electrodes distributed over the transfer zone, a electrode being arranged on each line of photo-elements, and these electrodes being controlled using a digital delay line, disposed laterally to the matrix> and in which the transfer control wave propagates.

FIG. 1 illustrates the principle of reading the charges stored under the photosensitive sites of the successive lines, the electrodes E1, E2 ... E i ... E n being associated with the lines of photoelements and respectively connected to the delay outputs 0,, 2t, (n-1) t of the digital delay line with nl stages,
R2, R3 ... R., Rn
As illustrated in FIG. 1, the length l of the voltage wave applied to the input of the delay line is such that it carries the potential of several successive electrodes at all times, in FIG. 1 three electrodes, at a positive V potential.

 Taking into account the charge represented by each electrode and the transit time of the wave in the delay line, of the order of ins per stage, interface circuits are arranged between each stage of the delay line and the line electrodes E1 ... Ei ... As, as shown in FIG. 2 which illustrates the structure of the matrix with organization of the vertical transfer by voltage wave. The digital delay line LAR, the first stage of which receives the voltage wave, has its output sockets connected to associated interface circuits CI. For the selection of photosite lines, a RADL line address register has its different stages also connected to the line electrodes E. via the interface circuits, CI.

ii
Each photosite consists of a reverse polarized diode controlled by a MOS transistor, the gate of which is connected to the corresponding horizontal electrode E i, while the other electrode is connected to a vertical channel (column) of associated transfer, hatched in FIG. 2. The outputs of the vertical transfer channels are connected to the corresponding stages of the horizontal output register RH, the last stage of which is connected to an output amplifier A, supplying the voltage read Vs.

The length l of the voltage wave, i.e. the duration
T of the pulse generating this wave, in relation to the rise and fall times at the output of the interface circuits, and the propagation time): across a stage of the delay line LAR define the number N d 'electrodes under which the charges N = T / t are distributed. The wave can extend over a hundred lines.

 The transfer efficiency is equivalent to that of a register with volume charge transfer, insofar as the speed of propagation of the wave remains lower than the speed of movement of the electrons under the effect of the longitudinal field produced in silicon through this wave.

For a high definition matrix of 1 inch format, (image height = 7mm), the average speed of the charges is VMOY = 0.7x106 cm / s. For the charges to effectively follow the movement of the wave, the longitudinal electric field must be greater than Emin such that
Emin VMoy / ll = 0.7x106 / 1350 = 0.05 V / pin.

 If we assume that the wave provides a transition of 10 volts in silicon, this transition can be distributed over a maximum distance DMAX = 10 / 0.05 = 200 m. For a matrix comprising 1080 lines with the step of 7 m, the time allowed for the vertical transfer of a line is approximately Ces and implies an elementary delay = 1, us / 1080, that is to say approximately lns.

 The switching time Tf of a line will therefore be less than 200pin x / 7pins, ie Tf <28ns. This time is compatible with the integration of interface circuits, the charge represented by a line electrode being less than 10pu.

The operation of the matrix shown diagrammatically in FIG. 2 is as follows: the photoelement consists of the reverse biased diode, photosensitive or connected to a photosensitive layer. The transfer of charges from the photoelement to the column channel is obtained by opening the MOS transistor, the gate of which, line electrode E., also ensures the transfer of charges in the column channel associated with this photoelement. on a line of photoelements is as follows: after the reading period of the previous line, at the start of the line blanking interval, a first pulse I1 is sent in the delay line
LAR so that the corresponding wave causes parasitic loads (due to over-lighting "blooming", or due to the fact that the reading registers can be photosensitive "smearing" or due to the current of darkness) towards the drains VD located at the ends columns. This first pulse I1 also has the effect of resetting to zero the line of photoelements which has just been read. The address register register RADL which comprises a "1" in the stage associated with the line which has just been read is then shifted by one line and a second pulse 12 is sent to the delay line. The passage of this wave at the level of the new addressed line creates on it an overvoltage used for the opening of the MOS transistor of the pixel and the transfer of the payloads in the column channel. These charges are then entrained by this second wave towards the end of the channel where they are directed via the horizontal register towards the signal output.

 Figures 3a, 3b and 3c show the evolution of the voltages on different electrodes as a function of time and their position.

 In FIG. 3a, the electrode E. addressed for reading the line L. sees its potential V increased until, as illustrated in FIG. 3b, the corresponding MOS transistor is open and the stored charges are therefore transferred to the column channel. Then, as the wave propagates, the charges are entrained in the column channel by the tension wave as illustrated in FIG. 3c. So that the charges do not spread over the entire height of the corresponding column before the arrival of the voltage wave, it is advantageous for the wave to reach the level of the line addressed at the time of addressing.

 In order to correctly carry out the transfer of the charges in the channel, it is important that the voltage pulse necessary for the reading of the photoelement is completed before the end of the passage of the second wave; this may require validation of the line addressing by the signal which traverses the delay line LAR.

 A possible embodiment of the corresponding device is shown in FIG. 5 where the output of a stage of row i, of the address register register RADL is connected to the input of a door AND, the other input of the door AND being connected to the stage of rank in of the delay line. The output of the stage i of the delay line LAR (point A) and the output of the AND gate (point B) are connected to the inputs of an adder circuit whose output is connected to the electrode E. The corresponding voltages at A, B and E. are also shown in this figure 5.

Figures 6a and 6b respectively illustrate in plan and in section the possible structure of an elementary point of such a matrix, described in more detail below in connection with its manufacturing process.
In a first step, on a type P bare substrate are formed by type N implantation of the channels Cv for the vertical transfer in volume by offset.

 In a second step, the two silicon grid levels are produced on the substrate to form the horizontal electrodes E .. the electrodes of odd rows being formed on the first level of silicon Sil and the electrodes of even rows being formed on the second level of silicon Si2.

 These electrodes are continuous in the horizontal direction, as shown in FIGS. 2 and Ga and extend around the elementary photosites, partially overlapping each other to fully control the vertical transfer channel Tv In the central zones limited by these electrodes, E. and E.

photodiodes are then produced, by N type implantation
The next step is to deposit aluminum electrodes on the electrodes E. to double them by making a contact between the silicon electrode and the aluminum electrode, this in order to reduce the series resistance which would be introduced with a silicon electrode alone. To make this contact without reaching the substrate, because the electrodes E. are close to the substrate, it may be necessary to deposit, before the silicon deposits and even before implanting N for the vertical channels, a thick oxide in areas intended for silicon / aluminum contacts. In FIG. 6a, this thick oxide has been represented under the contact zone, and a little wider than it.

 This thick oxide also acts as a barrier to prevent the dispersion of charges from a photosite to a neighboring photosite.

 To locate the charges even better, it is possible to deposit the thick oxide also on the area of the substrate not covered by the electrodes and not implanted N +, that is to say around the photosite, or to perform in the same zone an implantation P as illustrated in FIG. 6b which is a section along AA 'of the elementary area, that is to say a section at the photosite level. Such an implantation can be done without masking on the assembly, after the deposition and etching of the silicon electrodes. It does not affect the photodiodes and makes it possible to control the implantation in the areas required to properly locate the charges and isolate the diode.

 The invention is not limited to the embodiment precisely described in particular with regard to the production of the matrix.

Claims (7)

 1. Photo matrix sensitive to charge transfer, suitable for progressive scanning, comprising a set of photosites organized in n rows and p columns, characterized in that it comprises p vertical charge transfer channels in which the charge transfer is controlled by line electrodes (Ei) separating the photosite lines and widened at the level of the vertical transfer channels (CVj) to cover them, and transfer control means comprising a delay line (LAR) provided with n equidistant coupled outputs to the line electrodes (E.) and the input of which receives a transfer wave extending at a given instant over several stages (R.) of the delay line.  2. Photosensitive matrix according to claim 1, characterized in that it further comprises line addressing means (RADL) also coupled to the line electrodes for controlling the reading of the n lines of the matrix successively.  3. Matrix according to claim 2, characterized in that the outputs of the delay line and the outputs of the same ranks of the addressing means are connected to the inputs of interface circuits (CIi) whose outputs are connected to the electrodes of corresponding row lines (ex).  4. Matrix according to any one of claims 1 to 3, characterized in that each photosite of the matrix comprises a photodiode coupled to a MOS transistor whose gate is connected to the electrode (Ei) associated with the line to which belongs to this photosite, the last electrode of the transistor being connected to the vertical transfer channel (CVj) associated with the columns (j) to which this photosite belongs.  5. Method of manufacturing a photosensitive matrix according to any one of claims 1 to 4, characterized in that it consists, from a bare semiconductor substrate  - in a first step, to form the vertical transfer channels (Cv) by implantation in silicon,  - in a second step, to form on the substrate two levels of silicon line electrodes (Si1 and Si2) extending vertically over the transfer channel zones, the electrodes of odd rank being formed on the first level and the electrodes of even rank on the second level, or vice versa,  - in a third step, to implant the photodiodes in the central areas limited by the electrodes (ex).  6. Method according to claim 5, characterized in that aluminum electrodes are deposited after deposition of an insulating layer above the silicon electrodes; contact zones being provided between superimposed electrodes, to reduce the series resistance of the electrodes (E.).  7. Method according to claim 6, characterized in that a thick oxide is produced around the photosites to locate the charges and isolate the photosensitive diode.