Wehner`s Photoelectric CCD

Charles Douglas Wehner

26 January 2004

It is not generally known that Albert Einstein did not get the Nobel Prize for Relativity. He received the 1921 prize in 1922 for work that he did between 1905 and 1911 on the Photoelectric Effect. Max Planck had shown that each colour of light had its own energy. Planck made it possible to translate the colour into any desired units. So if we imagine the smallest-possible ray of light striking an electron, it would move that electron as if a voltage were moving that electron. So on the fringe of infra-red one would expect about 1.7 volts, and the voltage rises throughout the visible band until it becomes about 3.2 on the fringe of ultra-violet. The constant of proportionality that makes such calculations possible, h, became known as Planck`s Constant. It is proportional to the slope of the sloping line above. Einstein discovered that every metal has its own "paste" (p) by which the electrons are held in place. You have to pay (p) the price (p) of liberating those electrons. The alkali metals lithium Li, sodium Na, potassium K, and so on have the lowest voltages, generally about 2V. So if you have light on the fringe of ultraviolet, of 3.2eV, and a metal whose electrons are pasted in with 2V, the light will liberate electrons with an energy of 1.2V; the amount left over. We speak of "eV", of electron-volts, because light does not have a voltage. The electron-volts are the voltage of the electrons after they were struck by that light. The discovery that electrons are pasted into the atoms with different strengths of paste (p) led to a flurry of research to decipher the inner workings of atoms. After ten years, the Nobel Prize Committee had decided to award the prize to Einstein for discovering the Work Function, p. Photoelectrics had arrived. It became possible by careful choice of work-function to create an Einstein photoelectric cell that would see in the dark. It became possible to record sound as a stripe of varying density or width on a film. Cinema audiences could listen to sound delivered by a photocell. James Logie-Baird (nowadays known as John) conceived the idea of scanning an image onto such a photocell in the manner of Nipkow, and created the first television transmitters. Isaac Schoenberg of "Electrical and Musical Instruments" (EMI) made the first electronic television, with a TV tube - the Iconoscope - from Vladimir Zworykin. Then came the Vidicon, the image tubes such as the Image Orthicon, and the Plumbicon. Everybody had forgotten Einstein whilst they rushed to commercialise the spin-off from his work. If you mentioned Einstein, the knee-jerk reaction was to say "RELATIVITY".

In its simplest conceptualisation, the Vidicon has a screen of beads that are made of metal, metal-oxide, metal-carbonate or other material of low work-function. The front of that screen is free to "leak" electrically to the back. An anode is evaporated onto the front glass, in the form of an electrically conductive film. This anode is to receive the electrons emitted from the photosensitive screen. A grid is placed at the rear of the screen, and given a voltage that will slow down the electrons from the cathode gun, so that they alight with very little force. By scanning from the rear, therefore, one can measure the charge on any particular bead - which will be the result of exposure to light. There is even a photomultiplier version, in which the return beam from the cathode gun is amplified with very little noise contribution. When building a colour television camera, it was usual to take three tubes of this kind and to deliver an identical picture to each. However, the image-forming light was passed through a red, a green and a blue filter before arrival, so that each tube gave a signal for its own primary colour. This is known as "Additive colour photography", or RGB. Red and blue are widely separated, so it is the response of the green band that is critical. With dyed gelatine, dyed plastic or coloured glass, the response will typically be Gaussian. That is to say, unwanted light is absorbed and wasted. The absoption diminishes towards the peak, and then increases again. A Gaussian filter can never have an efficiency of 100%. First Invention WCB Photography The author was working as a technical author, design engineer and factory manager in photoelectrics, fully aware of Einstein`s contribution to the field. At that time, in the nineteen-seventies, it became clear to him that something approaching the "Holy Grail" of optical engineers could be reached - the "Brick-wall" filter, or square passband. The arrival of photoresistive and photoionic materials had complicated the science. It was - and is - customary to lump all these terms together under the heading "PHOTOELECTRICS", but this is not generally correct. The Image tubes relied upon CHARGE STORAGE on the photosensitive screen, or on a neighbouring screen to which the image is transferred. If storage effects could be minimised, something remarkable would be possible.

Where the photosensitive beads are truly photoelectric, and where the capacitance of those beads is minimum, it becomes possible to push back electrons due to the lower frequencies of light - the reds and yellows - so that the tube would see only high-frequency (high-energy) light. This would involve the incorporation of a colour threshold grid or coating, but in many instances an EXISTING vidicon could be used, by combining the threshold grid with the anode. The author dubbed the device the Einstein Tube, a COLOUR image sensor constructed from a SINGLE MONOCHROME one by the direct application of Einstein`s Photoelectric Effect. With no voltage on the grid, the tube sees WHITE (red plus green plus blue). When a voltage of about MINUS 0.7V is applied to the threshold terminal, the cut-off point shifts into the yellow-green and red is MISSING. When about MINUS 1.1V are applied, green is also missing, leaving only the blue. There were no frame stores in the early Seventies - they were just being invented. However, it was possible to extract the colour as a SUBCARRIER.

The differential of that cosine gives -sine. We see that it starts with a loss of red - an increase in cyan. After the minimum, we go from a lack of green (magenta) to an excess of green. Thus, by chopping the waveform in quadrature at TWICE its frequency, we can obtain the red-cyan and the green-magenta vectors. This gives us more than enough data for conversion to NTSC, PAL or SECAM. And what of the GAMMA, or luminance? It is fundamentally as shown in the first waveform - blue mixed with cyan mixed with white. If the subcarrier is asynchronous to the scanning (2.58 megaHerz in the States, 4.433 mHz in Europe), the luminance will look like that of a monochrome tube balanced for tungsten. Chroma interactions will be invisible. The particular delight of the concept is that the device can be constructed as a general-purpose one. There is no specific NTSC. PAL or SECAM concept designed into it. Thus, such a tube can be used for differing line-standards and different subcarrier frequencies. In those days there was no digital photography, but the author was actively involved in wet-process colour developing. That process uses the secondary colours yellow, magenta and cyan, and is known as "Subtractive colour photography", so that WCB can be seen to be an amalgam of existing techniques. Second Invention Virtual Filters With perhaps half a million dots on a television screen, it becomes very limiting to have to read each pixel in real time. Image tubes, with their storage effect, were invented precisely to raise the sensitivity by integrating the light over the period of a full frame. Where time is not a factor, such as in astronomical photography, it is perfectly possible to use frame-sequential colour techniques. An image is made by white light and stored. Then a cyan image is made and stored, and finally the blue. Storage in those days could have been on a photographic plate. There are techniques by which one image can be subtracted from the other - for example, by sandwiching a positive and a negative image together. The important point is that the absence of haste means that we no longer need to use sinusoidal waveforms. We can apply a colour-threshold voltage, wait for it to settle and take a picture before moving on to the next. Very exact control of the voltages becomes possible, and the user has a free choice of threshold within the available band.

The first waveform, WCBCWC, is preferable to the second, WCBWCB, because at higher frequencies they will become somewhat rounded. A glitch may ensue in the second waveform in some systems, as the threshold transits via cyan from blue to white. The essence of virtual filters is that the edges of the colour passbands may be held as voltages either by analogue or digital means. Semiconductors We have been looking at Cathode-ray tubes. It is noteworthy that there is no such thing as an Anode-ray tube. We know that there is only one type of electricity - the electron. So if we put positive onto the cathode of such a tube, no current will pass. This cosy world of given truths was disrupted at Bell Telephone Laboratories by John Bardeen, Walter Brattain and William Shockley. Whilst researching the properties of rectifiers, they found that the red oxide of copper responded to electrical fields. The oxide film would behave like metal - like a conductor. Under the influence of a field, it would drift towards the behaviour of glass - an insulator. It was a kind of transition resistor. The word transistor was being coined. They found that temperature and other influences limited the usefulness of the device. Used as an amplifier, it would hiss. They also found that it was not just an electric field, but also an electric current that would modify the behaviour of the material. A pair of fine needles were touched against the film, and current through one needle would alter the current through the other. This was the POINT-CONTACT TRANSISTOR. Other materials were found to display this behaviour - germanium, silicon and then the gallium compounds, for example. It was as if there were TWO kinds of electricity, positive and negative. The theory said that these materials would pass only ONE of the two kinds. A P-Type material would not let negative through. However, the equivalent of letting negative through would be to let positive go through in the reverse direction. Similarly, an N-Type material would not let positive through. But why the fine distinction? Does it matter whether positive is going from right to left, or negative from left to right? It matters if we want to explain the field effect. If only positive CURRENT CARRIERS can get through, and if a positive field pushes these away, then NOTHING can get through. Such a material, which can use only HALF of the available ways of conducting, is a halfconductor or SEMICONDUCTOR. Most modern semiconductors are made from ultra-pure silicon, by diffusing or implanting parts per million of P-type or N-type DOPANT. There are not too many other examples of electricity existing in two forms. There is the question of light itself, which some people consider to be made of electrons and POSITRONS. They explain its absense of mass by saying that positrons are anti-matter. These theories are too extreme for most electronics engineers, who prefer to think of positive electricity as HOLES - an absense of electrons. The FET Here we have a block of pure silicon which has a CHANNEL of P-type dopant implanted along the top. Connections are made to each end of the block. Then an insulating layer of silicon dioxide is created (PASSIVATION), and some metal is evaporated onto that. The device is a P-channel FIELD-EFFECT TRANSISTOR. If a negative voltage is put onto the metal film at the top, it will attract the positive MAJORITY CARRIERS, and the device will conduct along its channel. One end might be designated the SOURCE of any current that can now flow, the other becomes the DRAIN into which that current flows, and the isolated metal film, for obvious reasons, is called the GATE. The CCD It is possible to divide the gate into several pieces. If we do that, the conductive areas underneath them can be split into separate units. We will call these units WELLS. This does not stop up making the entire channel conductive. The multi-gate field-effect transistor is, in fact, the basis of the charge-coupled device. The trick of setting all gates negative will be used to create a RESET feature to sweep previous data out of the device in readiness for an exposure.

	Taking the image we have seen before, note the pattern of +--+--+--
	Now by means of ++-++-++- we narrow the wells. The majority carriers are obliged to migrate right to the narrower wells.
	By means of -+--+--+- we widen the wells towards the right. The charges in the wells spread out towards the right. Notice how a well has been formed on the extreme left, containing the reset voltage.
	The next step narrows the wells on the left again.
	And again the wells are widened to the right.
	And again narrowed on the left.
	The seventh stage brings us to where we were at the beginning, except that all the charges have been moved one place right, and the first well is reset.

We could, of course, reverse the sequence. If we now walked the wells off to the left by means of steps 6,5,4,3,2 and 1, we would be back to where we started from - apart, of course, for the right-most well. The charges in the wells can be moved about, regardless of what they are, by virtue of their coupling with the gate voltages. That is the meaning of the term. CCD has become synonymous with optical image sensors - but the concept may be used for other purposes such as the analogue delay-line used in echo units for sound. The gates are for obvious reasons known as SHIFT GATES. If light enters the silicon, it generally causes the photoionic effect to deliver an ELECTRON-HOLE PAIR. N-type silicon is deliberately introduced into the device to capture the electron. This leaves a positive charge in the well. The unfortunate side-effect of this is that infra-red light from the environment may also create electron-hole pairs indistinguishable from the wanted image. This raises the DARK CURRENT, and can make the image blotchy due to NOISE. In the photoelectric CCD, the author proposes to use P-type silicon only. Electrons released by environmental radiant heat will be obliged to wander about within the silcon until they lose energy and recombine with their respective holes. Only in the analogue-to-digital converters and logic circuits (far from the sensors) will N-type be used. Einstein photoelectric cells will be wired directly into the charge-coupled device. Such connections require a metal-to-silicon junction. It is well known in engineering circles that every such junction forms a Schottky boundary. Although many Schottky diodes have a very low threshold voltage, it has to be accepted that this is one of the critical points in the design. The lowest threshold and the sharpest diode-action are desired. The threshold has a temperature coefficient and a dependency upon the current passed. Thus, the stability of the photocell voltage depends on good design. Similarly, if the diode exhibits reverse leakage the behaviour of the simpler CCD design will be impaired. The action of light is to release electrons from the photocell plate, which are conducted away by the field plate. In this diagram, we have green glass, but quartz or any other transparent material will be suitable. A vacuum in the intervening space is desirable as we will see.

The absolute simplest concept is a chip of neutral silicon or neutral-doped silicon implanted with deep channels of P-type dopant. These channels are shown in pink. A single channel at the far right connects the rows. The top right-hand corner is shown in an unfinished state because it is here that the analogue-to-digital converter will be fabricated.

It is quite usual to use gold for the interconnections because of its stability and low resistivity. The quantities employed are small. What we see here are the lateral shift-gates laid down as vertical stripes. These are on top of the passivation, except for the left-most which is in contact with the silicon, to act as a reset. There is also a single vertical column of shift-gates laid down as horizontal stripes. In addition, we see the spots of metal that pass through the passivation and make contact with the P-type silicon. Aluminium is a common material for this purpose. They form the Schottky junctions, so a careful choice of material will reap benefits. A further layer of passivation is laid down, and the Schottky junctions are connected through this to the photocell plates.

The choice of material for the photocell plates is critical. What is needed is a material that can withstand brief periods of exposure to heat, such as when the sun is shone onto the cell array by the lens acting as a burning-glass. One of the best materials for this purpose is platinum, which can be laid down as a thin film by evaporation and can be rendered almost jet black by sputtering. A very dark photocell is desired, as this raises the quantum efficiency - but for obvious reasons, black paint simply will not do. The light must not only be absorbed but changed into electricity. Unfortunately, platinum has the highest of all work functions and is therefore useless on its own. Materials such as potassium have lowish work functions, but are chemically too reactive. There is NO ready-made metal that has a work function that will respond to red. ALL metals have a cut-off at least in the yellow. Fortunately, vast amounts of research into work functions have led to a choice of coatings. Barium oxide or carbonate may be used, and there are suboxides of nickel that are reputed to bring the work function of a photocell down to between 1 and 1.5 volts. Photographic film makers often extended the red response into the infra-red in order to boast a higher sensitivity without increased grain. If the infra-red response were easy to obtain with Einstein`s photoelectric effect, this would be an option. Unfortunately, it is in the red/infra-red region that we have a critical problem. The red cutoff may have to be chosen to be at 1.8 or 1.9 electron volts if a suitable coating cannot be found to bring it lower. The Schottky voltage, if badly designed, may be as high as 0.4 - but Schottky thresholds of 0.1 volts are quite common. So a worst-case design limit of 1.9eV for the red would need a Schottky threshold of 0.1 or less, and a work function of 1.8 or less. Wartime infra-red imaging systems in the 1940s used a mixture of caesium and rubidium on oxidised silver. Clearly, low work functions can be achieved. A report in September 1970 from Southampton University, UK, describes the work function of caesium alone as being from 1.795 to 1.82 volts when deposited on one form of single crystal tungsten, but 1.975 to 2.07 on another. The geometry of the substrate is therefore important. There have been reports of organic substances such as benzene lowering the work function by a tenth of a volt. Unfortunately, such materials are likely to evaporate. So to obtain the best-possible results, one is heavily dependent upon state-of-the-art research. Third Invention Field-plate Shutter The above design might be used in astronomy. All shift-gates are set to negative, and the reset voltage floods the chip. Then the field-plate is set to the lower limit of its passband, the shift-gates are set to create wells and exposure commences. The moment the exposure is over, a strong negative voltage such as -5V is applied to the field-plate. Electrons due to anything in the visible range are pushed back onto their cells. There is no frame park. None is necessary. Because there is no N-type silicon to form PN photodiodes, radiant heat will not generate any dark current. The only danger is from high-energy radiation, such as from X-ray stars. A mechanical shutter might be needed if this is a problem, but not if it is not. Thus all that remains is to step the data out of the array and into the analogue-to-digital converter. But consider the study of pulsars. These are stars that flash at several hundred times per second. If we are viewing at 25 frames per second, and if there is adequate light, we may wish to strobe those stars. For example, a 400Hz star would flash sixteen times per frame. It is possible to pulse the field-plate at this rate, and so bring the pulsar to rest in order to measure its frequency. Fourth Invention Chroma Shutter The above simple design is heavily dependent upon the diode action of the Schottky junctions. As the wells collapse, the charge has to be moved sideways, not up the connecting wire and back onto the photocell. Most everyday photography is tricolour, so that three sets of wells would be needed. Of course, it is equally possible to create four if one has set Red, Yellow, Green and Blue (RYGB) as ones design objective. The system is soft-centred, and so a change of software will give a change of behaviour. We will consider three as the most commonly desired arrangement: In this arrangement, shallow channels are fabricated to make contact with the deep well-channels. Additional gates, the CHROMA gates, will be deposited onto the passivation above these channels. The result is that we have created Field Effect Transistors, FETs, and the wells are gated wells. The ability of such FETs to isolate the wells from the photocells is much more distinct than that of the Schottky diodes. Indeed, it now becomes possible to use leaky Schottky diodes, and concentrate ONLY on lowering the Schottky threshold. Where the FET is off, the diagram portrays the channel as dark blue. Where the well is ON, but isolated, the diagram portrays it as pale blue, with a plus sign to represent the charge that is stored. These chroma gates can be used not only to separate out the colours, but also for the purpose of strobing, as previously described. This would be particluarly effective is the field-plate and chroma gates strobe together.

	The silvery dots represent the Schottky junctions. These are brought through the first passivation, upon which the shift-gate stripes and chroma-gate squares are laid.
	The silvery Schottky connections and square chroma connections are brought through the second layer of passivation. The chroma gates are now linked by horizontal stripes of conductor.
	The Schottky connections are brought through the third layer of passivation, and the photocells are laid down on top. Each cell makes contact with three Schottky connections, therefore, acting as a bridge.

The rectilinear array of metallisation allows the photocells to be laid out on such a grid as will protect the interior of the chip from light. Even though, as we shall see, the moire effect will not produce colour artefacts, the array should be protected against monochrome ones. This can be achieved by making the cell edges ragged.

It is quite usual to construct an analogue-to-digital converter at the output of the CCD array, because the analogue signal is weak and prone to be corrupted. A set of three would be preferable.

It would be a good idea to allow the user a choice of WCB or RGB, by constructing a binary subtractor between the W and C, and between the C and B. Underflow must be guarded against. A further variant is the construction of larger chips from smaller ones. Four such chips might be joined together, in such a way that the readout takes place simultaneously on each of four quadrants. When used for cinematography, four central processors can be used if desired, each working into its own segment of memory. This would allow a higher frame-rate to be achieved. Fifth Invention Synchronous Detection In the above diagram, the central cells have been shown without ragged edges in order to show their position. In an actual chip they may be as ragged as the others. Consider those cheap keyring novelty red lasers. They can be switched on and off very fast. Light travels at 300 million metres per second, so that in a millionth of a second it would travel about a hundred and fifty yards and return. If switched at 100 thousand times per second or less, the delay becomes insignificant. As the chip is soft-centred, it can be set to maximum bandwidth reception whilst the chroma gates alternate between two sets of wells, one set with the torch on, the other with it off. Then just TWO rows of wells are read out. These are the central line, the WHITE wells, used with the torch on, and the CYAN wells, used as white but with the torch off. The difference between these two readouts will show the pixel or pixels that were lit by the torch. This can form the basis for automatic focusing. The offset from centre of the spot can control both the positioning and speed of the focus motor. With a red laser beam controlled by a swivel mechanism, prism or edge of a lens, the focus-point will be visible to the user with his naked eye or via an optical viewfinder. It can even be incorporated into the liquid-crystal viewfinder. In a film studio, the red spot representing the focusing position can be visible at all times to the director and his crew, but invisible on the video. If the CCD responds to infra-red and an infra-red beam is used, the focusing spot will be invisible, but can be made visible in an LCD display as a synthetic crosshair or other graphic. Naturally, if the spot is not detected it will be deemed to be too far away. The software will set the lens to the hyperfocal appropriate to its aperture, as is usual. Various software tricks, such as to take the central line without spot, and average the brightness for exposure control, may be combined with the autofocus feature. The CCD can multitask between image-taking for an electronic viewfinder and central-area analysis for focus and exposure control. In a video camera, focus and exposure control can of course be running continuously during the filming due to the multitasking. Sixth Invention Chroma Multitasking

	This is our example graphic. The wheels will not be extensively manipulated in order that it will preserve the look of a truck.
	Taking three images in succession would cause the white rear of the truck to have black rather than cyan subtracted, thereby exaggerating the red. Similarly, a green fringe is created.
	If we change the field-plate voltage many times, and swap the active chroma gate as appropriate, the exposure delivers many colour stripes.
	However, each colour stripe is delivered only ONCE, and so becomes diluted by the overall exposure.
	The end result is motion-blur that is free from chroma artefacts.

It can be seen that the effect of chroma multitasking is to make it appear as if the photocells were doing three or more things at once. The sequence WCBC might be set up as a subroutine in software such that it takes a twenty-thousandth of a second. To make this possible, the conductivity of the field-plate must be high and its capacitance low. Similarly, the photocells should have only a fraction of a picoFarad capacitance if possible, for example by thickening their underlying passivation. If a twenty-thousandth of a second can be achieved, a thousandth of a second would have twenty cycles and the chroma error would be 5%. At a hundredth of a second, using two hundred cycles, the chroma error would be half a percent. These figures are quite sensible, and may be improved upon. Multitasking chroma error applies only to areas of motion blur. The Effect of Air Up to now, we have considered the ideal case. It is true that the flight time of the secondary electrons is brief and interactions between them will be few, but those with very little energy will be most vulnerable. The particular danger is that gas molecules in the space between the photocells and the field-plate will reduce the energy of the slowest electrons until they can be turned back to the cells. Gaussian probability plays a role here. The result is that the theoretical square passband becomes rounded into a sigmoid curve that tends towards a Gaussian one. If the field-plate capacitance can be tolerated, the plate should be brought perhaps to within a micron of the photocells and the space evacuated. The physical structure of the field-plate must be able to withstand atmospheric pressure at fourteen pounds per square inch, plus the pressures and accelerations of normal use without making contact with the cells. The Voltage Vector Effect Now comes a happy disappointment. Consider what would have happened if we had achieved a brick-wall effect. A yellow rose in daylight absorbs blue. So the red and green channels would both register the rose and it would appear yellow. However, a sodium lamp with its D-area having bright yellow lines at 0.5889953 and 0.5895923 microns might actuate the red sensitivity or the green sensitivity alone. The lamp would appear to be either bright red or bright green. And the yellow rose when LIT by sodium? Either a red rose or a cabbage! We began by saying that the electron-volt rating of light is the effective movement of an electron when struck by that light. It moves as if a voltage had pushed it. However, the movement is a VECTOR having both speed and direction. Electrons are not actually struck directly by the light. Light is absorbed into the metal, they raise electrons into the conduction band, and after a greater or lesser delay those electrons are flung out. By the laws of probability, some will go DOWNWARDS into the lower hemisphere that is the photocell itself. Indeed, it will be half. That gives us one F-stop loss of sensitivity, but as nothing was wasted up until now, we can afford this. Of the remaining electrons that fly UPWARDS into the hemisphere above the plate, the laws of averages say that we expect as many to arrive at 1 degree to the normal as arrive at 89 degrees to the normal. However, as we previously saw with single-crystal tungsten, surface geometry may bias the outcome somewhat. Generally, though, the average of all electrons will be at 45 degrees to the normal (such as 1 plus 89, divided by 2). The vertical component of their velocity will be Cos A times their speed, the horizontal Sin A times their speed, where A is their angle. So the average "vertical voltage" will be 0.707 times the ideal. But there are TWO axes to the CCD. Accordingly, if the average "vertical voltage" due to the distribution in the X-axis is 0.707, the effect of the distribution in the Y-axis will be to deliver 0.707 of this. The result is that the average upward voltage is only HALF of what we expected - another, and final, loss of an F-stop. So each colour delivers RANDOM electrons that average half the expected vertical voltage. The percentage of these electrons that get turned back onto the photocell declines reasonably linearly with rising light-frequency. With this in mind, it would be a good idea to have a colour response that extends barely into the infra-red, and use a POSITIVE voltage on the field plate for the white exposure. In this way, NOTHING in the gamma (monochrome W) channel is lost.

Einstein`s photoelectric effect allows a choice of negative voltage on the field plate to select the start of the cyan slope.
Similarly, the choice of restraining voltage for the blue exposure determines the onset of blue sensitivity.
We have two degrees of freedom in the control of exposures - field-plate voltage and TIMING. Longer timing increases the amplitude.
Cyan minus blue delivers the green. .
Amplitude adjustment by timing delivers the final green response.
Here the red and blue responses have been inverted, and put on top of the green. The yellow sodium D-area (thin black line) can be seen to deliver data in both the red and green channels.

It can be seen that the absence of abrupt transitions from one colour primary to another delivers an accurate representation even of a single spectral line. Binning The design of this image sensor gives almost 100% coverage of the image plane with photocells. There are no blind spots. It would therefore be a great shame to spoil the performance by crude interpolation of the high-definition image. Consider a dark pavement covered with white cigarette-ends. If the software picks out only every third pixel from every third row, eight pixels out of nine will be blind. Any camera-shake will cause those white dots to vanish and then reappear. This can be very distracting. A similar effect occurs with leaves on trees. Fortunately, because white, cyan and blue are segregated we can use standard binning techniques in one axis. The following diagram makes that axis horizontal, but the device may be swivelled through 90 degrees, so the X and Y axes can be considered interchangeable. The shift-gates are tied together in groups of 18, allowing binning in twos, threes and sixes. Also shown on the left is the tying together of the W, C and B chroma gates.

For binning in the other axis, one needs to devise a means of bypassing the various chroma CCD stripes in order to keep the colours segregated. We will call the device a BRACKET. It consists of a superficial P-type diffusion covered by passivation, and fitted with a gate. It is nothing more than an FET connection. Associated with each bracket is one or more BINNING SHIFT-GATES. These will normally be used for stepping the charges through the CCD, but for binning purposes they can INDEPENDENTLY be made to go positive, so that the wells underneath collapse and the charges are obliged to run through the bracket and concentrate themselves in a single well.

Here we see the situation for an image sensor with just one A to D converter. It is also shown for binning in twos or threes only. For binning in twos, bracket A is driven negative whilst binning shift-gate 1 is driven positive. Then A is driven positive, and 1 becomes negative. The output is read out. The system is stepped to the next colour, and the sequence repeated. Then again, and finally the system is rushed through the eighteen steps that bypass the empty wells. For binning in threes, brackets A and B work together whilst 1 and 2 work together. After the three colours are read out, the system is rushed through thirty-six steps before repeating.

Here we have an example with three A to D converters and binning in twos, threes or sixes. A and 1 work together, or A and B with 1 and 2, or A, B and C with all the binning shift-gates 1 to 5. Data is shifted up to the binning-point and binned. Then, in six steps, the next colour is moved up and binned. Finally, the third colour is moved up and binned. All three colours can be read out together - possibly as RGB. If binned in twos, eighteen steps follow with no readout. If in threes, thirty-six follow. If in sixes, ninety steps follow.

Although we do not need a frame park, a three A-D unit needs a single-pixel park as shown. The P-type channel for the output does not need to be straight. It is possible for it to turn through a right-angle for the last two or three wells, thereby allowing two or four small image sensors to be combined into a larger one - as previously shown. There are two main advantages to binning. Firstly, the distracting effect of the shimmering of cigarette-ends on the pavement and of the leaves on trees is eliminated. Secondly, the sensitivity of the image sensor is increased. In the latter case, binning in sixes allows a video recording to be made at one sixth of the lateral and vertical resolution, but at 36 times higher sensitivity - a gain of over five stops. Advantages The device is easily made with available technology. At the time of writing, there have been 99 years of research into work functions, starting with Einstein himself. There is therefore a wealth of expertise. At the present time, the limit of resolution of mass-produced lenses is 100 lines per millimetre. In a letter to the German government, the author pointed out that this invention would save the chip factory at Frankfurt on the Oder. The idea was to build CCDs to European standards. A DIN A10 chip would be very slightly oversized compared with a 35mm frame, would fit its image to any European paper and with 9.75 megapixels at 10 microns, 10 million pixels at 9.88 microns or 10,485,760 pixels (ten COMPUTER meg) at 9.76 microns cell separation would render 35mm film obsolete. Image detection takes place on the surface - not within the device. Thinning is not required. Unlike with the standard Bayer-pattern photoionic image sensors, red, green and blue do not penetrate to different depths within the wells. There will be extremely low dark-current due to the unipolar, rather than bipolar design. A separate reset circuit is not required. Resetting is accomplished by opening up all shift-gates and chroma-gates. This also discharges any charge held by the photocell capacitance. Quantum efficiency should be close to 25%. That is not some fudged figure. It is the percentage of the light cast upon the cells. In photoionic systems there is considerable loss of light due to the inefficiency of the colour filters - particularly the blue. Manufacturers often conceal this fact when quoting quantum efficiency, dealing with the percentage of light turned into electricity after it has actually entered the silicon. So we expect an increase in sensitivity of several stops. The total coverage of the chip area (with only small gaps for separation) brings a further gain of at least one F-stop. Colour balance can be controlled by means of the "virtual colour filters" - a unique feature - or by the timing, or both. The chip has great flexibility of software control. Anti-blooming diodes may be fitted, but care has to be taken not to jeopardise the intrinsically low dark-current. The colour fringes common with Bayer-pattern CCDs have been eliminated by multitasking. The definition of the image sensor is actually THREE or FOUR times that of the equivalent Bayer-pattern sensor, because the Bayer pattern RGRGRGRG followed by GBGBGBGB takes four pixels to make an image spot, whilst the photocells in this design deliver red, green and blue (via white, cyan blue) from the SAME cell. A Bayer sensor would therefore need thirty or forty million sensitive spots to compete with the abovementioned photoelectric A10. The Future Everything described so far has been realisable with existing technology. But there are concepts that need the boundaries of science to be pushed that little bit further. The first example leads to further miniaturisation of optical equipment. It is described as "FUTURE" because it makes demands on the glassmaker`s art that may just be too exacting. Mass-produced lenses today can just reach 100 lines per millimetre, or 2540 lines per inch. If green light is taken to be half a micron (micro-metre) in wavelength, each line will be about twenty wavelengths wide. If we consider, as is usual, a THIN lens, we can describe its thickness as X. The skew wave at 45 degrees will pass through a thickness of X / cos 45. That is root-two times the distance. Because the top and bottom of the lens can be thought to have zero thickness, the skew section can be imagined as a lens that "bulges" by root-two times X. The focal length is inversely proportional to the "bulge", so the skew focus at 45 degrees is 0.707 times the normal (right-angle) focus. The distance to a flat focal plane, however, is root-two times as far as the normal focal distance, when measured at a skew of 45 degrees. So the lens-maker is fighting against the geometry of nature whilst at the same time having to ensure that all COLOURS come to the same focus. Seventh Invention Focal Plane Curver The concept here is to use optic-fibre techniques to correct the geometrical distortion, leaving the lens-maker to deal with the colour error and perhaps some minor production tolerances on the geometry. Firstly, it needs extremely fine optic fibres. The process of stretching and folding am optic fibre block may snap individual optic fibres internally within that block. It is simply something that has to be tried. Secondly, if a parabolic curve is cut into such a block, the skew beam will meet up with less optic fibres on the side of the parabolic curve than the normal beam will meet. It is that same law again - 0.707. So the block must be TURNED UPSIDE DOWN and flattened again before the final curve is installed. It may take two such stages. A bundle of optic fibres is melted and stretched, folded upon itself, and melted together again. This is repeated until the individual fibres are below 5 microns in width. A preliminary curve is ground into the block of fibres. It can be seen that any ray striking at, say, forty-five degrees, will meet few fibres. The skew image will be reduced in size instead of enlarged. The block is flipped upside-down, so that the error is in the reverse direction. The third step is refiguring. This is a delicate stage that may need exact robotic control of time, temperature and pressure. The optic fibres are splayed out at the base to reverse the miniaturisation. Finally, the lower surface is lapped and polished, and then coated to act as the field-plate. The upper surface is given its final parabolic curve. A major problem may be to create such a field-plane curver without introducing ripple into the glass. The finished device might need only a very simple lens. This could be a Dollond-style cemented achromat - but then the two inner surfaces must match. A better design would be a single airspaced Gauss. This leads to a single degree of freedom in production - the adjustment of that gap - to take up tolerances. It is conceivable that perhaps TWO hundred lines per millimetre might be achieved. The second concept in the "FUTURE" category is so far-fetched as to require a fundamental inventive step. At the start, we dreamt of a brick-wall filter. During our studies we came across the voltage-vector effect as a major impediment. Such deviations from the simplicity of Einstein`s original concept are well known amongst researchers. There are such ideas as the Jellium model to explain the fuzziness of the work-function boundary. So we begin to think of nanotubes in front of the photocells that trap any secondary electrons that are more than, say, one degree from the normal. But these would reduce the sensitivity by 8100 or 13 F-stops. In addition, nanotubes would need an extreme REVERSED TELEPHOTO lens to deliver the light to the photocells. Only near-parallel light would arrive. One begins to think of magnetic focusing methods, and other means to deliver all secondary electrons regardless of their speed normally to the field-plate. There is an answer somewhere....... If achieved, one could take an image that includes the Sodium D region alone, and one that extends deeper into the red. This image might be stars taken by the wider bandwidth method. This image might be the D-region alone. The difference would be the Doppler-shifted stars that do not belong in the same galaxy. This ability to "TUNE IN" to a known part of the spectrum when taking photographs might be called SPECTROSCOPIC PHOTOGRAPHY. The ability to tap a key or turn a knob until one has created an image of stars in a galaxy, and then read off the Doppler shift by virtue of the restraining voltage, might be called SPECTROMETRIC PHOTOGRAPHY. Don`t hold your breath - it won`t be just yet. In the future perhaps..... Finally, there is the idea of returning to the Bayer concept with all its disadvantages. Bayer has a block of colour-sensitive spots: RG GB whilst the new image sensor might have WC BW and you may wonder why! Firstly, the sensitisers exist. We have the caesium-rubidium on oxidised silver for the white. There are materials like lithium (2.4V) or barium (2.5V) for the cyan, and samarium (2.7) or cerium (2.9) for the blue. These materials are simply coated as a monolayer on the cells. Secondly, with a positive field applied to the plate, all secondary electrons will curl upwards - producing the highest quantum efficiency of them all. A strong negative can still be applied after exposure, to turn the CCD into a frame park. Thirdly, if a program is written to apply a "stopping voltage", or "restraining voltage" to the plate, the threshold of all three primaries will rise in unison, and the slopes of the three curves will appear in unison. Fourthly, we are back to the simple design - we do not need the gated wells, with the FETs. Somebody might find a use for such a strange sensor. Criminals have taken control of the U.S. and British governments. Industry - other than the production of weapons of mass destruction - has been allowed to die. Nobody can buy or sell unless they have a secret code behind the forehead, or in the palm of the hand. This situation was forseen. The prediction stated "THEN WILL THE END COME". Surely, via the amygdala effect. The author was viciously tortured by the British government, and had to flee the country. He had lost his house and all his possessions. Under these circumstances, he was in no position to enter into protracted negotiations to sell this set of inventions. Rather than waste knowledge, the author put the information into the public domain. The information is NOT free. It is given to the world to hold in trust. Manufacturers should bear in mind their debt to the author when commercialising this design. He has been reduced to such poverty as to be unable to afford to buy his own product. Therefore, a fund should be held by manufacturers to reimburse the inventor. A photoelectric image sensor should be known as a WEHNER image sensor. A focal-plane curver should be known as a WEHNER curver. HOME (C) 1973-2004 Charles Douglas Wehner. Use freely but do not plagiarise.