1. Introduction

Silicon is the material of choice in current photodetectors. While this mature technology provides very sensitive and versatile photodiodes and phototransistors, it comes at a very high cost per unit area. Therefore, affordable silicon photodetectors always have a small active area. For most applications, elements with an area of a few square millimetres or less are fully adequate. If necessary, they can be connected to suitable focusing optics.

One application in need of large-area photodetectors is an interactive screen that traces the movement of a light spot on its surface. Other possible uses are for security equipment such as “light curtains” [1] or remote operation switches. We show that a surprisingly simple approach to such large-area position-sensitive devices uses a luminescent waveguide with silicon photodiodes attached in a regular pattern.

Extending the active area of a photodiode can be achieved by choosing different semiconductor materials that allow for large-area processing at low costs, such as organic semiconductors [2,3] that can be deposited in printing or low-temperature vacuum evaporation processes. A first example of building position-sensitive photodetectors has recently been presented [4]. However, this technology is still at a developmental stage, and it will take time before large-area organic photodetectors will be available where all issues concerning stability, reliability, large scale production and cost are solved. Until then, an alternative solution would be to find an easy way to increase the light acceptance area of a standard silicon photodetector.

One solution to this problem was found in a different field of application, namely in solar cells. A plate of a transparent plastic material doped with luminescent dyes can be used to harvest light and to channel it to a solar cell much smaller than the area of the plate exposed to the sunlight. This concept is called luminescent concentrator [5,6]. The basic idea is that the dyes inside the plastic plate absorb part of the sunlight and re-emit it as luminescence in random directions. Most of the emitted light is coupled into the planar waveguide modes of the plastic plate, thereby guiding the re-emitted light to the edges, where solar cells are affixed with a good optical contact.

An advantage of this concept is that it not only allows an increase in the active area of the photodetector, but can at the same time also provide information on the position of a localized illumination. If the sides of such a luminescent concentrator are covered fully by photodetectors, the position of a light point exciting the dyes inside the plate can be pinpointed precisely [7,8].

Mounting the photoactive elements at the edges of the concentrator plate might be the optimal configuration for solar cells, but in large-area sensor applications it leads to severe limitations. One point is that in large-area applications, flexibility is always a desirable property as it allows for high volume roll-to-roll production. Additionally, the product can be rolled up in a compact package, adapted to the desired application and is less susceptible to breaking.

To obtain flexibility, a polymer foil instead of a plate doped with dyes has to be used. As the edge of the foil is too thin to be connected firmly to a photodetector, the standard luminescent concentrator configuration cannot be used. To detect light signals hitting the foil, small detectors with a good optical connection attached to its surface may be sufficient to allow applications that require large areas and flexibility.

First approaches to this concept using an elastomer waveguide with four embedded silicon photodiodes have been shown recently [9]. If several such detectors are arranged in a regular pattern, the position of a localized light spot on a large surface area can be determined as the magnitudes of the signals depending on the distances to the detectors.

Here we present a flexible, large-area device made of readily available and inexpensive components that can trace the movement of a light spot across its surface in video speed and with an accuracy better than 10 mm. This comes close to fulfilling the requirements of applications such as interactive projection screens, which are used for computer games and presentations [10]. Further refinements will lead to more sophisticated applications such as positioning tasks in machines, large high-precision light curtains and interactive whiteboards.

2. Experimental

2.1 Setup of the position-sensitive device

A thermally stabilized 210x300x0.125 mm³ PET foil is used as the substrate on which the conducting lines are screen printed using conductive silver ink as shown in Fig. 1a . After printing, the ink is cured at approximately 110°C for about 20 minutes. The lines have a width of 0.5 mm, a thickness of approximately 0.01 mm and a resistance of about 0.4 Ω/cm, which results in a resistance of about 12 Ω along the longest lines.

 

Fig. 1 a Schematic of the device viewed from the top b Schematic of the device viewed from the side c Photograph of the device from the top

Download Full Size | PDF

A sheet of aluminium foil, which is directly connected to the ground lines of the foil and the readout electronics, is glued to the backside of the substrate. This reduces noise and background signals from electromagnetic radiation in the surrounding environment to a minimum level.

The ends of the printed conduction lines, shown at the bottom of Fig. 1a, are connected to a short ribbon cable with 24 leads. 12 of the leads are connected to the signal lines with every other lead being connected to the ground to minimize crosstalk between the signal leads. The ribbon cable is connected to a National Instruments SCB-68 connector box that is read out by a National Instruments DAQ-Card 6062E, which is a single 16 bit, 500 kS/s analog-digital converter (ADC) multiplexed to be able to read out 16 channels consecutively. This card is inserted directly into a laptop computer.

Between the signal and ground lines, SMD photodiodes (Everlight PD15-21C, marked in green in Fig. 1a and 1b) are affixed with epoxy resin, so that they form a square grid of 95x95 mm². The diodes are placed with the active area (~0.6x0.6 mm²) facing up. A 200 kΩ SMD resistor (marked in red in Fig. 1a and 1b) is glued right next to each photodiode. The devices are connected using conducting silver paste in such a way that each photodiode is mounted between a signal line and a common ground line, while the resistors are connected in parallel with one resistor per photodiode. Thus, the photocurrent generated in the diodes is transformed directly into a voltage signal and amplified to a usable magnitude.

A small amount of transparent epoxy resin is placed on the window of all photodiodes. Immediately afterwards, a dye-doped 200x290x0.3 mm3 polycarbonate foil (Bayer Makrofol LISA red) is pressed onto the photodiodes and fixed until the glue sets. The epoxy resin provides a good optical contact between the window of the photodiode and the polycarbonate foil. To avoid light reflection at the edges of the polycarbonate foil, a layer of black paint is applied directly to the edges of the foil.

To render the device mechanically stable, the edges of the substrate and the polycarbonate foil are taped together using black PVC insulating tape. The finished device is around 1mm thick, flexible and bendable with a bending radius of about 5 cm. We identified the connection between the ribbon cable and the substrate foil as the main source of failure, but as a miniaturized readout module could be fastened directly to the device, we do not assume this to be a problem in production.

A photograph of the finished device is shown in Fig. 1c. The vivid orange colour of the polycarbonate foil is striking, but underneath, the conducting lines, the photodiodes and the resistors are visible. At the bottom, the pads with the connections to the ribbon cable can be seen. The active area of the device is 285x190 mm² with a subdivision into 6 sectors of 95x95 mm² each. Each sector is surrounded by four photodiodes.

To use the device as an interactive projection screen, we stretch a sheet of thin paper (~60g/m²) over the surface that transmits 38% of the green laser intensity, but backscatters enough white light to make a projection on its surface possible. The transmitted intensitiy is enough to allow a fast and precise retrieval of the laser spot position.

2.2 Measurements

2.2.1. ADC readout

For many applications, such as a human interface device for computer games, a laser is the ideal input for the position-sensitive large-area sensor. To test the functionality, a 10 mW diode-pumped solid-state laser operating at 532 nm, TTL-modulated with 4 kHz at 50% duty cycle was used. Due to the long switch-on time of the laser emission, the average light power in this operation mode amounts to only about 1.5 mW.

The signals retrieved directly from the ADC measuring one photodiode output with the laser beam hitting the foil at different distances from the photodiode are plotted in the left column of Fig. 2 . In each measurement, 125 points are read out at a rate of 500 kS/s, so the total measurement time is 250 μs – exactly the duration of one period of the TTL modulation. Thus, the full change in the signal due to the switching on and off of the laser is always visible, while much slower signals, such as the 100 Hz flickering of fluorescent lighting, are visible only as a variable offset between different measurements. This allows fast measurements with efficient noise and background suppression without bulky and expensive electronics such as a lock-in amplifier.

 

Fig. 2 Left column: ADC signals taken with the light spot at various distances d from the current photodiode. Right column: corresponding histograms used for determining the amplitude of the signals.

Download Full Size | PDF

To retrieve a signal from these traces, a stable, simple and fast way of determining the amplitude is needed. One of the most efficient algorithms for this purpose is the histogram method, which strongly suppresses amplitude values that are overestimated because of over- or undershooting signals or random noise spikes. In this method, all signal values retrieved during the measurement period are grouped in a histogram with ten equally sized bins between the lowest and the highest signal value. Then the signal difference between the first and the last bin containing more than 10% of the signal values is chosen as the amplitude.

The histograms plotted in the right column of Fig. 2 demonstrate how the method works. By changing the threshold and the number of bins, the method can be tuned easily either to high precision (with the drawback of a higher sensitivity to noise) or to more stability (at the cost of precision). The parameters used in our case are set for high stability to avoid large deviations in the x-y-position retrieval.

2.2.2 Calibration

For x-y-calibration, the laser was scanned over the surface of the device in a square pattern in 9.5 mm steps using a goniometer moving a mirror relative to the laser source. At each point, the four photodiodes of the sector being measured were read out sequentially at an acquisition rate of 500 kS/s, and 125 samples were taken as a data set for each photodiode. This means that one data set of 125 measurement points always covers one period of the laser modulation (0.25 ms). The amplitude between laser on and off was determined according to the histogram method for each data set. Four amplitude values per diode were averaged and saved as the signal of the current diode at the current location.

2.2.3. Video speed operation

The readout sequence is outlined in Fig. 6 . First, all diodes are read out and the signals added together. If the resulting value does not exceed a certain threshold A, the readout is repeated until a sufficiently strong signal occurs. Alternatively, the readout sequence is initiated by checking if any signal exceeds a second threshold B. In such a case, the laser directly hits a diode and therefore the position is known and no further calculations are necessary.

 

Fig. 6 showing the readout scheme with which the position of the light spot is determined at high speeds. (A) and (B) are threshold values that have to be calibrated for the device depending on the exact operating conditions. A video (Media 1) shows the device in operation.

Download Full Size | PDF

In all other cases, the active sector has to be found by comparing the signal magnitudes of groups of photodiodes. To reduce noise, the four diodes of a sector can be read out again and an average value is used for position determination. With the obtained signals the position is recovered using the algorithm described in the results section. Both the readout of the four diodes of the active sector (n₁) and the full readout sequence (n₂) can be repeated several times and the resulting signals averaged to obtain a more precise result.

3. Results and discussion

The working principle of the device presented here can be summarized as a four-step process. First, light enters the polycarbonate foil and, if it has the right photon energy, is absorbed by the luminescent dye incorporated in the plastic. As shown in the black line in Fig. 3a , the foil is highly absorbent between 450 nm and 550 nm. Thus, blue and green light entering the foil is absorbed with an efficiency of more than 90%. In the second step, the absorbed light is re-emitted with a loss in photon energy because of the Stokes shift. This luminescence light is orange yellow with a spectrum as shown in the red line in Fig. 3a.

 

Fig. 3 a Absorbance and Photoluminescence spectrum of the polycarbonate foil doped with luminescent dyes b Sketch of the working principle of the device. The light coupled into the waveguide mode weakens while it spreads in the foil. Thus, more light is coupled out onto photodiode a, which is closer to the light point.

Download Full Size | PDF

In step three, most of the emitted light is coupled into planar waveguide modes of the foil due to the isotropic direction of the emission. The refractive index of polycarbonate is about 1.55, and therefore all light emitted at an angle of less than 50° relative to the surface of the foil is trapped in waveguide modes and channelled along the foil. As the emission spectrum of the dye is significantly red-shifted relative to the absorption and the polycarbonate is highly transparent, the light is absorbed only weakly during its travel along the foil.

Finally, in step four, the light is coupled out of the waveguide at certain points into small photodetectors distributed in a regular pattern over the surface of the foil. This is achieved easily by fixing SMD photodiodes with an active area of 0.3 mm² to the surface using a transparent epoxy glue that provides a good optical coupling between the polycarbonate foil and the transparent plastic window of the photodiode. As all the processes are linearly dependent on the light intensity in the power range used in our experiments, the signal detected by the photodiode directly reflects the amount of light entering the polycarbonate foil.

There is only one additional dependency and it relates to the distance between the point where the light is absorbed by the dye and the photodiode where it is coupled out. One contributing factor is the distribution of light in two dimensions over the area of the waveguide, which leads to a 1/x dependence. The other factor comes from waveguide losses due to scattering and absorption, most strongly by reabsorption in the dyes which then emit the light into a different direction. In a first approximation, these effects lead to an exponential decay with a decay constant α.

In Fig. 3b, the working principle of the foil is summarized. As the light hitting a certain spot loses intensity during its travel in the waveguide, signals of different relative magnitude are measured at photodiodes with different distances to the light spot. Only when the light shines directly on a photodiode, the signal has to be treated as a special case. However, this case can be identified easily because the signal is extremely high only at one photodiode, and therefore the location of the light spot is known.

In the left column of Fig. 2, the signals as retrieved by the ADC are plotted. The traces follow the change of the laser intensity added to a background of about 30-40 mV. This background is caused by the room light and changes with the flickering of the fluorescent tubes at a frequency of 100 Hz. As the whole measurement time is only 0.25 ms, the change appears as a background that varies only slightly during one measurement. To cancel out this background, the amplitude of the signal is determined by using the histogram method shown in the right column of Fig. 2 and described in detail in the experimental section.

The amplitude depends directly on the distance between the light spot and the measured photodiode. A light spot generated by a green LED on the surface of the foil is scanned, and thus four photodiodes arranged at the corners of a square with 95 mm side length are measured. Covering the whole area between the four photodiodes, the LED is moved to 121 points in a square pattern with 9.5 mm between points. Signals resulting from the light spot directly hitting a photodiode have to be treated differently, and the four amplitude values measured in this case are discarded.

This leads to a data set of 117 amplitudes per diode, which can be plotted against the distance between the light spot and the respective diode at the time the amplitude was recorded. The result is shown in Fig. 4 as a lin-log plot. A fit according to the dissipation mechanisms mentioned above has a dependence according to Eq. (1).

 

Fig. 4 Amplitudes of four diode signals versus the distance of the light spot from the photodiodes. The magnitude A derived from the fit is plotted for each graph.

Download Full Size | PDF

Deviations from the fit, especially at larger distances, are due to crosstalk between the signal lines. This is the reason for the increased deviation around 95 mm distance, where the signal is maximized in another photodiode. A further reason for smaller deviations is the anisotropy of the signal coupling to the photodiode because light coming from different directions is not coupled with the same efficiency. The reason for this is the rectangular shape of the photodiode, which leads to different outcoupling properties depending on the relative position between diode and light point. This becomes apparent as a spread of the signal over the whole distance range as could be described by a variation of A within the data set.

The most important parameter for the position recovery is the magnitude of A. As the optical connection between the photodiodes and the polycarbonate foil depends on the amount and distribution of the transparent glue, the magnitude can vary by a factor of about 1.5, in this case between 0.95 Vmm and 1.35 Vmm. Such a variation has an impact on the x-y-recalculation, making calibration by dividing the measured amplitudes by the magnitude value of the respective diode necessary.

To recalculate absolute position values from the amplitudes, a stable and fast algorithm has to be found. The function given in Eq. (1) cannot be inverted analytically by using basic functions, and most approximate solutions are either too complicated for fast computation on standard home computers or very sensitive to small deviations of the amplitude values. An approximate algorithm that is very sensitive to readout noise and therefore not very useful for fast and simple readout electronics has recently been presented for a very similar problem [8]. Here, a different approach using two steps of empirically determined equations is presented that needs very little computational power and is extremely stable against deviations in the signals.

The first and most important step is the combination of the amplitude values of four photodiodes to obtain a projection of the position onto a two dimensional plane. This is done by the algorithm given in Eq. (2), using the amplitudes I1, I2, I3, I4 from the four diodes around the light spot position in a clockwise order starting from the upper left.

If applied to the measurements or calculations of measured values, this algorithm gives a distorted projection of the x-y-plane. The projection of the fit curves according to Eq. (1) is plotted in Fig. 5a.1, that of the measured values in Fig. 5b.1. Again, the basis is a regular grid of 11x11 measurement points in a square of 95 mm side length with the points at its corners omitted.

Even though a rather large error is attached to the measurement points, as can be seen in Fig. 4, this algorithm is very stable because they are contained within a strict boundary. Now the distortion has to be removed and the scale fitted to the real dimensions of the system. Here we present a first attempt to provide an algorithm to remove the distortion relative to the x-y-coordinates. Further optimization will provide much better accuracy. The projection is modified the computations given in Eq. (3).

Using the ideal values from Fig. 5a.1 as plotted in Fig. 5a.2 (black squares), the mean deviation from the real position values (red circles) is around 3 mm. This can be reduced by employing a more refined algorithm to remove the distortion.

The values from the real measurement are not as symmetric around 0 as the values derived from the fits. This is probably due to anisotropic coupling of the light from the waveguide modes to the photodiode. As this asymmetry is similar in all sectors of the device with 12 photodiodes, it can be taken into account once by fitting the parameters of Eq. (3) to obtain the smallest error. Thus, parameters of 0.26 for the x- and 0.23 for the y-direction are determined, which leads to an average deviation of the recovered from the real position of about 8 mm.

 

Fig. 5 a Projection of the fitted ideal amplitudes onto the x-y-plane b Projection of the measured amplitudes onto the x-y-plane

Download Full Size | PDF

While this may at first sound like a rather low precision, it must be taken into account that the device can be scaled up easily to a size of several square meters, where an accuracy of one centimeter in absolute position detection is remarkably good. However, the main advantages of the device and readout are speed and simplicity. One readout cycle takes 0.25 ms, which amounts to 3 ms for reading out all 12 diodes on the A4-sized device and determining the position of the laser beam.

The accuracy can be increased by multiple averaging or by adding more diodes. With 100 diodes, nearly 1 m² could be covered and determining the position would still take, in theory, only about 25 ms using the current modulation frequency. Faster modulation would directly increase readout speed and, simultaneously, further reduce the effect of unwanted low-frequency background signals.

By using different laser modulation frequencies and applying demultiplexing algorithms before the amplitude determination, several input devices can be used and clearly identified simultaneously, allowing for a true multiuser input capability in a human interface device.

A measurement sequence used to trace the position of a laser spot on a 12-diode device is outlined in Fig. 6. Either the whole sequence or parts of it can be repeated to achieve higher stability. If a signal is detected and the light spot is not on a diode, the time of one cycle is given by Eq. (4).

4. Conclusion and outlook

In summary, a large area photodetector device based on luminescent waveguides was designed and fabricated that allows video-speed tracing of the position of a light spot on an A4 area at an accuracy better than 1 cm. The luminescent waveguide is used to extend the active area of small silicon photodetectors, resulting in large-area, light sensitive foils. Due to the modular nature of the device, scaling up to much larger areas is easily possible. The materials and measurement devices used are low-cost and compatible with mass production. Thus, a large-area input device for controlling computer games and presentations projected onto a screen can be fabricated.

Further studies and optimisations are needed to improve the precision and sensitivity of the device. Better algorithms for position recalculation, faster modulation of the light source and adaption of the dyes inside the foil to reduce the amount of background light captured in the foil will further improve the performance of the device and make possible manifold applications. This new approach will probably renew the scientific interest in luminescent waveguides and result in new concepts on how to use the principles of light trapping and light guiding in modern technology.