Digital pixel sensor
The digital pixel sensor (DPS) was developed by Stanford University in California. It is an image sensor that is based on the fundamentals of CMOS sensors , but due to a special scanning method, it has significantly greater dynamics and in many cases a significantly better signal-to-noise ratio than conventional sensors. In addition, frame rates of up to 10,000 images per second can be achieved under suitable lighting conditions.
history
The sensor type was brought to maturity around 1999 and the principle on which it is based was applied for a patent. Stuart Kleinfelder, SukHwan Lim, Xinqiao Liu and Abbas El Gamal were involved in the development. The commercial implementation as a semiconductor and product is in the hands of the Californian company pixim, which mainly focuses on the surveillance cameras . In 2006, for example, the company won the product award in the sensor technology category of the German-language trade journal Elektronik for its sensor model D2500.
construction
The structure of the converter is designed in such a way that it has a downstream counter register in its cells which, when the optimal exposure point is reached, i.e. H. shortly before saturation, can be frozen and read out later. Memory cells which have still not reached the optimum point above a certain counter value are recorded by lowering the comparison voltage as a function of further counter steps. By digitally reading out the counter register, a brightness value is determined for each pixel.
Combined with the method of “correlated double sampling” (CDS), in which double and multiple sampling are used to improve the result, further increases in sampling precision are possible. In addition, this makes it possible to automatically compensate for the variable dark offset that is basically present in image sensors. In particular, the dynamics of the output signal can thus be increased. There is an expansion of the scanning capacity, which can only be expressed to a limited extent as a binary quantity in the form of the number of bits; it can rather be compared with the concept of mantissa and exponent, as developed by Konrad Zuse for computer technology. Some sources still speak of a 17-bit resolution, for example. Comparable regular sensors are often classified with 10 bits, rarely 12 bits or even more, which are simply technical limits due to the noise in the system.
Since DPS can transmit the measurement data non-destructively for evaluation via the scanning already implemented in the direct pixel environment, there are various possibilities for the targeted use of this systematic property. A general improvement with regard to interference effects (such as blooming with CCD sensors , or voltage noise with APS / CMOS ) results from the very small distances between the sensitive element and the comparator that controls the digital memory.