In order to design an electro-optical sensor properly (or any other kind of sensor: capacitive, inductive, with stress gauges, piezoelectric, Hall effect, micro-wave, ultrasonic,...), it is advisable to know the state of the art, i.e. most of the sensors that already exist in the domain of interest, in order to evaluate the possible sources of difficulties, particularly when one wants to introduce a product on a new market : one has to get used to the procedures and norms before introducing a device that may be more complex than usual, which generally means additionnal training and maintenance costs. It is hence necessary for the designer to optimize each phase of its design : operational specifications, principle of operation and architecture, modelling, technical specifications of components and subassemblies, evaluation and test benches, breadboards and prototypes (see design sequences, figure 1 below).
The "operational specifications" of a sensor are a list of its functions and constraints. They are generally defined by the customer, but they may also be established by the designer himself : these specifications must be as precise as possible, about all the missions that are expected from the device, its performances, the environmental conditions of use, hierarchy or priority between functions, when it is a multifunction sensor. Depending upon the sensor complexity, operational specifications may add up to a few pages or correspond to thick documents. Most often, the principle of operation is not mentionned here, neither are its technical specifications, but rather the results to obtain.
Starting from the analysis of the sensor operational specifications, the designer will have to define its principle of operation, i.e. the physical (here, optical) phenomenon which the whole system will rely upon to reach its goals. He must lay down the main parameters and their influence on the expected performance, the characteristics of the optical and electrical signals, the signal processing techniques, and then define the sensor subassemblies. In case several principles of operation and architectures are possible in order to solve a given problem, it is necessary to pursue the analyses far enough (i.e. by taking into account the existing technologies) for the advantages and drawbacks of each solution to appear.
At this level, the architecture of the proposed sensor is converted into a « block-diagram », that describes its « electro-optical chain ».
After defining the architecture of the sensor, the designer starts simulating its performance by means of a « black box » model in which each subassembly is represented by its inputs, issued from the preceding black box, and by its outputs, delivered to the following one, and by its « transfer function ».
Modelling of an electro-optical system is usually separated into two parts : the first one, that goes from the source to the detector (or up to its its preamplifier), must make sure that the electrical output from the detector is strong enough to be processed with success. For that purpose, the model defines the radiometric budget of the sensor, i.e. it evaluates the optical signal, or flux, that is incident upon the detector. This budget ends up by the computation of the signal to noise ratio at the output of the detector, inside the electronic bandwidth of the sensor : this parameter is fundamental in the performance of the sensor and if its value is found to be too low, then it is worthless to start modelling the whole sensor since actual electronic circuits will further degrade this SNR, and no processing technique, whatever its quality, will be able to restaure the situation.
Modelling procedures and corresponding softwares must be specific to each part of the sensor. One can find numerous computer aided design softwares dedicated to optics (emission and propagation of light, optical properties of surfaces and media, optical design, detectors), to mechanics and electronics (signal modulation, demodulation and filtering, analog and digital signal processing, image processing).
These softwares help eliminate the boresome side of computations and lead to finer and finer results because they allow the designer to test and analyze rather large numbers of configurations. However, the designer must not forget to be critical of the results thus obtained, because the validity of such programs rests upon the soundness of the model he himself has chosen.
The results from simulation are then used to define the techinal specifications of all the components and subassemblies in the sensor, taking into account the state of the art. This phase will profit from the experience of the designer with previous designs, and is optimized by consulting the literature (technical journals, internet), taking courses in continuing education programs, going to conferences,...
One must not wait till the instrument has been designed before one starts to think about ways of testing and evaluating it : the designs of both the sensor itself and of its test procedures and benches must be carried out in parallel, because the test definition always brings some help into the design of the device, at all levels (components, subassemblies, system). In case one has to design complex electro-optical sensors, or if there is a small number of items to fabricate, as is the case in defense and space applications, test and evaluation equipments generally represent a non negligible part of the project : a wrong definition or an underestimation of the costs for example may be fatal to the projector even to the company.
To conclude about this design phase, a « golden rule » , to be followed as strictly as possible when setting up breadboards and prototypes, is that every component and subassembly must be controlled and tested before assembly, if one wants to avoid costly dissassembly and reassembly efforts in case of unexplained failures at the system's level.
There may not be possible to put the chosen principle into operation because the necessary technology is not available yet. In that case, one has to evaluate the chances for such a technology to get through and how long that will take, by closely following the progress being achieved in the research labs involved in the corresponding domain.
As was mentionned before, it is of the upmost importance to make operational specifications as precise as possible. The designer must insist on clearing up any ambiguity, especially on which goals are vital and which ones are not, if the sensor is multi-functional.
By definition, the design of an electro-optical sensor is a multidisciplinary job, and it must be well balanced in the evaluation and study of the various contributing elements, without neglecting or underestimating any one of them : any weakness concerning one component or another (source, detector, propagation medium, optical instrument, mechanical parts, servo-controls, filtering, analog and digital signal processing, image processing, display, software,...) will affect the global performance of the sensor.