Real data always comes with some noise, and it is important that the software does not worsen the noise spectrum (in pixel as well as in data space) more than necessarily. With the command CREATE/RANDOM_IMAGE, images with various random--number distributions can be generated. For most purposes it will be sufficient to just add such an image (using COMPUTE/IMAGE) to the frame with the test pattern. This would simulate the case where there is only detector read--out noise. If for some reason the simulation of photon noise is more desirable, this case can be approximated by scaling (again using COMPUTE/IMAGE) the amplitude of a suitable noise distribution (output of CREATE/RANDOM_IMAGE) with the square root of the pattern before it is added to the pattern frame. Evidently, all kinds of combinations of these two cases can be used.
If the STMODEL context is available (activated by SET/CONTEXT MODEL in a MIDAS session), then the command HPOIS may also be used for the simulation of noise.