Cambridge Research Systems Logo Cambridge Research Systems - Tools for vision science CRS Products
CRS Research
CRS Support
CRS Research topics menu Computerised Visual Stimuli
Topics in Computerised Visual Stimulus Generation

Topics in Computerised Visual Stimulus Generation by Tom Robson is published as a chapter of Vision Research, A Practical Guide to Laboratory Methods, edited by Roger Carpenter and John Robson, OUP 1998.


2. How do computers represent images
2.2 What is meant by bandwidth?

Once we have the idea that images or stimuli can be represented in the frequency domain (both spatial and temporal) it is natural to consider what are the highest and lowest frequencies present as any system that will have to display or manipulate these stimuli will need to work equally well for all these frequencies. In other words, in order not to introduce unnecessary distortion into our stimulus, the reproduction system should have a flat frequency response across this range. The bandwidth of the system must exceed the bandwidth of the signal. In practice the observer of the stimulus will place upper and lower bounds on the required bandwidth as there is no point in reproducing signals if they are can't be utilised. For reference, the frequency response of a human, very roughly, is about 0.1 to 40 cpd in the spatial domain and 0.1 to 70Hz in the temporal domain. It is difficult to characterise these bandwidths precisely as they depend significantly on the viewing conditions including such things as the luminance and contrast of the stimulus itself and what else is happening in the neighbourhood. Please beware that the frequency response of a component of a visual system, such as a retinal cell, may be significantly higher than the frequency response of the visual system as a whole and this should be considered when probing internally. It is mush easier to measure the bandwidth of a machine such as a CRT display and this is often quoted by the manufacturer on his data sheets; but more of this later.

The spatial bandwidth of a device is normally determined by its physical size at the low frequency end and the density with which it can produce dots at the high frequency end (the dot pitch). It is worth noting here that the bandwidth of a CRT display, but not a scanner or laser printer, also depends on the ability of the video circuitry to turn the electron beam on and off fast too. In fact it is not just the monitor that has this problem but the visual stimulus generator as well so it is convenient to lump them together and talk about the bandwidth of the system as a whole. This electronic bandwidth limits the ability of the system to reproduce horizontally adjacent pixels and therefore has the affect of reducing the resolution in this direction. Vertically adjacent pixels are unaffected by electronic bandwidth as they are really separated, as far as the electronics is concerned, by the time it takes to scan one line. A high quality visual stimulus generator/display system will have a video bandwidth of at least 100MHz which corresponds to rise and fall times of about 3.5ns. (A good way to measure the bandwidth is to generate a stimulus of parallel dark and light bars and measure the average screen luminance. Do this for horizontal and then vertical bars and for different separations. When the bars are horizontal, changing the spacing between them shouldn’t affect the total luminance but when vertical the bandwidth reduction caused by the video system combined with the non-linearity of the display will create a reduction in overall luminance.)



Figure 1 What happens to the luminance when the bandwidth is poor

arrowPrevious
  Nextarrow


^ Back to top



   
contact us