The lack of self heating in the LED source and the rapid switching of its drive current would tend to suggest that a very linear relation exists between 'on-time' and total light output.

This assumption was checked by the use of Neutral Density (ND) filters placed between source and CCD. A range of ND filters with Density ranging from 1 to 2.6 were obtained and their true densities measured at the wavelength of the LED used. The measurement was done with a photodiode.

The source was then mounted in front of a CCD camera with a small space between into which could be slipped the ND filters. A 10ms pre-flash exposure was then taken and the peak signal measured using IRAF 'imexamine'. The ND filters were then inserted one at a time and the length of the pre-flash exposure calculated so as to deliver the same amount of light to the CCD. For example, when the ND2 filter was inserted the calculated pre-flash would be approximately 100 x 10ms. If as the pulse length changed, the depth of the exposure on the chip also changed , then it would indicate that the source itself was non-linear. Note that with this method the linearity of the CCD used to make the measurements is not important , since the total exposure in all the measurements will be approximately the same.

With this method, the brightness of the LED during 10ms, 91ms, 353ms, 1036ms, 1973ms and 3631ms long pre-flash pulses was measured successfully. A great deal of care was required to compensate for the fact that there were small differences between the demanded and actual pre-flash times (due to subtle bugs in the timing software). The results are tabulated below :

Pulse width	Difference in brightness compared to 10ms pulse
10ms	0
91ms	+0.32%
353ms	+0.6%
1036ms	+0.4%
1973ms	+1.17%
3631ms		+0.5%

This certainly suggests that the shorter pulse widths should be avoided for linearity measurements. These results were confirmed when the linearity of TEK4 was measured. The CCD gave an excellent linearity except for the measurements that used pulse widths less than 100ms, where the non linearity of the CCD dropped by 0.9%. This could of course be a genuine feature of the CCD but it seems more likely that it is an artefact of the LED source.

In conclusion, the Linearity source is linear to better than 1% for pulse widths between 100ms and 8s.

Thanks to Richard Hijmering for making these measurements.