Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

Julio Vazquez wrote:
> As a commentary, I agree the Zeiss "optimal" settings are not strictly 
> "Nyquist", but I would also argue (and here I am venturing well beyond 
> my comfort zone) that "Nyquist" was developed originally for other 
> purposes than microscopy, and there is a fair degree of arbitrariness 
> to it. Basically, if you sample a theoretical periodic signal, the 
> more you sample, the closer you will get to representing the actual 
> signal. The value of 2.4, in my somewhat simplistic understanding, is 
> just a compromise where you get a fair representation of the sampled 
> signal, while minimizing resources (data storage space, processing 
> time, etc...), but at 2.4 samples per cycle, you only get a crude 
> representation of the original signal (although you do preserve the 
> frequency, which you wouldn't if you sampled at 2x or less). 
Hi Julio

There is nothing arbitrary about digital sampling of signals. The 
mathematics of Nyquist-Shannon sampling is clear and is valid for 
discrete sampling of _all_ continuous signals. If you wish to retain all 
the information in a digital representation of a analog (or continuous) 
waveform (of any number of dimensions) you must sample at a sufficient 
rate to capture all the (spatial) frequencies contained therein. To 
sample at less than this rate does not jut degrade the output, it makes 
artifactual data appear. All you have to do is decide where the 
modulation transfer function of the instrument becomes close enough to 
zero (across all relevant dimensions) and sample at just over twice that 
frequency. At that frequency, the representation is not 'crude' it is 
perfect as it contains all the information needed to reconstruct the 
original signal, picture or volume...

My 2c

Cheers Mark