Dear Richard, 
   I guess it depends a lot on how many spectral channels you need. We 
probably wish to have perfect high resolution spectra, but often they are not 
really needed unless you wish to analyze the "fine" structures of spectra. 
Would you make unmixing, then the number of channels and required resolution 
will be somehow related to the number of fluorophores to be unmixed and their 
minimal separation.

This preamble to say that a prism-based system such the Leica SP* equipped 
with the 2 (if I remember correctly) additional spectral channels already 
provide high photon-efficiency and 5 spectral channels. The lambda scan 
option of the Leica is quite slow and therefore useful only for photostable and 
usually fixed samples (also be aware of some non-linearities of the detection 
system which will "deform" the spectra).

The spectral options of Nikon and Zeiss will be somehow less photon efficient 
compared to the collection of 3-5 parallel channels with a Leica, but certainly 
comparatevly faster and more photon efficient than scanning a higher number 
of channels.

We are using a prism based EMCCD spectrograph for spectral detection 
coupled with a confocal getting 128 spectral channels with pixel dwell times of 
100-400 us). This allows to combine the high efficiency of prism based 
detection with the high quantum efficiencies of EMCCDs. I guess that if you 
need a high number of channels this would be the optimal solution, but often 
fewer channels already do the job.

Regards, 

Alessandro Esposito
Laser Analytics Group
Dept. Chemical Engineering and Biotechnology
University of Cambridge