[BioC] backgroundCorrect(method="normexp") problems

[Palmer, Lance]

I am just starting to do microarray analysis.  I have been trying to normalize 4 chips (two dye-swap experiments done in biological replicates)
 
After doing backgroundCorrect(method="normexp") , one of the arrays has all of its $G values at NaN.

This does not happen when any other method is used.
 
Can someone tell me why this is, and if this can be fixed?
 
Thanks
Lance Palmer