Q: EMAN1/FAQ/Filter When I lowpass filter a map the resolution looks too high, and when I combine multiple filters, the results aren't what I expected.
A: The behavior you're seeing is exactly correct. While there was a time at which in icosahedral reconstructions, people tended to use a 'hard' cutoff filter, eventually they realized how nasty the ringing artifacts this produced were, and switched to smoother filters. In most cryoEM work, people use Gaussian filters nowadays, though in some instances you may also see things like Butterworth or bilateral filters. The properties of a Gaussian filter are exactly what you observe. Say you apply one gaussian filter
F'(s)=F(s) exp(-B1 s^2)
then you apply a second
F(s)= F'(s) exp(-B2 s^2) = F(s) exp(-B1 s2) exp(-B2 s2) = F(s) exp(-(B1+B2) s^2) ie - the 'B factors' in the filter are additive. The next point is the 'apparent' resolution, or more accurately, 'resolvability' of the map after filtration. The way these filters are defined typically in cryoEM is based on their 1/e width in Fourier space. ie lp radius = nx * Apix / filter ie - box size=128 A/pix = 2.4 lp = 30 radius = 128 * 2.4 / 30 = 10.24 (Fourier pixels) B=1/radius^2 However if you look at something like the Rayleigh criterion, a '30 A lowpass filter' can be 'resolved' to much higher resolution (there are several different criteria which could be applied. In single particle reconstruction 'resolution' is a measure of noise levels, NOT resolvability. Due to the additive nature of Gaussian and any other non-cutoff filter, the current resolvability of the map impacts the filter you must apply to achive a particular resolution. ie - if you have a 15 A filtered structure and apply a 30 A filter to it, you won't get the same result as if you start with a 30 A filtered structure.