Defocus refinement is only justified when the images have strong signal, for example, when a large virus is imaged. If the defocus is refined (either per micrograph or per particle), the value for defocus uncertainty restrains the change that can occur during refinement. However, it is not a hard cutoff and will only put a "penalty" on defocus values that deviate from the original values by significantly more than the uncertainty value. It is probably wise to set the uncertainty value to something small, such as 50, and then run a few cycles of refinement to see if the reconstruction actually improves (make sure you check the density as the FSC curve can be unreliable).
Defocus refinement is only
Defocus refinement is only justified when the images have strong signal, for example, when a large virus is imaged. If the defocus is refined (either per micrograph or per particle), the value for defocus uncertainty restrains the change that can occur during refinement. However, it is not a hard cutoff and will only put a "penalty" on defocus values that deviate from the original values by significantly more than the uncertainty value. It is probably wise to set the uncertainty value to something small, such as 50, and then run a few cycles of refinement to see if the reconstruction actually improves (make sure you check the density as the FSC curve can be unreliable).