In perception based coders the encoding process is controlled by the global signal-to-mask ratio vs. frequency curve. If the necessary bit rate for a complete masking of the distortions is available the coding scheme will be transparent, i.e., the decoded signal is indistinguishable from the source signal. If the necessary bit rate is not available, the global masking threshold serves as a spectral error weighting function: the resulting error spectrum will have the shape of the global masking threshold. In practical designs of perceptual coding we cannot go to the limits of masking or just noticeable distortions, since post processing of the acoustic signal (e.g. filtering in equalizers) by the end user and multiple encoding/decoding processes may demask the noise. In addition since our current knowledge about auditory masking is very limited, the hearing model used in the design of particular perception based coder may not be accurate enough. Therefore, as an additional requirement, we need sufficient safety margin in practical designs of coders.