I have developed a version of the median filtering algorithm described in XAPP953 and earilier, in more detail in a paper by Chaitali Chakrabarti (mentioned in XAPP953 bibliography). Everything is working almost fine. Why almost? I cannot find information about handling a situation, when 2 or more samples of the same value come into the processed window at once (that is the case in my simple 3x3 implementation, where 3 samples com in and out of the window on every clock cycle). Am I missing something? From what I understand, if the bits resulting from comparation of old samples with the new samples and the new samples among themselves are equal, the 0's and 1's that are then counted to evaluate the rank, the resulting rank would be also the same for all incoming samples, and such situation is not acceptable. I know this post may not be strictly connected to this group's area of interest, but please do not consider it as spam. I just wanted to make sure what do I do wrong, before I try to make any unnecessary modifications.