When building excerpts for presenting a document in search results (using either Cached Copy or the body section of the document, or the original document), search.cgi concatenates context chunks. Every context chunk consists of a search query word padded with its original context from the left and the right sides and can be up to ExcerptPadding * 2 + length(word) characters.
The default value is 40.
The number of context chunks concatenated into a document except can be approximately calculated as ExcerptSize / (ExcerptPadding * 2 + length(word)). So a larger ExcerptPadding value makes every context chunk longer but reduces the number of context chunks that can fit into ExcerptSize.