When recording values within a regular collection, each reading is a separate document. The number of entries within the index increases linearly. If there are 50 readings, there are 50 index entries. If there are 10x the readings, the index grows to 10x the number of entries (although compression helps with the byte-size).
When using buckets, we generally have fewer index entries because we index the bucket "header" rather than each individual entry. There are different approaches to grouping readings into buckets - this is an example of time-based bucketing (we round-down the seconds to the nearest "5"); an alternative is count-based bucketing.