You should not base your schema on implementation details of the database server you are using.
Only the fields that differs are updated.
But despite the fact that only the fields that differ are updated, the whole document is written back to disk. And that’s for efficiency reasons, compression would be hard on a field by field basis.
Could you please provide closure on some of your other threads by marking some post as the solution?
That helps the forum being useful by indicating to readers that they can trust the answer provided.