Spice Size Estimation

Based on AWS documentation, to estimate Spice size each text and numeric value uses 12 bytes x no. of rows and text uses (24 bytes + text length) x no. of rows.

My questions are:

  1. For text length is that the size of the field or the size of the value in the field?
  2. If I have 2 data sources linked in a dataset one with 2 text columns (varchar 10) and a 1000 rows joined to a dataset with 2 text columns (varchar 20) and 100 rows, how do i calculate estimated size? Is it (34x2x1000) + (34x2x100) or 34x4x1000. ie, is all the data in a dataset combined into a single tabular file?