To express the magnitude of the input vector field we use arrows or other glyphs which scale with the magnitude; but it seems natural to equate visual impact with area, and so scale that the area of an arrow is proportional to magnitude. One chooses a value for the constant of proportionality: the scaling.

In choosing the scaling we decide the sizes of the arrows, and as the scaling increases, the number of arrows that one could fit into a plot reduces. In fact one can use the --number option to specify the number of arrows in the plot (approximately), the program will then bisect on scalings to find one with the specified number of arrows. This behaviour is the default, with the --number set to 200. Experiment with that value, and when you're happy with that, take the --scale value as printed by the program verbose output

Using the --scale option has the advantage that one does not need to run the bisection to find the scale from the --number, that can take an annoying amount of seconds.

------------------
   scale      num
------------------
  0.0000   472.43
  1.0000     1.32
  0.5000     2.02
  0.2500     2.68
  0.1250     3.37
  0.0625     4.82
  0.0312     7.93
  0.0156    13.31
  0.0078    22.88
  0.0039    38.32
  0.0020    61.54
  0.0010    93.63
  0.0005   133.72
  0.0002   179.17
  0.0001   226.58
  0.0002   198.82
  0.0002   211.32
  0.0002   204.78
  0.0002   201.73
  0.0002   200.26
  0.0002   199.53
------------------
use --scale 0.000180 to avoid autoscale