Why does applying a filter using arithmetic or the digital filter function, result in a different scaling factor on the y-axis?

All expressions in Arithmetic are evaluated by first converting any source channels to unprefixed units (e.g. if the source channel is in mV, it is converted to V by dividing by 1000). The actual unit itself is not used, for example you can add a channel in Volts to a channel in Amps - Arithmetic will just numerically sum the values.

 

Therefore if the units have not been set with the appropriate prefix (eg mV), they will be output in the equivalent no of Volts- which would affect the outputted number by a factor of 1000.

If the data has a digital filter applied, the correct factor will be applied automatically.