Make auto scale actually useful on trends

Auto scaling should be the default in the modern age however we seldom use it does not work as expected.

The problems I have, it needs to be consistent when looking at different time ranges and dates, at the moment every step will adjust and move around. if the period is too small PI will zoom in too far making it look like wild things are happening, it is too long a time then various instrument glitches or zeros, will obscure the actual data. Surely it is possible to keep some statistics on typical values and variance, filtering out outliers and making reasonable trends automatically.

The other problem we have is some tags are related and need to be linked to the same scale, a common example we have a Setpoint (SV) and a process variable (PV), on trends that are auto scaled the SV will be a straight line with 1 or 10 changes zoomed in probably on the centre line, the PV will be on a different scale and zoom, when half of what you want to see is that the process is being controlled (the PV is tracking the SV). We recently inherited an extra plant and these guys had 8 auto scaled trends solely to decide if their thickener was controlling to set point, I rationalised this to one manually scaled trend where you can actually understand the process. They think I'm a magician, but I couldn't help think that if Pi worked as advertised what they had done should have been enough.

Note I can't usually use the single scale option as things like MV should be on the same trend but typically have a different scale.