We tested the general problem that you describe by modifying our ScrollApplication1 example to produce a large dynamic swing of data, from += 5000 to += 30000 and more. We made the following changes to the program:
Increased the inital values of the currentTemperature variables:
double currentTemperatureValue1 = -5000.0;
double currentTemperatureValue2 = 5000.0;
Increased the default range of the initial display by setting the currentTemperature default minimum and maximum display values
currentTemperature1.setDefaultMinimumDisplayValue(-10000);
currentTemperature1.setDefaultMaximumDisplayValue(10000);
currentTemperature2.setDefaultMinimumDisplayValue(-10000);
currentTemperature2.setDefaultMaximumDisplayValue(10000);
and increased the data simulation to include a strong +- bias in the data so that each temperature ramps towards + or = 30000.
currentTemperatureValue1 += 1000 * (0.65 - ChartSupport.getRandomDouble());
currentTemperatureValue2 += 1000 * (0.35 - ChartSupport.getRandomDouble());
counter++;
currentTemperature1.setCurrentValue(currentTemperatureValue1);
currentTemperature2.setCurrentValue(currentTemperatureValue2);
When running the example we could find no problem with the ScrollFrame autoscaling. It seemed to auto-scale properly in both direction. See what you can do to reproduce your problem using this, or any of our example programs.