Finding the linear portion of an opamp output

[AlfBaz]

Currently I'm simulating a single supply inverting opamp (Ideally I'd like non-inverting but I don't know how to get "fractional" gain with one)

The input spans from 0V to Vcc, which is also the supply volts for the opamp and its not rail to rail.
The output starts off flat, then goes into a linear transfer section and flattens out again.

I want to limit the input span to the linear region of the output, is there an easy way to determine that?

Keep in mind that when I bread-board this I'll collect actual values and want to do the same to these points

I'm using ltspice and can export the data and graph it in an external program such as Graph or excel

The only method that springs to mind at the moment is take 2 points in the middle of the linear section, creating a linear equation for them and then checking the deviation of the actual values from the derived line

Are there any simple methods I'm missing or is my circuit fundamentally flawed

[Rufus]

Add a trace

v(VCC) - v(Vin) - v(Vout)

It will be zero in the linear region - you decide how far away from 0 is non linear and measure those points.

A 'fractional non-inverting amplifier' is just a voltage divider which you can buffer with a unity gain op-amp if you need to.

[AlfBaz]

Add a trace

v(VCC) - v(Vin) - v(Vout)

It will be zero in the linear region - you decide how far away from 0 is non linear and measure those points.

A 'fractional non-inverting amplifier' is just a voltage divider which you can buffer with a unity gain op-amp if you need to.

Excellent

The only down side to this is having to measure 3 voltages simultaneously on the bread board assuming some drift in Vcc. I guess it depends on how much accuracy I need, haven't gotten that far yet

I do have a scanning card on my hp3457A and I just checked, those relays change over really quick when I drop the NPLC. I would need to run some tests to see what effects this would have on the adc if the voltages on each channel where very different.

With that amount of work it might just pay to just sample the output and extrapolate linearity from the middle of the span.

Nevertheless love the elegance of your solution