Firstly I am getting readings that fluctuate somewhat. The analogue read value changes by 1-2 in what appears to be a oscillating pattern. (period of about 1.5seconds). Any thoughts or do I just need to live with it? The change appears to be happening in the INA125. Tying the analogue input to other voltage sources, such as the ref voltages on the INA125 gives stable results. Any suggestions for filtering, or changes to the circuit?
The datasheet for the ATMega chips will state a 10 bit precision, +/- 2 LSB accuracy. In other words the most you can rely on is 8 bit accuracy.
It seems that the appropriate route would be to add an additional ADC converter to get more bits of accuracy. I'm no expert though and also still experimenting myself.
Secondly, I am not able to obtain results that agree with real world weights. I have made the changes to the code required for the reference values and tried several different values. I have also tried a smoothing using code based on the Smoothing example in the IDE. But what I am finding is the output isn't linear. So the map function results to values that don't agree with the real weights of objects. It appears better at the lower end with the heavy weight skewing the results.
I don't think load cells are expected to give a linear curve for weight vs voltage. The thing to do would be to sample readings with different weights (0Kg, 10Kg, 20Kg, etc.), plot those using a spreadsheet, and then use that spreadsheet's facilities to find the curve of best fit. The resulting formula will give your voltage to weight conversion.
Third problem, any advice on getting the output of the INA125 to run near 0volts for no load on the scale and nearer to 5 or 3.3 at full load. (100Kg).
The solution is to balance the bridge using a resistor (resistors) between the excitation and output lines. I'm not sure that's a good idea though as any resistors you add to the bridge will introduce new problems with the difference in sensitivity to temperature of the load cell's resistors and your resistor. Best bet is to grin and bear it?