My calibration tables are relatively small (I think) at about 31 elements that are signed 2-bit integers.
That means that the values are in the range 0 to 3. Is that right? Or, did you mean 2 byte integers?
My calibration tables are relatively small (I think) at about 31 elements that are signed 2-bit integers. My numbers are typically in this format: 0.235, 1.399, -4.211, etc. I converted them all to integers by multiplying by 100 so I wouldn't have to deal with floats. After retrieving the number from the calibration array, I multiply by 0.01 to get the number back to where it needs to be as a float. I am not sure if this is a waste of effort or not.
How often do you read/manipulate the array? Once in setup()? Then, yes it's a waste of effort. 25 times a second? Then it may or may not be.
A 31 element array is using only a small part of SRAM. What else is using the rest of SRAM? Or, are you trying to solve a problem that doesn't really exist?