how can increase Variable-precision arithmetic

My calculations is so what need more than 32 digits is considered in calculations after fixed point. I used, for example digits(100), in first of my codes, but I did not get right answers. Any one can help me?

Answer 1

[Walter Roberson]

Are you referring to the Symbolic Toolbox?

Could you give an example of a calculation that does not give the answers you expect?

Do you have any symbolic floating point constants such as solve('1.23+x^2') ? If you do then those can end up dragging the precision down, as the symbolic engine will not always treat that as a fraction (e.g., 123/100) or as having implicit trailing 0's (e.g., 1.230000000000000). If you have any fixed-point numbers then you should assume that the symbolic toolbox will treat the expression the same way that scientific calculations are done -- at the lowest precision involved in each sub-expression.