RE: HP48-HP50G : Error functions ERF(z), FRESNEL integrals, PROBIT, etc.

[Gil]

HP50: ERF(x), ERFc(x), iERFc(x) and iERFc(x), with x a real number
Now significand or mantissa with 11 to 12 accurate digits
(even for the inverse functions iERF and iERFc).

Note that using UTPN does not give always the full significant digits (or best solution).
The critical point of x lies then at about 0.114294.
Depending on the values of x, an integral was used accordingly (instead of UTPN).
—> For computational time, EMU48 is then highly recommended.

For ERF(z) and ERFc(z), with z a complex number (a b):
- the imaginary part of the resulting complex number should also give the correct 11-12 digits of the mantissa;
- the real part of the resulting complex number, however,
may not show all the mantissa digits (only some incomplete digits), depending on the case.
This issue still should be fixed up.