MATLAB: Code that was working under R2012b and do not work any more under R2016a: fcts solve and eval

MATLABsolveSymbolic Math Toolboxvpasolve

Hi, I have following short code that was working under Matlab R2012b (see result below) and do not work any more under R2016a!
Could you examine this!
Thank you
fm    = 1000;
BW_Hz = 100;
delta_OMEGA = BW_Hz/fm;
ai = sqrt(2);
bi = 1;
syms x;
y=bi^2*x^8 -delta_OMEGA^2*bi*x^6  +(delta_OMEGA^2*ai^2-2*bi^2-2*delta_OMEGA^2*bi)*x^4 -delta_OMEGA^2*bi*x^2 + bi^2;
alphas1b=solve(y,x)
alphas1c=eval(alphas1b)
R2012b : OK ——–
alphas1b =
                                              1.0360030147808337582053932803524
                                             0.96524815635941929264411150626075
                                            -0.96524815635941929264411150626075
                                             -1.0360030147808337582053932803524
  - 0.035333235097304614354828859673636 + 0.99937558630254651692473420806315*i
  - 0.035333235097304614354828859673636 - 0.99937558630254651692473420806315*i
     0.035333235097304614354828859673636 + 0.99937558630254651692473420806315*i
     0.035333235097304614354828859673636 - 0.99937558630254651692473420806315*i
alphas1c =
     1.0360          
     0.9652          
    -0.9652          
    -1.0360          
    -0.0353 + 0.9994i
    -0.0353 - 0.9994i
     0.0353 + 0.9994i
     0.0353 - 0.9994i
R2016a : ERRORS ——–
alphas1b =
 -root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 1)^(1/2)
 -root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 2)^(1/2)
 -root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 3)^(1/2)
 -root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 4)^(1/2)
  root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 1)^(1/2)
  root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 2)^(1/2)
  root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 3)^(1/2)
  root(z^4 - z^3/100 - 2*z^2 - z/100 + 1, z, 4)^(1/2)
Error using evalin
Undefined function or variable 'z'.
Error in sym/eval (line 11)
s = evalin('caller',vectorize(map2mat(char(x))));

Best Answer

This is due to a change in the design of function 'solve' starting in R2014b. The solver now returns exact representations of the solutions. To get numeric results, you may want to try function 'vpasolve' instead of 'solve' in the newer release.
As a workaround, please replace line 'alphas1b=solve(y,x)' in your code with the following code to call the numeric solver:
alphas1b=vpasolve(y,x)
Related Question