Different results for formula in Matlab and Excel

I derived a formula using a long calculation and it gives satisfactory answer (1759) when variables are replaced in Matlab

f__prime_c=4000;
kd=44.928;
epsilon_o=2*4000/(57000*sqrt(4000));
epsilon_ct=0.003;
epsilon_c=0.001264;
epsilon_ult=0.003;
z=150;

double((3*f__prime_c*kd*(epsilon_c^3-3*epsilon_c^2*epsilon_o+2*epsilon_o^3))/(500*epsilon_o^2*epsilon_ct)-(9*f__prime_c*kd*(-epsilon_ct^2+epsilon_o*epsilon_ult)*(-z*epsilon_ct^2+2*epsilon_ult+epsilon_o*epsilon_ult*z))/(1000*epsilon_ct*epsilon_ult^2))

But when I bring the same equation in a cell in excel ,answers is different.

I had made sure that replacement is correct but everytime I replace variables with cells or even with numerical values in excel ,answer is different from Matlab.

The Numerical values for variables are same in matlab and excel.

Any apparent reasons for the problem?

Btw answer is 1759 in matlab and -4332 in excel

[DGM]

I don't know why your excel spreadsheet doesn't match, because I can't guess what you put in it. Considering that I can't even get the same results in Matlab, I have no idea.

f__prime_c=4000;
kd=44.928;
epsilon_o=2*4000/(57000*sqrt(4000));
epsilon_ct=0.003;
epsilon_c=0.001264;
epsilon_ult=0.003;
z=150;

% the original
R0 = double((3*f__prime_c*kd*(epsilon_c^3-3*epsilon_c^2*epsilon_o+2*epsilon_o^3))/(500*epsilon_o^2*epsilon_ct)-(9*f__prime_c*kd*(-epsilon_ct^2+epsilon_o*epsilon_ult)*(-z*epsilon_ct^2+2*epsilon_ult+epsilon_o*epsilon_ult*z))/(1000*epsilon_ct*epsilon_ult^2))

% it's a lot easier to read giant expressions if they're split up somehow.
An = (3*f__prime_c*kd*(epsilon_c^3-3*epsilon_c^2*epsilon_o+2*epsilon_o^3));
Ad = (500*epsilon_o^2*epsilon_ct);
Bn = (9*f__prime_c*kd*(-epsilon_ct^2+epsilon_o*epsilon_ult)*(-z*epsilon_ct^2+2*epsilon_ult+epsilon_o*epsilon_ult*z));
Bd = (1000*epsilon_ct*epsilon_ult^2);

% the results are the same anyway
R = An/Ad - Bn/Bd

Both give 1759

R0 =
         1759
R =
         1759

[Osama Anwar]

Sorry 1759. Edited.

[Osama Anwar]

[Osama Anwar]

[Osama Anwar]

The problem is in Bn but I can't figure it out.

I have compared the excel values with matlab values but still unable to figure out the real cause.

[Osama Anwar]

1st pic is of the cell with variable names (R72) and second one is using cell values (S72)

Answer 1

[Steven Lord]

The problem is operator precedence. Here are the rules for operator precedence in Microsoft Excel and here are the rules for MATLAB. In Excel negation (unary minus) takes precedence over exponentiation so -epsilon_ct^2 is evaluated as (-epsilon_ct)^2 while in MATLAB unary minus is at precedence level 4 while exponentiation is at levels 2 and 3 so -epsilon_ct^2 is -(epsilon_ct^2).

epsilon_ct = 0.003;
resultFromMATLAB = -(epsilon_ct^2)

resultFromMATLAB = -9.0000e-06

resultFromExcel = (-epsilon_ct)^2

resultFromExcel = 9.0000e-06

When I modified the formula in Excel to use parentheses like resultFromMATLAB above its answer agreed with the answer from MATLAB. Or you could rewrite that term slightly to reduce the need for parentheses, since subtraction is lower than exponentiation in both Excel and MATLAB.

epsilon_o=2*4000/(57000*sqrt(4000));
epsilon_ct=0.003;
epsilon_ult=0.003;
term1 = epsilon_o*epsilon_ult-epsilon_ct^2

term1 = -2.3426e-06

term2 = -(epsilon_ct^2) + epsilon_o*epsilon_ult

term2 = -2.3426e-06

term3 = (-epsilon_ct)^2 + epsilon_o*epsilon_ult

term3 = 1.5657e-05

[Osama Anwar]

Thanks man. I replaced - by +(-1)* and it worked.

Answer 2

[Jan]

It looks, like you have a typo in the code you have typed in Excel.

[Osama Anwar]

I copied the matlab formula as it is and not typed it in excel.

[Image Analyst]

But did you define the range correctly so that the named range points to the actual correct cell? You keep forgetting to attach the Excel workbook. Why are you not attaching it???

[Osama Anwar]

Attached.

[Osama Anwar]

@Image Analyst Yeah I did that

Answer 3

[Osama Anwar]

Excel file