How do I convert an unsigned integer to a signed integer without causing an overflow or using TYPECAST in Simulink 7.9 (R2012a)?

When complying to the DO-254 standard you must detect arithmetic overflows as errors in a Simulink model. This means that using a Data Type Conversion block to convert from an unsigned integer to a signed integer is not possible if an unsigned N-bit integer >= 2^(N - 1) is to be converted, as this will cause an overflow.

Further more when generating HDL Code a calls to the function TYPECAST are not supported.

Answer 1

[MathWorks Support Team]

The workaround is to carry out the conversion manually, in Simulink or in MATLAB. As an example the MATLAB code to convert a 16-bit integer is outlined below and the Simulink alternative can be found in the attached model 'convert.mdl'.

isNegative = int16(bitget(u,16));

convertedValue = int16(bitset(u,16,0)) + (-2^15)*isNegative;

The code checks the most significant bit of the unsigned integer to determine if the value will become a positive or negative signed integer.

-If positive the converted value is just a simple conversion to a signed integer as no overflow will take place.

- If negative the the first 15 bits are converted to a signed integer and the smallest signed integer value (-2^15) is added to achieve the wrapped signed value of the original unsigned integer.

Answer 2

[Kiran Kintali]

Use reinterpretcast to convert an unsigned number to signed number without causing overflow. The stored integer bits of the number will be reinterpreted.

 >> a = fi(7, 0,3,0);
 >> a.bin
    111

 >> b = reinterpretcast(a, numerictype(1,3,0));
 >> b.bin
    111 % notice no change in stored integer bits

 >> a

a =

     7

 >> b

 b = 

    -1 % notice change in real world value