CODEWITHSUNDEEP
matlabsimulinkfixed-point
Papa Grisha
Papa Grisha

Reputation: 11

Cannot round correctly a MATLAB fixed point number to the Nth decimal place

For the purposes of my application the result of a fixed point calculation needs to be rounded to the 2nd decimal place first and then rounded to integer. So, if you consider a number like 0.4962 after you round it to 2 decimal places you get 0.5. And once you round that to integer you get 1. Problem is some values are not getting rounded correctly.

Since MATLAB's round function does not support Nth decimal place rounding for arguments of type fi I'm using the following algo:

  1. Multiply the number with 10 ^ (desired decimal place)
  2. Round result to integer using the Round rounding method
  3. Divide result by 10 ^ (desired decimal place)

Let's consider this code:

format long g
a = fi(33.495, 0, 64, 40)
b = fi(0, 0, 64, 40)
math = fimath('RoundingMethod', 'Round', 'ProductMode', 'SpecifyPrecision', 'ProductWordLength', 64, 'ProductFractionLength', 40)
b(:) = round(math.mpy(a, fi(100, 0, 8, 0)))
b(:) = divide(numerictype(0, 64, 40), b, 100)

Unfortunately the output of this isn't 33.5 as expected, but 33.49 Here's the complete output:

          33.4949999999999

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 40
b = 
     0

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 40
math = 
        RoundingMethod: Round
        OverflowAction: Saturate
           ProductMode: SpecifyPrecision
     ProductWordLength: 64
 ProductFractionLength: 40
               SumMode: FullPrecision
b = 
        3349

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 40
b = 
          33.4899999999998

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 40

What do I do?

Upvotes: 0

Views: 376

Answers (1)

A-T
A-T

Reputation: 356

Matlab supports rounding a number upto a specified number of decimal points.

var = 0.4962
var = round(var,2) # (gives you 0.50)
var = round(var) # (gives you 1.0)

Upvotes: -1

Related Questions