Numerical Logical Statement Error

Can anyone explain me what's actually happening? Why is the result true, although they are obviously not? How do I fix this problem?

K>>Us
Us =
  -1.7280e+03
K>> P(ct).L
ans =
  -1.7268e+03
K>> err
err =
   1.0000e-03
K>> (Us-P(ct).L) <= err
ans =
  logical
   1

[Stephen23]

"Can anyone explain me what's actually happening?"

Sure, lets just take a look at the value:

>> -1.7280e+03 - -1.7268e+03
ans = -1.2000

Is -1.2 less than 1? Yes, it is (all negative values are less than 1).

"How do I fix this problem?"

The standard way is to use abs:

>> abs(-1.7280e+03 - -1.7268e+03) < =1
ans = 0

Answer 1

[James Tursa]

Both of your numbers are negative, so I don't think this test does what you intended. E.g.,

>> ct = 1;
>> Us = -1.7280e+03
Us =
       -1728
>> P(ct).L = -1.7268e+03
P = 
  struct with fields:

    L: -1.7268e+03
>> err = 1.0000e-03
err =
   1.0000e-03
>> Us-P(ct).L
ans =
   -1.2000
>> (Us-P(ct).L) <= err
ans =
  logical
   1

You probably want an abs( ) wrapper:

>> abs(Us-P(ct).L) <= err
ans =
  logical
   0

And, if the values are very close to the err, note that floating point calculations might not give you the result you expected either, so you may need a tolerance on the tests.

[Limanan Nursalim]

Thank you!!

Answer 2

[madhan ravi]

Just search with a tag floating-point and you will find lots explanations dealing with floating numbers , but obviously Us is lesser than the other.

>> vpa(-1.7280e+03)
ans =
-1728.0
>> vpa(1.0000e-03)   % obviously it's positive
ans =
0.001
>> 

Answer 3

[Steven Lord]

Display the quantity (Us-P(ct).L). It is indeed smaller than err.

Now the absolute value of (Us-P(ct).L) is not smaller than err, but that's not what you asked MATLAB to compute.