Floating point calculation in STM32L431CBT6

[DKhan.1]

Dear all,

I am having some issues with precision of floating point calculations with STM32L431CBT6.

As one can see that when I do floating point calculation I am getting 1.23000002 instead of 1.23.

Can anyone suggest more efficient and accurate way to do this kind of calculation. I need very precise value for my application.

Thanks and kind regards,

Dibyendu

[Uwe Bonnes]

I think, convert_int() can be replace by atof(), with maybe byte[4] set to 0. Otherwise flot number have some small inaccuracy.

[Danish1]

The C language has two types of floating-point number - float and double.

float uses 32-bits to represent a number and gives between 6 and 9 digits of precision.

double uses 64 bits and gives twice that.

So just use the word "double" where previously you used "float".

Be careful with scanf.

[Andrew Neil]

---

@Danish1 wrote:

So just use the word "double" where previously you used "float".

---

@DKhan.1 but note that the Cortex-M4's floating-point hardware is only 'float' - so 'double' will be done in software...

[Andrew Neil]

It is in the nature of Floating Point that you will get such errors

https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html 

Have you considered working in (scaled) integers ... ?

 

[Tesla DeLorean]

Use doubles to carry more "precision", but be aware that the binary number space can be very unforgiving with certain numbers, as decimal can be with 1/3

People use different methods for financial / book-keeping for example.

If this is important truncate the numbers, or use integer methods to get the rounding/forms that you want / need.

Be using atof() or atod()

[Karl Yamashita]

---

As one can see that when I do floating point calculation I am getting 1.23000002 instead of 1.23.

Can anyone suggest more efficient and accurate way to do this kind of calculation. I need very precise value for my application.

---

1.23000002 is a more precise number than 1.23.

Or do you mean you want a number with no more than 2 decimal places?