Important changes to forums and questions

All forums and questions are now archived. To start a new conversation or read the latest updates go to forums.mbed.com.

Convert float to 4 Bytes and then back again

Topic last updated 17 Apr 2019, by Ayoub Ayoub. 17 replies
Tom Gregory
24 Mar 2011

I need to convert a float to 4 bytes so it can be sent on a 8 bit data channel. I then need to convert these 4 bytes back to a float. Can anyone advice me on how to do this.

Thanks

Igor Skochinsky
24 Mar 2011

The simplest way is just to treat it as a buffer.

 float x;
 i2c.write(addr, (char*)&x, sizeof(x));
 ...
 i2c.read(addr, (char*)&x, sizeof(x));

If you need to access each byte separately, use a union:

 union float2bytes { float f; char b[sizeof(float)]; };
 
 float2bytes f2b;

 float x;
 f2b.f = x;
 for ( int i=0; i < sizeof(float); i++ )
   send_byte(f2b.b[i]);
 ...
 for ( int i=0; i < sizeof(float); i++ )
   f2b.b[i] = read_byte();

 x = f2b.f;

Note that you need to make sure the sender and receiver both use the same endianness and floating-point format.

Andy K
24 Mar 2011

Something like this should work. Use Google to find out more about C unions. You''ll also want to make sure you know how much C data types take up space wise.

typedef union _data {
  float f;
  char  s[4];
} myData;

#include "mbed.h"

Serial pc(USBTX, USBRX);

myData q;

int main() {
  q.f = 1.234;
  pc.printf("%c%c%c%c", q.s[0], q.s[1], q.s[2], q.s[3]);
}

I'l leave you to work out the reverse from chars back to float :)

Yoko Hama
25 Mar 2011

simplest way

float f; uint8_t *p = (uint8_t*)&f;

for (int i = 0; i < sizeof(f); i++) printf("0x%02x ", p[i]);

Tom Gregory
26 Mar 2011

Thanks for your help.

Luis Linares
18 Oct 2011

My solution is to employ typecasts. This makes it, of course, endianness dependent, which as explained in here, is little-endian for the LPC1768. For the examples, let's assume the float starts at position 3 of the byte array. so:

char my_array[10];
float my_float;

my_float = *( (float*)(my_array + 3) );           // from array to float

From float to array, any of the above solutions works as well :)

steve childress
18 Oct 2011

In C, the union type is often used. Something like

union floatUnion_t
{
    float f;
    char bytes[sizeof float]
} floatUnion1;
Javier Sing
06 Dec 2016

Andy K wrote:

Something like this should work. Use Google to find out more about C unions. You''ll also want to make sure you know how much C data types take up space wise.

typedef union _data {
  float f;
  char  s[4];
} myData;

#include "mbed.h"

Serial pc(USBTX, USBRX);

myData q;

int main() {
  q.f = 1.234;
  pc.printf("%c%c%c%c", q.s[0], q.s[1], q.s[2], q.s[3]);
}

I'l leave you to work out the reverse from chars back to float :)

Hi, im trying to convert a float variable in to hex and store it in a new variable, will this work for doing this?

melvin awesome
21 Jul 2017

Troubleshooting!

Hi guys i need some help with my code. i trying to convert readings in byte array to float with union but can't seem to work. i manage to get reading of the float but the reading is 0.00 Here is my code:

main.cpp

#include "mbed.h"

typedef union _data {
  float f;
  char  buffer2[4];
} myData;

myData q;

//--------------------------------------
//| Hyperterminal configuration         |
//| 115200 bauds, 8-bit data, no parity |
//--------------------------------------

Serial pc(SERIAL_TX, SERIAL_RX); //PC to STM32 Board
//Serial transmit(PC_4,PC_5); // STM32 writing/reading with Arduino
Serial uart(PC_4,PC_5);     

//DigitalOut myled(LED1);
DigitalIn BlueButton(USER_BUTTON);  // This is Blue-Button and is on NUCLEO-L153RE
DigitalOut pc_activity(LED1);
DigitalOut uart_activity(LED2);
 
#define Pressed 0
#define NotPressed 1



int main() {
    char buffer[5];
     char buffer1[4];
    
    //float my_float;

    pc.printf("Select one of the following...");
    pc.printf("\r\n");
    pc.printf("i = deice type and version number");
    pc.printf("\r\n");
    pc.printf("r = take a single pH reading");
    pc.printf("\r\n");
    
    uart.baud(115200);                          //Setting of baud rate
    int printready =0;
    int i =0;
    while(1) {
        
        if(pc.readable()) {                     //If PC port got reading/get input of char
            uart.putc(pc.getc()); 
            
             //wait(0.5); 
             //uart.clear();   
                     //Sending char to arduino
            pc.printf("Message Sent!\r\n");
            
            //pc_activity = !pc_activity;       //LED indication
        
        
             while(uart.readable()) {                   //UART excute only when receive reading            
            char tempc = uart.getc();
           // if(tempc != 'v')
                buffer[i] = tempc;//uart.getc();
                
            wait(0.5);
            i++;
        }
         printready = 1;
        }                                       //sending working 
              if(printready == 1){
        pc.printf("uart is %s",buffer);
        for(int j = 0; j< 4; j++)
            buffer1[j] = buffer[j+1];  
       pc.printf("\nnew buffer is %s",buffer1);
       pc.printf("\r\n");
       
        //my_float = *( (float*)(buffer1 + 3) );    //not working
       
      // q.buffer2[0]=buffer1[1];
     //  q.buffer2[1]=buffer1[2];
     //  q.buffer2[2]=buffer1[3];
     //  q.buffer2[3]=buffer1[4];
       
        pc.printf("Float:%.2f", q.f );
        printready = 0;
        }  
    }
}
randy ransier
18 Jan 2018

some old functions i had laying around if anyone wants to get into bit bangin stuff you can split and or rejoin 32bits and 16bits to and from 8bits just ignore the printf's they are for testing on a local compiler, the math is in the shifts (32 bit are down a little)

////  // splits a 16bit into 2 8bits

rightshifty16 (int16_t argn)
{
unsigned char bytes[2];
bytes[0] = (argn >> 8) & 0xFF;
bytes[1] = argn & 0xFF;
printf("0x%02x 0x%02x\n", (unsigned char)bytes[0],
                        (unsigned char)bytes[1]);
}

////  // combines 2 8bits to a 16bit

leftshifty16 (unsigned char argb[])
{
int16_t integers;
integers = (argb[0] <<8) + argb[1];
printf("%d\n", (int16_t) integers);
}

////  // splits a 32bit into 4 8bits

rightshifty32 (int16_t argn)
{
unsigned char bytes[4];
bytes[0] = (argn >> 24) & 0xFF;
bytes[1] = (argn >> 16) & 0xFF;
bytes[2] = (argn >> 8) & 0xFF;
bytes[3] = argn & 0xFF;
printf("0x%x 0x%x 0x%x 0x%x\n", (unsigned char)bytes[0],
                        (unsigned char)bytes[1],
                        (unsigned char)bytes[2],
                        (unsigned char)bytes[3]);
}

//// // combines 4 8bits to a 32bit

leftshifty32 (unsigned char argb[]) 
{
int32_t integers;
integers = (argb[0] <<24) + (argb[1] <<16) + (argb[2] <<8) + argb[3];
printf("%d\n", (int32_t) integers);
printf("0x%x\n", (int32_t) integers);
}
Steve Mylroie
14 Feb 2018

Using union for byte to int32 works on x86 architecture processor but not on ARM based processor since the ARM architectures do not support sub word addressing A byte on ARM occupies a full word in memory on a ARM processor

Zoltan Hudak
15 Feb 2018

It seems that union works also on ARM processors. Check the code below:

#include "mbed.h"

union Float {
    float    m_float;
    uint8_t  m_bytes[sizeof(float)];
};

float        pi;
uint8_t      bytes[sizeof(float)];
Float        myFloat;

int main()
{
    pi = 3.141565f;
    
    printf("\r\n*******************************************\r\n");
    printf("*******************************************\r\n\r\n");
    printf("pi = %.6f\r\n", pi);
    printf("bytes = [ %d, %d, %d, %d]\r\n", bytes[0], bytes[1], bytes[2], bytes[3]);
  
    printf("\r\n***** Conversion by using type casting *****\r\n\r\n");
    
   *(float*)(bytes) = pi;  // convert float to bytes
    printf("bytes = [ 0x%.2x, 0x%.2x, 0x%.2x, 0x%.2x ]\r\n", bytes[0], bytes[1], bytes[2], bytes[3]);
    
    // send bytes over data channel ..
    
    // receive bytes ..
        
    pi = *(float*)(bytes);  // convert bytes back to float
    printf("pi = %.6f\r\n", pi); 

    printf("\r\n********* Conversion by using union *********\r\n\r\n");
       
    myFloat.m_float = pi;   // assign a float to union
    printf("myFloat.m_Float = %.6f\r\n", myFloat.m_float);
    printf("myFloat.m_Bytes = [ 0x%.2x, 0x%.2x, 0x%.2x, 0x%.2x ]\r\n", myFloat.m_bytes[0], myFloat.m_bytes[1], myFloat.m_bytes[2], myFloat.m_bytes[3]);  // get the bytes   
    
    // send bytes over the data channel ..
    
    // receive bytes and assign them to the union bytes: myFloat.m_bytes[0] = received_bytes[0] ..        
   
    pi = myFloat.m_float;   // get the float back from the union
    printf("pi = %.6f\r\n", pi);    
}
Ayoub Ayoub
17 Apr 2019

It works

Zoltan Hudak wrote:

It seems that union works also on ARM processors. Check the code below:

#include "mbed.h"

union Float {
    float    m_float;
    uint8_t  m_bytes[sizeof(float)];
};

float        pi;
uint8_t      bytes[sizeof(float)];
Float        myFloat;

int main()
{
    pi = 3.141565f;
    
    printf("\r\n*******************************************\r\n");
    printf("*******************************************\r\n\r\n");
    printf("pi = %.6f\r\n", pi);
    printf("bytes = [ %d, %d, %d, %d]\r\n", bytes[0], bytes[1], bytes[2], bytes[3]);
  
    printf("\r\n***** Conversion by using type casting *****\r\n\r\n");
    
   *(float*)(bytes) = pi;  // convert float to bytes
    printf("bytes = [ 0x%.2x, 0x%.2x, 0x%.2x, 0x%.2x ]\r\n", bytes[0], bytes[1], bytes[2], bytes[3]);
    
    // send bytes over data channel ..
    
    // receive bytes ..
        
    pi = *(float*)(bytes);  // convert bytes back to float
    printf("pi = %.6f\r\n", pi); 

    printf("\r\n********* Conversion by using union *********\r\n\r\n");
       
    myFloat.m_float = pi;   // assign a float to union
    printf("myFloat.m_Float = %.6f\r\n", myFloat.m_float);
    printf("myFloat.m_Bytes = [ 0x%.2x, 0x%.2x, 0x%.2x, 0x%.2x ]\r\n", myFloat.m_bytes[0], myFloat.m_bytes[1], myFloat.m_bytes[2], myFloat.m_bytes[3]);  // get the bytes   
    
    // send bytes over the data channel ..
    
    // receive bytes and assign them to the union bytes: myFloat.m_bytes[0] = received_bytes[0] ..        
   
    pi = myFloat.m_float;   // get the float back from the union
    printf("pi = %.6f\r\n", pi);    
}