Nicholas Outram

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.

9 years, 3 months ago.

read_u16 on the Nucleo F401- least significant nibble equals the most significant nibble?

The code below is a simple demo of reading the ADC from pin A0 on the Nucleo F401.

As I increase the voltage on pin A0, so I surprised to see non-zero values in bits 0..3 I was expecting zeros. I understood this device to have a 12 bit ADC, and that the bits should be shifted left 4 bits for cross-platform compatibility.

However, they are either zero or equal to bits 12..15

It's easy to mask out, but I don't like not understanding the cause.

Can anyone explain why this is happening.

Nick

simple ADC example

#include "mbed.h"

//Function prototype
void doSample1Hz();

//Global objects
Serial pc(USBTX, USBRX);
AnalogIn POT_ADC_In(A0);
DigitalOut led(LED1);

//Shared variables
volatile static unsigned short sample16 = 0;

//The ticker, used to sample data at a fixed rate
Ticker t;

//Main function
int main()
{
    //Set baud rate to 115200
    pc.baud(115200);

    //Set up the ticker - 1Hz
    t.attach(doSample1Hz, 1.0);

    while(1) {

        //Sleep
        sleep();

        //Display the sample in HEX
        pc.printf("ADC Value: %X\n", sample16);
 
    } //end while(1)
} //end main

//ISR for the ticker - simply there to perform sampling
void doSample1Hz()
{
    //Toggle on board led
    led = !led;

    //READ ADC as an unsigned integer - store in static global variable
    sample16 = POT_ADC_In.read_u16();
}

If the 4 least significant bits were always 0 then the range would be from 0x0000 to 0xfff0, that's not the full range for a 16 bit output. In order to use the full range the value should go from 0x0000 to 0xffff. Copying the 4 most significant bits into the 4 least significant bits is a simple way to achieve this in a nice linear way that avoids any big jumps in the output.

posted by Andy A 17 Sep 2015

Some years ago I worked as a DSP engineer and I never encountered this technique. Thanks for highlighting it.

posted by Nicholas Outram 17 Sep 2015

1 Answer

David Lowe
9 years, 3 months ago.

If you inspect your target's mbed-src, specifically analogin_api.c, you'll see the 12 to 16 bit conversion:

uint16_t analogin_read_u16(analogin_t *obj) {
    uint32_t value = adc_read_u32(obj);

    return (value << 4) | ((value >> 8) & 0x000F); // 12 bit
}

Thus the 12 bit value is normalised to the full 16 bit scale.

Accepted Answer