dsPIC33E ADC with external voltage reference

Question

I am using dsPIC33EP64MC202-I/SP to read analog inputs with the code shown below. It is using a 2.5V voltage reference on pin5 (VREF+) as ADC reference voltage. When I connect 1.1V to pin3 (AN0) ADC1BUF1 (CH0) reads 0x1F5, which converts to 1.223V (501*(2.5/1024)). AN3 (CH0) reads 0x3FF, which seem correct. What is causing the error in reading AN0?

To test the voltage reference, I changed the code to use AVDD (connected to 3.3V) as voltage reference by setting AD1CON2bits.VCFG = 0. Then AN0, ADC1BUF1 reads 0x155 = 341*(3.3/1024) = 1.099V, which is the correct answer.

So I am wondering why the 2.5V reference is not working as expected?

void initAdc() {
    ANSELAbits.ANSA0 = 1;       // set AN0/RA0 as analog input
    ANSELAbits.ANSA1 = 1;       // set AN1/RA1 as analog input
    ANSELBbits.ANSB0 = 1;       // set AN2/RB0 as analog input
    ANSELBbits.ANSB1 = 1;       // set AN3/RB1 as analog input 2.5V
    TRISAbits.TRISA0 = 1;       // set AN0/RA0 as input
    TRISAbits.TRISA1 = 1;       // set AN1/RA1 as input
    TRISBbits.TRISB0 = 1;       // set AN2/RB0 as input
    TRISBbits.TRISB1 = 1;       // set AN3/RB1 as input 
    /* Initialize ADC module */
    AD1CON1 = 0;
    AD1CON1bits.SIMSAM = 1;     // Simultaneous sampling 
    AD1CON1bits.SSRC = 7;       // Auto convert
    AD1CON1bits.ASAM = 1;       // auto sample 
    AD1CON2 = 0;
    AD1CON2bits.CHPS = 2;       // all four channels
    AD1CON2bits.ALTS = 0;       // Always uses channel input selects for Sample MUXA. MUXB not used
    AD1CON2bits.VCFG = 1;       // External VREFH = VREF+, VREFL = Avss 
    AD1CON3 = 0;
    AD1CON3bits.SAMC = 0x1f;    // Auto-Sample Time for 31 TAD
    AD1CON3bits.ADCS = 0x0f;    // TAD = 16TP

    AD1CON4 = 0;

    /* Assign MUXA inputs */
    AD1CHS0bits.CH0SA = 3;      // Channel 0 Positive Input Select for Sample MUXA bits
                                // Channel 0 positive input is AN3
    AD1CHS0bits.CH0NA = 0;      // Channel 0 negative input is VREFL 
    AD1CHS123 = 0;              // CH1 = AN0 (voltage FB), CH2 = AN1 (I_OUT), CH3 = AN2 (temp) 
    /* Enable interrupts */
    IFS0bits.AD1IF = 0;
    IEC0bits.AD1IE = 0;

    /* Enable ADC module and provide ADC stabilization delay */
    AD1CON1bits.ADON = 1;
    __delay_us(20); 
} //initAdc

Answer 1

With external 2.5V ADC reference, when I connect 1.1V to AN0, it reads 1.223V (501*(2.5/1024)) in SW. But the actual voltage connected is 1.1V. This is error I am trying to troubleshoot.

ADCs are not perfect and you should probably take a few more measurements. Here are where the errors can arise: -

• Basic measurement error i.e. DVM not quite perfect
• Reference voltage not 2.5 volts but maybe a bit lower
• ADC Zero error - can be in the realm of 1% and is sometimes specified in LSBs
• ADC Gain error - full scale can be "out" by +/- 1% on a lot of ADCs and this happens even though the reference accuracy may be perfect.
• Integral non-linearity error - divergence from the perfect straight line that represents gain. This can be several LSBs in magnitude
• Differential non-linearity - this is the factor that dictates how imperfect the step accuracy is. Think of it like a staircase where the height/depth of each step has a random factor involved.

If you add all these up, you might approach a total error of maybe 5% and 5% in a full-scale of 2.5 volts, is 0.125 volts of potential error and can turn a real input voltage of 1.1 volts to 1.225 volts as perceived digitally.

Answer 2

I had a similar problem with an PIC24HJ128GP504 and a LM4050-3.3 as external ADC reference. I discovered that the voltage of the LM4050 dropped from 3.3 V to 3.05 V as soon as the ADC started to use it. The reason: I chose a 4k7 resistor to operate the LM4050 at 5 V with 360 µA - this is within the range of 45 µA to 20 mA and seemed reasonable for 10 µA flowing into the ADC. The solution: After reading the manual more carefully I realised that this is the current while ADC is OFF - obviously the LM4050 circuit was not able to keep the 3.3 V stable when loaded. As I found no information for the real current while ADC is ON I used a 1k and the issue was gone.