Electronic – DAC to ADC errors

adcdacerror

I am testing my setup of MCP4725 DAC connected to MCP3221 ADC. DAC's output is connected to ADC's input. I'm measuring the difference between set values in DAC and measured values in ADC and I'm getting these strange patterns.
Edit:x axis: values set in DAC (bits), y axis: DAC value – ADC value (bits)

Minimums and maximums are repeating every 128 bits. Also, errors shift from positive to negative going through the full range. I have tried different delays between setting DAC value and measuring ADC value and I get the same result. So I think it is not background noise.

Edit2:Schematics:

I use REF3033 to power both ADC and DAC. But the same pattern is present when connecting them to LF33CV. Also, adding additional capacitors betwen input/output and ground doesn't change the pattern, just shifts it up or down.

Could you please help me figure out why is this happening and how to avoid it?
I'm really sorry if this is something basic I'm missing here.

Best Answer

DAC output voltage accuracy can be estimated by looking at the data sheet and the output error limits: -

Then you need to understand what terms like INL mean: -

The graph above shows INL (integral non-linearity). Basically, it means that the DAC output could have to +/- 14.5 least significant bits of error from the ideal straight line. Given your DAC is 12 bits (4096 LSBs), the INL produces an error of +/- 0.35%.

But you will also have a zero and a gain error: -

The gain error can be up to 2% of full-scale and the offset error can be up to 0.75% of full-scale.

You can also get an understanding on the error patterns if you delve deaper into the data sheet. For instance this graph on page 5 shows the effect of INL and notice how certain patterns repeat as per the picture in the question: -

Remember also this is just the DAC - a similar picture will exist for the ADC.

Related Solutions

Electronic – How to use AVR ADC with multiplier

int can only be 0-255

That's not correct. In the C programming language, an int is always at least 16 bits wide, even on 8-bit micros. A char is 8 bits wide.

But when using inttypes.h, I strongly recommend using the explicit types like uint8_t,int16_t or uint32_t.

Your real error is using _delay_ms() with a variable parameter. This is not allowed, as it is a macro involving floating point calculations - at least in avr gcc. You are supposed to use something like this:

void DelayMs(int delay) {
  while(delay--) _delay_ms(1);
}

Electronic – How to get a linear relationship between a voltage divider and an ADC

You state that at 128, you measure about 500mV. This means at full scale (256) you measure about 1V.

According to the datasheet (14.12) with:

// Set ADC reference to AVCC
ADMUX  |= (1 << REFS0);

You actually do the following:

This bit selects the voltage reference for the ADC, as shown in Table 14-2. If this bit is changed during a conversion, the change will not go in effect until this conversion is complete (ADIF in ADCSRA is set).

Table 14-2.

Voltage Reference Selections for ADC
REFS0   Voltage Reference Selection
0       VCC used as analog reference.
1       Internal Voltage Reference.

Paragraph 14.3 states:

The ADC converts an analog input voltage to a 10-bit digital value through successive approximation. The minimum value represents GND and the maximum value represents the voltage on VCC or an internal 1.1V reference voltage.

So in conclusion, you are setting the ADC reference voltage to 1.1V and with that knowledge your findings are pretty accurate.

Best Answer

Related Solutions

Electronic – How to use AVR ADC with multiplier

Electronic – How to get a linear relationship between a voltage divider and an ADC

Related Topic