ADC READING ERROR

Ali Tatari

Hi,

I am using MSp430F248 in my circuit and the problem is 12-bit ADC readings are little off approximately by 10 to 15 counts. It is two plane board with Digital ground and Analog ground tied together. So basically there is only one ground. I added 10uf 6V caps to pin 1 and pin64 i.e DVCC and AVCC as mentioned in user guide document. I calculate the counts that I should expect by measuring the scaled down voltage that is being input to the ADC pins and using the formula given in datasheet i.e Vadc = Vin * 4095 /vref+. I am using internal Vref+ -2.5V and Vref- = 0V. Can someone please give me some idea what could be the reason for it. I am really having tough time with this.

Ali.

over 13 years ago

0 Lenio Cacula over 13 years ago

TI__Genius 16700 points

An error of 15 in 4096 is less than 0.4%, which represents 0.9mV in a 2.5V range. You have to have a very stable input voltage and a high precision voltmeter to determine that this is a problem with the ADC. Normally the electrical noise is enough to justify that difference. Are you averaging your reading? You can always improve your dynamic range by doing so. Increasing the sample rate and adding an appropriate low pass filter to the circuit also helps.

0 Ali Tatari over 13 years ago in reply to Lenio Cacula

Prodigy 250 points

Low pass filter to ADC inputs?? In my system I am making calculations based off the ADC readings and it is important that I get good results. I have all the necessary capacitors to counter for grounding issue but still it is the same problem. It varies sometimes it 15 counts off and sometimes it is more than that.

0 Lenio Cacula over 13 years ago in reply to Ali Tatari

TI__Genius 16700 points

Have you measured the noise level in your input? Good grounding is important but not sufficient to eliminate all of the noise. Also, if your input is oscillating (anything different than a super flat DC level) you should compensate for aliasing and that is why the recommendation for a low pass filter. That filter will depend on your sampling rate (Nyquist frequency - you can read up on it).

0 Ali Tatari over 13 years ago in reply to Lenio Cacula

Prodigy 250 points

I have measured the noise level. This is a good suggestion but which sampling rate you are talking about. ADC sampling rate.My input is DC power supply and it is super DC.

0 Lenio Cacula over 13 years ago in reply to Ali Tatari

TI__Genius 16700 points

If your input is a DC signal, you do not need to worry much about aliasing. In that case any sampling rate is ok. The averaging would still be important to reduce your error. Is the noise below 1mV?

0 Ali Tatari over 13 years ago in reply to Lenio Cacula

Prodigy 250 points

The input to ADC is low in noise about less than 1mV.

0 Lenio Cacula over 13 years ago in reply to Ali Tatari

TI__Genius 16700 points

If your noise is right at 1mV, it would represent a reading of 15 with a vref of 2.5V and justify the difference you are having.

0 Jens-Michael Gross over 13 years ago in reply to Lenio Cacula

Guru 227245 points

Lenio Cacula said:
An error of 15 in 4096 is less than 0.4%, which represents 0.9mV in a 2.5V range.

It's 10mV. 15*2.5/4095 = 0.009157509.

A noise of 1mV will result in a count of ~2, not 15.

However, offset and gain drift with temperatureas well as VCC, and the calibration values (if available and used) are for 30° and 3V VCC.

**Attention** This is a public forum

MSP low-power microcontrollers

MSP low-power microcontroller forum

ADC READING ERROR