Compare Atmega 328 ADC With High Resolution ADC

[MrAl]

Hello there,

Anyone else do any comparisons of this chip vs a higher resolution ADC?
This chip is typically found on an Arduino such as the Uno.
High res ADC can be any chip with higher than 10 bit resolution.

I found about 5mv different max, after adjusting the code factor for the carefully measured +5v line to the Uno.  It varies a bit too though.  Measurement point close to 4.000 volts.

A good idea would be to sweep the whole range from 0.000v to 5.000v but i wanted to start with this single measurement.

[ralphd]

These guys have done some detailed analysis of the ADC:
http://www.openmusiclabs.com/learning/digital/atmega-adc/in-depth/

[MrAl]

Hi,

What i was looking for mostly was a comparison of the DC accuracy to other ADC's perhaps or high resolution volt meters.

[madires]

I think it would a good idea to look into voltage references first before discussing quantization errors and conversion methods. The ATmega328's internal bandgap reference can vary widely and Vcc isn't an accurate voltage either.

[ralphd]

I think it would a good idea to look into voltage references first before discussing quantization errors and conversion methods. The ATmega328's internal bandgap reference can vary widely and Vcc isn't an accurate voltage either.

You should see within 2% of the datasheet typical Vbg for a specific voltage and temperature.

[mikerj]

Hi,

What i was looking for mostly was a comparison of the DC accuracy to other ADC's perhaps or high resolution volt meters.

You need to expand on that.  ADC accuracy isn't defined by a single number, and in the case of the AVR there will be separate accuracy specifications for the internal voltages reference (quite poor) and the ADC itself.

If you are attempting to measure voltage by assuming a perfect 1.1v reference and zero gain and offset error, your worst case error could be pretty bad (the nominal reference voltage can vary by up to +-0.1v so that's 10% off the bat).

[MrAl]

Hi,

Thanks, but i dont really care about the reference, i just care about the ADC itself.  When i apply a known reference how well does it actually read the voltages.

I had found that the 1.1v reference can be quite a bit off, but i assume that once it is measured it can hold for the same temperature.  I cant find a temperature plot for the reference, so all i can do is assume that it varies like any other band gap reference around 1.1v, which isnt too much.  But i dont want to concentrate on the reference itself, just the measurement accuracy given a constant known reference of 5.000 volts.

In other words, if i measure 4.100 volts is it really 4.100 or 4.101 or 4.102 or 4.103 or 4.110 etc.  And this would be after a single point calibration (i know multiple points can do better, but i am trying to characterize the ADC without extensive calibration).

Here is how it would go...
Measure a constant DC voltage with a high quality meter.
Measure with the ADC with an ADC adjust factor of (float) 5.000 volts.
Using the ADC reading and the quality meter reading, compute the true adjust factor.  Say it comes out to 4.935 volts.
Change the 5.000 to 4.935, then test again to verify, then test over the whole range say at 0.100v, 0.500v, 1.000v, 2.000v, 3.000v, 4.000v.
Compare the readings from the ADC with those 6 quality meter readings.

[ralphd]

I cant find a temperature plot for the reference, so all i can do is assume that it varies like any other band gap reference around 1.1v, which isnt too much.

I'm guessing you didn't read all 660 pages of the datasheet, and in particular page 615. :-)

[MrAl]

I cant find a temperature plot for the reference, so all i can do is assume that it varies like any other band gap reference around 1.1v, which isnt too much.

I'm guessing you didn't read all 660 pages of the datasheet, and in particular page 615. :-)

No i just clicked, "Agree" (har har) :-)

Thank you for pointing that out.  I've been reading so many data sheets lately that my eyes are falling out.  That will help very much, and any other information like that will help.

What else would help is some actual tests and measurements.  Similar to what i posted previously but whatever anyone has done already would be nice to see too.
I know i could probably find the spec for the ADC basic accuracy too, but i wanted real world data.