Finding maximum source impedance for A/D converter (SAM3X8E ARM Cortex-M3)

Question

I would like to find out what is maximum recommended source impedance for A/D converter on Atmel SAM3X8E ARM Cortex-M3 microcontroller.

I tried to follow the example of this excellent answer, which also explained the maximum recommended source impedance for Atmel AVR 328P MCU (being < 10k). There are even quotes from datasheet in the linked topic, giving this answer specifically. Therefore I also checked the Atmel SAM3X datasheet, hoping to find this information. But I had no luck spotting the answer from page 1317 onwards, which deal with ADC.

I might add that I intend to use 12 bit resolution capability of the ADC.

How should I be searching for this information and where to find it?

Answer 1

There are charts on pages 1408 and 1409 of that datasheet which give ADC max. source impedances vs ADC frequencies.

According to that chart, the 'worst case' acceptable source impedance is 353Kohm for 10-bit resolution @ 1MHz frequency.

Answer 2

With many ADCs, the maximum source impedance is ultimately dominated by pin leakage current, in particular those embedded in microcontrollers when multiplexed with digital I/O functions. This is due to a number of causes, but always exists.

Charge redistribution devices are a bit more complex.

This has nothing to do with the time taken to charge the sample capacitor (which determines the maximum source resistance vs. sample rate)

On page 1380 of the data sheet we find that Vdd powered pins have a worst case low input leakage of 30nA when the pin is at 0V; as this Is the higher of the leakage currents, I will use this value to figure out the maximum source impedance.

To prevent greater than 1% error, we must drive the pin with at least 100 times the leakage (3 microamps), so at midrange (1.65V if you are converting across 3.3V) we get 550kohm, which lines up well with the maximum sample rate vs. source impedance from the previous answer.

If I wanted 0.1% error or less, I would keep the source impedance below 50k. Note that it is common to drive an ADC with a very low impedance source with a device designed for the task to ensure input leakage is not an issue.

This is an estimate, of course, but it seems reasonably accurate.

Answer 3

Per your first reference, Cs/h is charged through a 1..100k resistor

What you need is the output impedance of your driving circuit to be able to drive the ADC input fast enough that the reading is not skewed by the slow charging of the sense cap.

So what you are looking for is an output impedance of your drive circuit << 1Kohm (for fast sampling cases) or << 100k (for slow sampling cases).

In practice I like this to be as small as possible and drive almost everything that is a time varying signal using an OPAMP on the input to the ADC.

Power lines and other slow varying or low impedance inputs I would limit to be < 1/3 of the expected input resistance imposed by the sampling time.

In many cases, I sample the same pin twice and disregard the first reading which is skewed by the input in the penultimate multiplexer setting.