I wish to measure a voltage that may vary between 11 V and 15 V with the ADC of an AVR32DA32. I'm using a voltage reference of 2.5 V and picked a 5.1 kΩ and a 820 Ω resistor to create a divider to bring the voltage range down to 2.078 V and 1.524 V.

Now admittedly, I chose these resistor values almost completely arbitrarily, other than to bring the voltage below 2.5 V while using common resistor sizes. I set off looking to justify my resistor choice or find a different solution and saw a few places recommending to use a voltage follower after the divider to minimize the input impedance of the ADC.

But in the data sheet I found the "Recommended Impedance of Analog Voltage Source" to be 10 kΩ. This confused me because it seems to contradict the idea of minimizing the input impedance.

1. First of all, is there a piece of the puzzle that I'm missing here, or maybe something I'm not understanding?
2. Would it be best to follow the suggestion of the datasheet and (assuming I'm interpreting it correctly) set up the voltage divider to have an output impedance of 10 kΩ, following that the impedance of a voltage divider is $$\\frac{R1\cdot R2}{R1 + R2}\$$? Or should I rather seek to minimize the ADC's input impedance?

• That ZAIN spec does not make sense.. What is the part number? Aug 22, 2022 at 3:18
• I see nothing after a glance over the ADC specs that suggests that an input impedance well below 10 kOhm would do any harm. Aug 22, 2022 at 6:04
• Seems like a sloppy datasheet. The maximum source impedance would be 10k$\Omega$. Anything lower is fine (better, in fact). Aug 22, 2022 at 6:08
• @RussellH the part number is AVR32DA32. Aug 22, 2022 at 14:51

When the ADC acquires, it charges its sampling capacitor from the input during a fixed amount of time. If source impedance is too high, voltage on the sampling cap will not settle fast enough to reach an accurate value before the sampling time ends. This is why they specify a maximum source impedance. A lower impedance makes settling faster, which provides no extra accuracy if it already has settled to 1LSB, but it won't hurt either.

If you don't want your voltage divider to waste current, you can use resistor values that set the output impedance to the recommended value, so $$\R1 \| R2 = 10k\$$, something like 71.4k and 11.5k.

You can use higher resistor values, but then you need a filter cap at the input of the ADC, and its value must be higher than the ADC sampling cap multiplied by its resolution (4096 for 12 bit). The idea is that when the ADC connects its cap (say 2pF) to the filter cap (say 10nF) then the voltage drop on the filter cap will be less than 1LSB (1/4096). In this case the ADC sampling rate should be low enough so voltage on the filter cap has time to settle between acquisitions. This is useful for battery powered systems.

The input impedance of an ADC cannot be minimized. It is not desirable to do so. It is fixed and often can be quite low (10k to 50k). I have seen high input impedances > 1M. It is desirable to minimize the source impedance, which is $$\R_{1}//(R_{2}+R_{Vs})\$$.

The input may also expose a sampling capacitor that will rely on current from the signal source. The information is usually specified in the datasheet, but I cannot find it in the AVR32DA32 that you are using.

I think the data sheet source impedance $$\Z_{AIN}\$$ is either miss labeled as an analog input impedance (AIN) or the description is incorrect as a source impedance.

If $$\Z_{AIN}\$$ is a source impedance it should be listed as a maximum value.

If it is an input impedance the typical (or minimum) value would be appropriate.

An assumption cannot be made here. As a maximum source resistance, the input impedance would be very high in order to resolve 12 bits with < 1LSB of loading. $$2^{n}<\frac {Z_{AIN}}{Z_{src}}$$

As an input impedance, the $$\Z_{src}\$$ would need to be very small.

You should ask Microchip to clarify. It can be measured but not with an ohmmeter. Need to measure voltage drops which means a test PCB. Yuck!

A second point is that the sample time can be made longer, in one of the ADCs registers, for high source impedances. This means that the charging current for the sampling capacitor does come from the signal input.

I speculate that the datasheet meant that the source impedance can be as high as 10k. So in your case an increase in sample time may be required. Your voltage divider is probably ok.

You can increase $$\R_{1}\$$ and $$\R_{2}\$$ if you want to reduce loading on your voltage source. You will then increase the sample time to make it work.

Try to keep the source impedance low and the ADCs input impedance high. If either cannot be met then include a unity gain buffer in between them.