In the previous post, "Why do I get funny readings from my ADC", I wrote about analog inputs which generally feature a sampling capacitor and a solid state switch network that connects it to the input and then to the conversion circuit. As it turns out the analog inputs in many of those Arduino class boards are actually a little different. They almost always have a large resistor in series between the input and the switched capacitor network. This reduces the current draw at the cost of taking more time to charge the sampling capacitor. The manufacturers may have intended this as a kind of "fool-proofing". But that extra resitor, placed internally in front of the switched sampling capacitor, also "expert-proofs" the analog input. It prevents us from being able to provide an appropriate, low impedance front end circuit with adequante charge or current source for the sampling capacitor.
In this page linked above, we described kickback and sampling window issues associated with the switched sampling capacitor network, and then we described what circuits work as a good front end for that kind of input. In short, that page describes how ADC's are supposed to work and how to use them. In this post we will show you how things work with that extra series resistor that the manufacturers built into the input. The lesson is that if you are making careful measurements for your research, you may want to avoid those internal analog inputs and instead use an external ADC, as we did for the boards we created for the SPI instrumentation. project.
Okay, lets look at some examples. Following are some spice models of those Arduino, Teensy, etc., inputs connected to a thermistor resistor voltage divider network.
Teensy 4.0
The MCU chip for the Teensy4 is an IMXRT1060 family chip by NXP. Here is the equivalent circuit from their datasheet, IMXRT1060CEC Rev. 3, 03/2022, page 64. Notice the resistor in front of the switch.
The input resistor in the SAR is actually one of a set of resistors. The resistor and conversion clock, and samping period, are selected by control registers in the MCU. For the demo, I chose three combinations from low and high values for the input resistor and the fastest and a mid range value for the sampling period.
And here is fast and fast. As you can see, the sampling capacitor almost reaches the input voltage by the end of the sampling period.
Here is slow and fast. This is even less good. There is a large error in the voltage reading.
And here is slow and slow. This works okay, but it really slows down the adC.
And now, lets try an opamp buffer with that. Let's stat with fast and fast. As you see it is also not so great, the charging curve has to be a lot faster for 16 bits of accuracy.
The takeaway so far is that the series resistor built into the analog input turns any low impedance source into a high impedance source. An ideal voltage source, zero impedance and infinite current capacity, will not make the input better than the time constant set by the internal RC.
UNO R3
Now I promised you a model for the UNO R3, here it is. They are only 10 bits, so precision is limited to begin with. But, we should at least be able get that much from it. As we see here, it is still not so great.
Concluding remarks
Recall that in our previous post, a simple opamp buffer was sufficient to give us a good voltage reading from the ADC, albeit with some spikes propagating back through our circuit. Adding a small RC as a charge reservoir, gave us pretty good performance without kickback. Here, because of that large resistor inside the chip, we are stuck. There is not much to do about it as a general solution. If you want to do precision work, you need to use an external ADC.
Be sure to check out the SPI instrumentation project. There we offer high precision circuits with analog-digital conversion on the board, and leverage the processing power of the an Arduino or Teensy to create flexible high-end instruments.