More detailed specification of SAADC input impedance?

[snygaard]

I’m searching for some more detailed information of the SAADC input impedance.

Please see attached file for details and questions.

Nordic Semiconductor question 20190315.pdf

[Carsten Wulff]

a) The R_INPUT modells leakage path's, for example, leakage of ESD clamps, thus this is a hard question to answer, there will be device to device variation, and large temperature variation (lower resistance when hot). What we've measured is in excess of 50 M Ohm. In general, we're here into leakage currents of nA, 

b) R_INPUT does not change with gain setting. For the other settings I think it depends on what you mean by "change". During sampling, the network of transistors connected to the pad will be different from the transistors during idle, so there is a theoretical change in the amount of leakage (pA). R_INPUT is not a physical resistor, but rather a model of all the leakage paths from the pad, to VDD, and ground. 

c) The value for R_ladder is given for typical process, however, this could vary up to +-20 % for a typical silicon process technology

d) C_sample will depend on process variation (device to device), which could be up to +-20 % in a typical silicon process technology, in addition the C_sample changes with gain setting.

Maybe you could elaborate on your source resistance? It would be helpful to understand your application. 

[snygaard]

Thank you for the input.

I'm using an ADC as battery monitor using a simple resistor divider and want to predict the obtainable precision of the voltage measurement. The nRF52810 is running at regulated 1.8V and during battery monitoring single ended ADC input and internal reference is used. I was looking at this case for inspiration:

https://devzone.nordicsemi.com/b/blog/posts/measuring-lithium-battery-voltage-with-nrf51

and was wondering how Mohm in resistor divider is compatible with ADC datasheet values of Rinput in the same Mohm range.
For nRF52810, the datasheet says "Rinput >1Mohm (typically)".
If using an equivalent source resistanse of e.g. 400k, it makes a significant difference of the ADC value sampled, if the equivalent Rinput is 1Mohm or 10Mohm.
I expect you to have a specification stating a typical and a guarenteed minimum value of Rinput over the temperature range. Else the precision of the ADC measures cannot be predicted.

[Carsten Wulff]

a) The typical condition (R_INPUT > 1 MOhm) does not cover temperature, process or voltage variation. The typical operating conditions is 3.0 V and 25 C, as specified in Recommended operating conditions

b) See page 10 in https://www.nordicsemi.com/-/media/DocLib/Other/Product_Spec/nRF52832PSv14.pdf?la=en. There you'll find the following text

"Electrical specification tables, containing performance data which apply for the operating conditions
described in Recommended operating conditions on page 20."

That means that if the el-spec table specifies min/max it covers the full voltage range and temperature range. 

[snygaard]

a) So R_INPUT can be even lower than 1Mohm over operational range?
Can you provide me a minimum specification of R_INPUT (over temperature, process and voltage)?

b) For me it is not obvious that the text "Electrical specification tables, containing performance data which apply for the operating conditions described in Recommended operating conditions on page XX" means the following:
- Typical values is valid for nominal recommended operating conditions.
- Minimum and maximum values are valid for the recommended operating condition range.
I believe I above have tried to mention that enterpretation as an option, but anyway, if thats the general way to read your datasheet, I have learned that now.
Sometimes communication can be difficult.

[Kenneth]

Hello,

The R_INPUT value is a question that pop up from time to time when reading the datasheet, since it does state >1Mohm as a typical value, which initially may seem very concerning. The good news though is that it is much higher, and I have still not seen anyone raise a concern after going to production or show any measurement that give any indication that the leakage current is higher than the few nA range.

The explanation behind the >1Mohm value is that there has not been any simulation on the bias/leakage current for the input, simply because it is very, very low, and it is not an issue. If you need a number for your calculations, you can safely presume that worst case you will need to bias the input pin with up to 0.1uA (but for all practical purposes it will be 10-100-1000x lower than this, even worst case I doubt it will ever go above 10nA).

Best regards,
Kenneth

[snygaard]

Thank you for an explanation of the ">1Mohm" statement.
You shall consider changing the specification in the next datasheet revision.

[skvery]

10 nA is 100 M and to get 12 bit accuracy the source impedance should be less than 24 k and for 8 bit less than 400 k.