# Discrepancy on the ADC input impedance in the ATtiny datasheet 1..100k or 100M?

I found a discrepancy in the ATtiny datasheet. While the section explaining the ADC shows a simplified schematic with a 1...100 kOhm series resistor, the characteristics table shows 100 MOhm.

The ADC is optimized for analog signals with an output impedance of approximately 10 kΩ or less. [...]

With that max. 100 kOhm series resistance in the sample/hold path it makes sense to me, that they say that the source impedance should be about 10 kOhm. But with the 100 MOhm impedance from the characteristics table I'd say requiring 10 kOhm source impedance is unnecessarily high.

Table 21-8. ADC Characteristics, Single Ended Channels. T A = -40°C to +85°C
...
Table 21-9. ADC Characteristics, Differential Channels (Unipolar Mode). T A = -40°C to +85°C
...
Table 21-10. ADC Characteristics, Differential Channels (Bipolar Mode). T A = -40°C to +85°C
...

What is the reason for that discrepancy? Or can I assume the 100 MOhm is a wrong value and should actually be 100 kOhm typical?

• The actual discrepancy is in 1..100k range, versus the recommended 10k analog source impedance. I would root for a linear buffer with 20-50 Ohm output impedance in front of the ADC pin. Feb 2, 2017 at 22:45

The 100 Mohm is a static value which pertains to leakage. The active value you first quote is when the ADC is sampling an input signal. In effect, both are in parallel.

If you read further on, they explain why they recommend less than 10kohm source impedance. It is so that the sampling capacitors can charge up in a negligible time. If you use a higher source impedance, they will take longer to charge, and initial readings will be low. The ADC input does have a >100Mohm input impedance at DC.

• Sure, I read that sampling part. Any explained it well with "In effect, both are in parallel". There was my misunderstanding. So, when sampling is done (it's "holding"), the input impedance is 100M. Feb 2, 2017 at 20:55

It's helpful to understand a bit about the data sheet generation process at a large silicon vendor.

The technical editor responsible for the data sheet typically gathers information from Design Engineering, Test Engineering, Applications Engineering, Development Tools Support, Marketing and possibly external early-adopter customers and consultants.

Electrical models in the data sheet are often simplified models of complex behavior, and the parameters in them are the results of many hours of measurements, calculations and production parameters gathered from different teams.

A lot of the information is typically "hot off the press" and the data sheet editor is under a lot of pressure to release the initial copy.

Before first release, the document is circulated for sign-off review to design, test, applications, devtools, marketing and finally, legal.

The point is that many disciplines provide input into the document, and good datasheet editors are miracle workers in how few errors slip through with so much data from so many sources.

Conclusion: Discrepancies happen all the time, and you should not be surprised by them (or blatant errors) in datasheets - just read the legal disclaimers in them.

Also, when you download a datasheet, always look for the errata on that datasheet and make sure you read it first.

When you do find a discrepancy you can not or do not want to solve, the vendor's support system can always be used to figure out the facts, but it typically takes time and can set back your design schedule. That's why it's good practice to design with popular products that have been in the market for 1-2 years - other early adopters have found/fixed most of the bugs in the data sheets for you!