I am looking the STM32L443CC Datasheet (DS11421 Rev 5 October 2019) and trying to figure out how you come up with the values for Table 66 Rain.

[KiptonM]

Doing my due diligence, I did find someone else asked the question in 2019 but there was no response.

https://community.st.com/s/feed/0D50X0000BaM9PoSQK?t=1669134764953

What it did show was the equation, but not where he found it, and no description of the variables.

  

I am guessing CADC is the capacitance of the ADC. Which I am guessing is in Farads, and the datasheet says it is 5 pF.

I am also guessing that N is the number of bits of Resolution, which is 12 in my case. But I do not understand why there is 2 added to it.

Is fADC the clock frequency of the ADC? The table uses 80 MHz

I am guessing RADC is the internal ADC resistance. which is not specified in the datasheet like the capacitance is.

Then there is k. Which I am guessing is sampling time. Do they take off the .5 to make it maximum? Seems like the wrong direction.

In any case it does not work. Also the datasheet does not specify the difference between the fast and slow channels. What is different in the HW?

[Tesla DeLorean]

I think it comes from the STM32F4

https://community.st.com/s/question/0D50X00009XkXk3SAF/stm32f4-adc-impedance-calculation

This from DM00102166 STM32F401xD STM32F401xE Data Sheet Rev 3

"The formula above (Equation 1) is used to determine the maximum external impedance allowed for an error below 1/4 of LSB. N = 12 (from 12-bit resolution) and k is the number of sampling periods defined in the ADC_SMPR1 register."

AN2834

https://www.st.com/resource/en/application_note/cd00211314-how-to-get-the-best-adc-accuracy-in-stm32-microcontrollers-stmicroelectronics.pdf

[waclawek.jan]

https://community.st.com/s/question/0D53W00000hObQkSAK/radc-of-stm32g473

JW

[KiptonM]

I seem to have missed the part that explains why my answer is not even close to their answer. It looks like I figured out the identity of the variables correctly. So the answer should match. What am I doing wrong, or is the datasheet wrong?

So the 1/4 value explains the N+2 in the natural log term. That I understand why.

But I still do not understand why my value does not match their values if I use the same inputs.

[AScha.3]

> Also the datasheet does not specify the difference between the fast and slow channels. What is different in the HW?

more switches/resistance in path from pin to adc input.

 

[KiptonM]

So they are saying the equation is wrong since it does not account for those other variables.

[KiptonM]

Something has to cause the sampling rate to be different. Most likely lower resistance or lower capacitance to make the sampling rate higher. But no details.

[waclawek.jan]

What's your question, anyway?

For your application, just follow the table in DS.

The formula is okay, except that - as Igor Cesko said in the thread I linked to - RADC is not a constant. Most probably dominated by the mux's FET resistance, that also explains the "fast" and "slow" channels - there are only that much big (wide) transistors providing for the "fast" channels, to spare silicon area. Instead of giving a characterization of these transistors for all possible temperature/supply (=gate) voltage/BOOST on/off/ etc., which then is almost impossible to explain to the users, they decided to simply give the worst-case application-relevant RAIN value.

JW

[KiptonM]

The question is what is the real equation? The equation does not match the table. I cannot find the values that make the equation work. Which means something is wrong.

I can understand if they use worse case but give me the numbers or equation used so I can verify. I have found too many documentation errors to trust anything from any semiconductor company datasheet blindly. If they provide the equation, then they should provide the numbers they use. And someone else said temperature was also a factor. Temperature is not in the equation. So did they account for that in the "worst case"? That could explain why I cannot make the numbers work.

Today I am reporting a typo in the L6474 Programming Manual. Mistakes happen. I make enough mistakes on my own, to not have to worry about something wrong in a datasheet. As Reagan said, "trust but verify".