MPC9808 breakout board - deciding on pullup / pulldown restistor values

[alcs2000au]

Hi guys

I was wondering why in so many schematics always 10k ohm pullup / pulldown resistors seem to be used. The reason I am thinking about it is because the MPC9808 is a high precision temperature sensor and one would want to minimize any potential self heating. Any recommendations for the circuit below? Esp32 board with 3.3v will be used. SMT dip switch so the address can be set on the board. Thanks.



http://ww1.microchip.com/downloads/en/DeviceDoc/25095A.pdf

[Paul Moir]

Are you concerned with the resistors heating the MPC9808?  If not they're value is pretty irrelevant to the input current since that'll be the <5uA supplied/sunk by the input pin no matter what. 

If you are talking the resistors, then yes they can be much higher.  Hauling down 3.3V to 0.3V with 5uA max is is 600k so technically anything under that will work.  10K would be habit, eg you'd need this for old TTL LS, and noise resistance.  But your resistors don't need to be near the MPC9808 thermally I would think.

[alcs2000au]

Thanks, that already makes it a lot clearer!

It was this part of the datasheet that got me thinking about it:

[Paul Moir]

Oh yes, I understand what you mean now.  I was focusing on the address lines since they're the only ones with pull-downs.  I don't understand why you have series resistors in SDA and SCL; I was assuming they were for some EMI or protection reason but I now think you meant them to be pullups?
When the MCP9808 talks back to the EP32, it'll have to sometimes pull down a pull-up resistor on SDA.  The value of the pull-up resistor is driven by the bus capacitance and how fast you communicate with the MCP9808:
http://www.ti.com/lit/an/slva689/slva689.pdf?ts=1587934417753
(TLDR - go to figure 3).
If you can communicate slow enough with the MCP9808 and keep it close to the ESP32, you could have a pretty high value for the pull-up resistor.  Note that the pullup resistor only heats the MCP9808 when the MCP9808 is sending a zero on the i2c bus, which can be pretty infrequent.  Say you figure you can get  away with 10K.  You'd think that would be 33uA which wouldn't be fantastic.  But since it's only actually sending a zero say 1% of the time, it's only "0.33uA" from a thermal perspective.  Obviously this is dwarfed by the 200uA the MCP9808 consumes while running, but I expect you'll have it in sleep a lot if you want to minimize self-heating.