Hello team,
My costumer Prodrive (platinum design house) has asked me for below. As I don't have practical experience with this, I ask you for your opinion.
----------------------------
For a project we are using the TMP175 temperature sensors. These have a user selectable I2C address with ‘trimode’ A[2:0] pins which also can be left ‘float’ for even more selectable I2C addresses.
We have many of these sensors on a product and need the additional ‘float’ selection on A[2:0]. Because of bad experience in the past we normally decouple these trimode pins to prevent crosstalk or noise influencing these signals, especially when an assembly variant requires these pins to pull-up/down or float, the short traces already provide an antenna function. We now used 100nF decoupling and discovered during measurements that the internal biasing is so low that it takes many seconds for the 100nF to charge to the desired level. Even with a 10Mohm multimeter the internal biasing is overruled to GND! Of course because of power consumption these input currents cannot be high.
Questions:
1) What is the internal biasing current/resistor divider used for the ‘float’ setting? Is this indeed the “leakage input current Iin” of max 1uA?
2) What are the ‘float’ voltage thresholds?
Options:
a) Leaving these pins really ‘open’ (no capacitor at all). We have bad experience with this because of noise susceptibility because of the short traces (for assembly options) attached to this pin
b) Use a smaller capacitor (1nF?). Then we need to know what the min/typ bias current or bias network is to calculate the acceptable settling time for this pin
c) Is it possible to provide a ‘hard’ mid-level with an external resistor network. Then we need to know what the ‘float’ detection voltage levels are for these pins.
What option is suggested?
----------------------
Hope you can help to give suggestions how to 'terminate' these floating inputs.