Filtering input of MAX9611 high side current sensor

[aberco]

Dear everybody,
I'm running in a hardware question regarding a power-meter project using a integrated high side current sensor.

I have decided to go for the MAX9611 which is a "plug and play" I2C device, and have completed the project which is running nicely. I have followed religiously the datasheet so that the A/D conversion is computed in software and displayed exactly as specified. I used a 10mOhms shunt resistor wired properly (2 fat traces for the current path, and 2 fine traces running from the resistor to the MAX9661). Prototype all mounted on a PCB.

1) When the input of the shunt are left floating, values measured for I and U (common mode voltage) are erratic. What would be the best way to clamp the shunt in order to get a steady 0 when left floating? I tried putting 1MOhms to ground on each terminals of the shunt, which fixes the readings for U, but not really for I.

2) When grounded, I do not get a true "0" output on the ADC. I could fix that using an offset in software though, if that is normal. I don't know yet if my readings compare well to a calibrated multimeter.

Any help would be appreciated

[aberco]

A little more brainstorming on my project which seems to give quite good readings whatsoever. I have attached a picture of the project measuring current through a 100ohms load.

I have looked at possible filtering on the shunt in order to get more stable readings. The datasheet only show direct connexion of the MAX9661 to the shunt but adding a filtering network could possibly help in having more stable readings when the consumption of the load is fluctuating. Only problem would be to carefully match the values of both Rf, which may otherwise degrade significantly the measuring accuracy.

Also adding D would be important for inductive load, but would limit low voltage measurement capability

Have not yet managed to fix readings that are not zeroed when load or input are left floating.

[sean87]

Please don't ban me for bringing up this thread  

For keeping the reading at 0 when floating, one solution is to get the raw reading out of I2C when the sense terminals are grounded and see the raw ADC output, then compensate for it in software.

It is not possible to make it in the analog world exactly 0 (AFAIK).