Connecting nets

Thanks.

Moving ground to the load side of the sense resistor does not fix the potential charge/discharge path from PACK+ to the communication ground (GND) which bypasses the (Q2, Q4) protection FETs or the communication lines. That move will also give SRP the larger voltage swing with current flow which is not the recommended operating condition.  Moving the ground of the electronics to the PACK- with the low side switch will not work since most signals to the AFE will go out of range and damage the part when the protection FETs open. The methods mentioned before are options, you must chose one suitable for your application.  One of the straightforward approaches is to isolate the SMB interface using a part such as the ISO1540.  It must have power for each side though and this adds to the cost and load on the battery.

It is difficult to measure low current through the 1 mOhm sense resistor. The sense resistor should be sized to keep the dynamic range of the battery in the dynamic range of the inputs and available settings.  If you have a low current application an the battery electronics is a significant portion of the load, you should be able to move the GND to the load side of the sense resistor so that the system can measure and gauge the pack electronics current.  Remember that the recommended range of the SRP is small.  Although you can measure the current, the protection FETs can't turn it off.  Although it is larger than self discharge current, normally the electronics current is accepted as a quiescent current and approximated with self discharge current.  To reduce the electronics current the AFE can be placed in ship mode which reduces the current, see the ISHIP in the datasheet.  When the bq78350 is powered by the AFE it will be turned off.  If you go to ship mode you will need a way to boot the AFE again when the battery is desired.

Powering the bq78350 with a separate efficient supply may be practical, but the AFE is not, the AFE should not operate at nominal voltages below the cell inputs except REGSRC.  If REGSRC is supplied from an efficient supply there is still the LDO to REGOUT and the designer would need to decide if that is a sufficient improvement or if both REGSRC and the MCU load should be provided from efficient supplies.

How should we deal with possible swing?

PACK- and BATT- will differ only by mV when the protection FETs are closed.  When the protection FETs open the signals may be many volts different depending on whether a load or charger is attached.

Q1:  Referring to the FET figure you included above.

Q2: Right, translation is OK as needed within the specifications of each part in the translation chain.  the concern is when the FETs are off.  For example when the battery is at 40V but it protects, perhaps it got too hot, if there is any load PACK- pulls to approximately PACK+ or 40V from the battery standpoint.  If you are referencing the PACK-, then you see PACK+ at approximately 0V and SMBC and SMBD at approximately -35V.  The ISO part allows you to talk to the battery to ask it what is wrong if you have an alternate power source for your system processor.  If you have the isolator but not an alternate power source, your system processor may have lost power when the battery protected.

If you don't have an isolator and an alternate power source for the system processor and it is still attached to SMBC and SMBD, then it creates a leakage path to continue to discharge the battery through the processor and SMBC and SMBD lines.

Q3. Perhaps I placed that poorly, see paragraphs at Q2 above.

Q4. Perhaps poor word choice.  If GND is referenced at the SRN filter connection side of the sense resistor, the SRP side of the sense resistor and the SRP pin will move positive or negative with respect to GND.  The amount will vary with the current and the size of the sense resistor.

It may be informative to purchase an EVM and demonstrate the part operation in various modes and see the signals move.