BQ76920: Understanding of Balancing adjacent cells

Hi Paul,
It is confusing, and perhaps the wording is not the best. Remember that the cells are in series, and that the circuit will have both voltages to VSS (single ended voltages) and voltages across single cells or between pins (differential voltages). Of course the single ended voltage is a differential voltage with respect to the same point, VSS.
If the nominal cell voltages are 4V as in the example, the nominal voltage for each input VCn with respect to VSS when not balancing is 4V x n where n is the cell input. So VC2's nominal voltage is 8V. VC3's nominal voltage is 12V. VC4's nominal voltage is 16V. When a cell balances, the sense input at the top and bottom of that cell are pulled together internally. If both cells 5 and 4 are balancing, VC5 and VC4 pull together, but VC4 and VC3 also pull together, so they are all 3 pulled together. The average point of these voltages is the cell 4 potential or 16V above VSS.
It may help to draw out the voltages on a graph. In the section you reference, you will probably find that some combinations produce voltages which might be acceptable, others will not. The recommendation is not to balance adjacent cells.

Hi Paul,
The cell balancing current is low, it will not completely bypass the cell, just reduce its current so that it does not charge as much. If you wish to stop charge you can do than, then alternate balancing between the cells with higher charge.

Hi Paul,
There are many algorithms which could work for balancing, select one which is suitable for your system. The one you mention should work fine, it will have energy loss since you are stopping charging, but if the higher cells have reached their upper limit charging should be stopped. Cells 2 and 4 could be balanced simultaneously in the scenario above, but that would be more complex and put more heat on the board. One cell at a time is certainly acceptable.
One question may be why the cells became so imbalanced. If cell 1 has much greater capacity it may be operating over a range of 3.2 to 3.8V while the other cells operate over 3 to 4.2V. Cell balancing to 3.8V may not help run time. If the cells were matched in capacity but at different states of charge, you would expect the cell at 3.8V to limit the discharge of the battery from UV and the battery would have a limited useful capacity. Cell balancing to 3.8V would reduce the charge in the battery that cycle, but next cycle the battery should have longer run time since the cells would be approximately matched in state of charge and any balancing adjustment would be small.
Normally it is best to assemble matched cells and have the battery balance system maintain rather than make large adjustments. You probably do want your balance algorithm to be able to handle large offsets, but at some point you may want to determine that something unusual has happened to the battery and you may want to stop its use.

Hi Paul,
Best is subjective. Often best is what meets the need, you can understand and accomplish within the space, cost, and time constraints. Active balancing is possible, you will find the EMB1428 and EMB1499 chipset if you search on TI.com. You will find various papers on the internet.