See my cad pic,
Yellow indicating where the reflector tape is mounted.
90 degrees on, 90 degrees off, 90 degrees on, 90 degrees off, as it spins.
With the low resolution counter method in bang-bang, the difference between (I am going from memory from a couple of weeks ago) in a 10 ms loop, it was 9, 10, or 11 counts, which equated to 3000rpm, 3300rpm, 3600rpm. We realized that if the tape did not start at 180 degrees apart, this was an issue and an error. As we increased the loop time to get more counts, we were reducing the motor output changes which increased the error.
We are using the "gear tooth" mode, and a timer avg over two counts, which is one revolution. This really reduced the error in not placing the reflector tape exactly 180 degrees apart.
Does that make more sense?
Edit, Banner part number DS18VN6LP