¹The possibility of implementation errors in the complicated sensitivity calculations has been largely eliminated by the software self-test option, thereby making errors an unlikely reason for slow learning.