15.16.2  Errors 

The error is computed using DY during a convolution.  The NDY if present
is ignored in the computation, and it is set to DY when done.

No error calculation is done for inverse convolutions.