There is a bug in set-up of the input filtering. The function accepts a channel pair number, but this pair number is used as a direct channel number when setting up the filter.
Code currently in SDK 2.5.0 (and earlier)
/* Input filter available only for channels 0, 1, 2, 3 */
if (chnlPairNumber < kFTM_Chnl_4)
{
reg = base->FILTER;
reg &= ~(FTM_FILTER_CH0FVAL_MASK << (FTM_FILTER_CH1FVAL_SHIFT * chnlPairNumber));
reg |= (filterValue << (FTM_FILTER_CH1FVAL_SHIFT * chnlPairNumber));
base->FILTER = reg;
}
This is incorrect as the check in the code expects a channel number, not a pair number. The bits in the FILTER register are also for channels not pairs.
My proposed correction is as follows.
/* Input filter available only for channels 0, 1, 2, 3. Thus pairs 0 & 1 only (channels 0 & 2) */
if ((chnlPairNumber*2) < kFTM_Chnl_4)
{
reg = base->FILTER;
reg &= ~(FTM_FILTER_CH0FVAL_MASK << (FTM_FILTER_CH1FVAL_SHIFT * (chnlPairNumber*2)));
reg |= (filterValue << (FTM_FILTER_CH1FVAL_SHIFT * (chnlPairNumber*2)));
base->FILTER = reg;
}
This change is in-line with the rest of the code that converts the pair into a channel number when setting the register values.
