This question might have been better posted in the NMRPipe group. Normally, in NMRPipe, gradient enhanced Varian 2D data is converted with "-yMODE Rance-Kay". This actually runs an NMRPipe macro to perform gradient shuffling of pairs of adjacent 1D vectors during the conversion, but it is run "in background", and you don't see it explicitly: | nmrPipe -fn MAC -macro $NMRTXT/var_ranceY.M -noRd -noWr \ In an interleaved version of the gradient-enhanced data, there are two possible formats. In the case array="phase,IPAP", the innermost acquisition loop is the Rance-Kay phase encoding. So, adjacent pairs of vectors are Rance-Kay pairs, and you can convert this data with "-yMODE Rance-Kay". Then, later processing steps can use COADD to select or combine channels. However, when array='IPAP,phase' the IP/AP alternation is in the innermost acquisition loop. This means that manipulation with COADD must happen *before* gradient shuffling. In these cases, we convert with "-yMODE Complex", and then explicitly include gradient shuffling just afterwards, for example:

nmrPipe -in test.fid \
| nmrPipe -fn COADD -cList 1 -1 -time -axis Y \
| nmrPipe -fn MAC -macro $NMRTXT/var_ranceY.M -noRd -noWr \
| nmrPipe -fn SOL \
| nmrPipe -fn SP -off 0.5   ... etc ....
  -out dif.ft2

Also, one more tip: When dealing with this kind of data, it can be handy to extract the IP and AP channels separately ("COADD -cList 1 0" and "COADD -cList 0 1") so that they can be inspected first to find the optimum coefficient for cancelation of IP and AP signals. Then, use the stand-alone program "addNMR" to combine the two channels to form scaled sum and difference, for example: addNMR -in1 A.dat -in2 B.dat -c1 1.0 -c2 1.2 -out sum.dat -add addNMR -in1 A.dat -in2 B.dat -c2 1.0 -c2 1.2 -out dif.dat -sub