If I may hop on to this old post and add to the answer already here,

1: abs2 and abs1 in TopSpin are for automatic baseline correction, not necessarily for solvent suppression. Baseline correction in nmrPipe is performed using the 'BASE' (linear) or 'POLY' (non-linear) functions. The POLY or SOL functions can be specifically applied to the time domain prior to Fourier transform as Evgeny Fadeev has stated earlier for solvent suppression.

2: 'First' is the level of the first contour level, and 'Factor' is the distance between contour levels (i.e. a factor of 1.3 means the next contour level is 1.3 times more intense.). In my group, we usually use a factor of 1.4 (~ square root of 2), but you can use whatever value looks good to you. Determining the noise level in nmrDraw (using 'T') will automatically set the 'First' contour level just above the calculated noise. This usually gives a decent spectrum. There is no 'have to' set these values, it is what you want them to be to get the best looking spectrum.

3: I assume by 'square layout' you either refer to looking only at the left side of the spectrum (using the EXT function, described earlier), or do you mean the scaling of the dimensions? If you are converting this into Sparky or some other program, and you use non-Sparky axis titles, then Sparky will assume all the dimensions are 1H. You can check this using the 'ucsfdata file_name' command in a unix shell. The ucsfdata man page describes how you can change the values. Once you fix that, then the spectra will look the way you are used to with appropriate scaling.

4: You can phase shift an already Fourier transformed spectrum, as long as you haven't deleted the imaginary parts. If you have, then you need to recreate them using something like a Hilbert transform (HT function), to then process them. But, in general, what you want to do is to process the 2D spectrum, adjust the phase using 1D planes in nmrDraw, and re-process the data with the (added) phase correction values you find. If you are processing 3D or 4D data, it is best to find the phase values for the series of 2D projections (same as TopSpin) rather than attempt to phase the entire spectrum.

As a final note, I will also specify that the phase values you find in TopSpin will be the same in nmrPipe, except with opposite sign. This will be why when you use the same phase values from TopSpin, you will end up with double-dispersive peaks, and that's why it looks different!

The biggest difference that in nmrPipe you get to write scripts, which are plain shell scripts composed of mostly chained calls to nmrPipe (chained through Unix pipes).

For the solvent suppression nmrPipe has these commands:

nmrPipe -fn SOL #low frequency digital filter
nmrPipe -fn POLY -time #polynomial baseline correction in time domain

For the phasing, there is a way to run scripts from nmrDraw (the companion display program) but for me typing the phase values directly into the script works better. I have a default script where phase parameters are set to 0, then I run the script, adjust the phase in nmrDraw, copy the values into the script and re-run it. Then usually some more adjustment will be needed, most likely just small corrections where it's easy to add/subtract the values and re-run the processing again.

For extraction of a region, use EXT (example for 5-12ppm in one dimension, add a similar line for the second dimension):

nmrPipe -fn EXT -x1 5ppm -xn 12ppm -sw