Coherence selection by PFG naturally results in P-type (echo - the sense of the frequency modulation is the same in t1 and t2) or N-type (antiecho - the sense is opposite) data selection depending of the gradient ratio used to dephase-rephase the desired magnetization. This procedure permits to obtain frequency-discriminated spectra in a single scan but the undesirable doubly-absorptive/doubly-dispersive lineshapes (so-called phase-twisted) of the signals prevent to present phase-sensitive data. 

In the P/N approach, two data sets which correspond to the N- and P- peak coherence-pathway selection are separately recorded by reversing the sign of one or several gradients. Pure absorption spectra are then obtained by combining the echo and antiecho spectra after reversing the w1 sign of one of the data sets. This approach generally affords a decrease of the signal-to-noise ratio by a factor of sqr(2) with respect to an equivalent phase cycled experiment in which both cosine and sine modulated data are simultaneously acquired. An analysis of the different sensitivities obtained from several HSQC experiments which use gradients has been made ( 94JMRA70-111 ).