PROCESSING YOUR DATA
 
Phase corrections
The phase of an oscillation refers to the angle which corresponds to a particular point along the rise and fall of the function. For example, the cosine function, as shown below, is at a maximum (y = +1) at an angle of 0 degrees, at a minimum (y = -1) for 180o, at an intermediate positive value (y = +0.5) at 60o and 300o, at an intermediate negative value (y = -0.5) at 120o and 240o, and at nulls (y = 0) at 90o and 270o. A phase shift occurs when the cycle is moved to the right or the left. To see the effect of shifting a cosine function by 60o, move your mouse into and out of the area below:

In an idealized NMR experiment, all frequency components are at intensity maxima at the very start of the FID (time, t = 0). We could say that they have a phase shift of 0o. However, during a real NMR experiment, the instrument cannot detect the signal at exactly t = 0. The very strong rf pulses used for excitation overload the receiver, rendering it useless for a short period of time, called the deadtime. Since the FID components continue to evolve during the deadtime, they will all have different phases when the receiver finally begins to detect signal. Position your mouse over the illustration below to see a simulation of receiver deadtime:
Note that the phase shift varies with the frequency. In this case, the phases of the higher frequency component (light blue and green) s are shifted less than those of the lower frequencies (orange and red). This suggests that the phase errors will vary as a function of frequency.
 

If each individual component was observeable, as shown above, the phases could be corrected before the data was transformed. However, since the observed FID is the sum of the discrete frequencies, this cannot be done. Instead, phase corrections are performed directly on the spectrum when needed.

The FT of a signal with no phase shift yields a symmetric peak with no baseline twist. Such a peak is said to be properly phased:
 
An FID with a phase shift of 90o, on the other hand, will produce a peak with a badly skewed baseline, as shown below:
Such an error must obviously be corrected before the spectrum can be properly analyzed.
 

As hinted above, FID with more than one component will exhibit phase shifts that vary across the final spectrum. In practice, phase corrections are made by first performing a zero-order correction on the largest peak in the spectrum. Your goal is to produce a symmetric peak flanked by flat baselines on either side. If done manually, phase corrections are a trial-and-error process. With experience, you will be able to estimate an initial guess at the appropriate correction. Small increments are then added to or subtracted from this estimate into order to arrive at the proper correction.

The figures below illustrate this approach. In the first example, the phase correction is guessed to be 45o. If you move your mouse over this spectrum, you will see that this has improved the situation significantly. However, this value is clearly a slight overestimate, because the peak is still asymmetric, and the baseline still skewed. Furthermore, the direction of the twist is opposite to that in the starting spectrum (i.e., the right side is now higher than the left). In the second figure, the effect of subsequently performing a phase correction of -3o is shown. The resulting peak is now properly phased.
 

Because of the frequency dependence of phase errors, zero-order phasing is not usually sufficient to correct the entire spectrum. Peaks near the phased resonance may look fine, but those farther away will still exhibit some residual misalignments. Once the zero-order correction has been done, a first-order correction is performed on a peak far removed from the largest. Again, the aim is to produce a symmetric peak with flat baselines on either side.

The examples below demonstrate this process. The first figure shows a spectrum in which the two peaks clearly require different phase corrections. As you move your mouse over this illustration, you will see the effect of phase-correcting (zero order) the larger peak. The appearance of the smaller peak has changed, but it obviously needs further work. A estimated first-order correction of +20o clearly overshoots the appropriate value, as illustrated by the second set of figures. The third set of reveals a spectrum in which both peaks are properly phased, after a subsequent first-order correction of -10o.
Most spectrometer systems now include automatic phasing routines that can do these operations with minimal operator intervention. However, in many cases, the spectrum may benefit from some small, additional manual corrections.
At this point, all the peaks in the spectrum should be properly phased. If they are not, you may have some artifacts present in your spectrum. For example, you might have a glitch from an uncorrected dc offset in the FID, or a spurious peak from foldover caused by a too-small spectral width.