Removal of coherent noise

In pulse NMR, there is always noise coherent with the pulse sequence because of the transient nature of the experiment. This is in addition to the usual random noise as well as other systematic noises such as the 50 or 60 Hz hum from the ac source which are not synchronized with the pulses. These non-synchronous components of noise can be reduced either by appropriate filtering or by multiscan averaging or both. In this section we will deal with noise coherent with the pulse sequence. Even if such noises were very small, they will add upon signal averaging (just like the desired signal) and may, therefore, be troublesome in experiments where very many FID s are accumulated due to inherently poor S/N. 

The best solution to any problem, of course, is to treat it at the source. Much of the incoherent as well as some of the coherent noise are due to cross-talk between different 

Since the noise coherent with the FID is usually a ringing of some kind generated by the high intensity rf pulse, such ringing must be suppressed as much as possible. Some of the major sources of ringing were discussed in section VI.B.5. 

The most obvious way to cancel unwanted signal coherent with the FID is to accumulate the total signal and the signal without a contribution from the nuclei in the sample and subtract them. The sample contribution may be removed in several ways. The sample may be physically removed the magnetic field may be changed to go off resonance or the magnetization may be saturated by some pulse sequence. The 

The more common of the two methods reverses the phase of the rf carrier for exactly one-half of the sequences. Then the signals from the two halves are subtracted. Reversing phase will turn the desirable FID upside down so the subtraction will simply accumulate the two halves together. If the spurious transients are not coherent with the rf, they will not respond to the phase reversal and will cancel in the process. The phase reversal can be performed on successive sequences with concomitant juggling of the data, or N signals can be 

---