Water suppression, whether in NOESY or simple 1D, is always an exercise in compromise. The "best" method will depend on your sample, the pH and how large your molecule is.

Exchange:
If your are trying to preserve exchangeable protons (NH or OH) you have to be careful of saturation methods such as "presat". Even watergate (without flipback) does saturation. Wet does saturation, but as with watergate without flipback pulses, the exchangeable proton magnetization is recovered during the acquisition and relaxation delays. The important difference is that presat is on during the relaxation/mix periods and will attenuate signals because of "saturation" and cross-relaxation (noesy). The sample pH will affect the exchange properties. The nature of the sample (folded protein vs. extended peptide) will determine how many exchangeable protons are exposed to the solvent, thus influencing the choice of suppression method.

Jump-Return will have the least amount of compromise by it has the problem of a non-linear excitation, restricted bandwidth of full excitation, and poor overall suppression. But it does let you see how much magnetization is possible.

Effect of Molecular Size:

If you are studying large proteins you can use gradient methods without too much problem with diffusion (watergate or excitation sculpting- dpfgse, or double pulsed-field gradient spin echo), but T2 becomes an issue. Long echo times permit strong gradients, but each echo loses signal because of natural T2. This means, typically, that the spin echo time in watergate is limited to a few millisecons. If the shaped pulses in soft-pulse watergate are more than a few msec long the echo time T2 losses might be too severe. Using 3-9-19 watergate removes the need for shaped pulses in the spin echo delay, but it gives a non-linear excitaion profile).

Wet just has a single pulse for excitation and so does not have a T2 problem. The Wet portion is done only during the relaxation delay and mix period where the desired magnetization is along +Z, so T2 is not an issue. However, it is a saturation experiment so enough relaxation delay must be used to recover desired magnetization.

Excitation Sculpting (dpfgse) methods give very good suppression, and are not saturation methods, but the echo penalty is twice as much. So if you are looking at small molecules or peptides where lines are narrow, this is an excellent method.

Soft-pulse watergate with selective pulses of 1 msec can also give good water suppression. However, the best suppression is only obtained when the soft pulses and hard pulses are accurately calibrated. You can optimize with respect to the offset as well.

Overall, 3-9-19 is the "easiest" method (with flipback)but has the above pitfalls. Wet has the fewest compromises.

There are many other methods for water suppression that can be investigated in order to optimize for any one sample.

George Gray NMR Consultant for Agilent Technologies