If you try to lock, and then turn the sweep on, can you see a ringing signal? If you see only a weak signal you might get somewhere by increasing the lock power as suggested.

If your system is capable of gradient shimming, i.e. it has a lockswitch or 2HTX board, or is a new system with LTRX, then you can easily acquire a 2H spectrum and see if there is any signal at all and how broad it is, and also check the 2H tuning.

To do this, make a new experiment in an even expno, read the parameter set gradshim1d2h, and change the pulse program to zg2h. Do getprosol to read in the 2H pulse, and change AQ to something sensible (a few seconds). Set SI to half of TD for processing. You can then acquire the spectrum with the command xaua, (don't use zg), which will deal with turning the lock off and minimising the lock power to avoid the lock pulse breaking through into the spectrum.

You can also check the 2H tuning at this point with wobb - it won't be perfect, but if it's some MHz off tune, that may indicate that detuning is the problem. When you're done, iexpno to an odd numbered expno and acquire a spectrum again with xaua, which will sort out turning the lock back on and so on.

One thing to add to Pete's good advice. There is a further problem at high temperature - rapid sample convection can interfere with both lock and shimming. If the lock is very unstable when you eventually find it, it may be worth using a short sample (e.g. Shigemi) or (better) a smaller ID NMR tube to reduce convection. If you get a good lock but can't gradient shim, changing the gradient shimming pulse sequence can help (JMR 154, 325).