Happy new year to you too! Please see below for some answers:
we use methanol temperature calibration to make sure the relaxation measurements are taking place at the same temperature. Interestingly, I see significantly diverting signals in my HSQCs / TROSYs which I believe are due to temperature differences. I don't see diverging signals between T1, T2 and Hetero-NOE measurements, so at least the data at one field seem to be consistent in any case.
There are so many factors involved with temperature control that nothing would surprise me. Just note that the temperature must be identical for all relaxation measurements otherwise the dynamics of your system, especially the global tumbling, will be different in each experiment and there will be no way of analysing the data correctly (the dynamic changes are not linear).
One strategy to overcome this was to * record a HSQC/TROSY at different temperatures at the different spectrometers (conveniently by using Brukers multi_zgvt),
Did you use the trosy experiment for relaxation measurements? Otherwise it shouldn't matter too much.
* reference the two sets of spectra to a residue signal which seems to be at exactly the same position regardless of temperature
As long as the experiments are not for a dynamics analysis, then this is ok. NMRPipe can do this automatically if you know the real sample temperatures of the experiments.
* select the temperature where the NH spectra are nearly 100% identical
If temperature calibration is important, then you should run the experiment on methanol or ethylene glycol to calibrate. Otherwise if the spectra are the same but just shifted because of temperature, then it's not so important.
What is your experience with calibration strategies? Do you see any diverging NH signals at different magnets?
I have seen cases of large temperature divergences, and cases where the temperatures are all the same. It really depends on the hardware setup (see http://www.nmr-relax.com/manual/Temperature_control_calibration.html). The only thing that is important is that you use the same sample temperature as measured on MeOH or ethylene glycol for each relaxation experiment. You can modify the pulse sequence to preserve the OH magnetisation and then see the 1D for any fid, or run a quick 1D after running a short version of the experiment. Regards, Edward