For a very good discussion of Linear Prediction, Manimum Entropy and all other aspects of NMR data processing I recommend:

"NMR Data Processing" by Jeffrey C. Hoch and Alan S. Stern, Wiley-Liss (1996). ISBN 0-471-03900-4

It is a bit dated, but is is oterwise quite excellent, and quite readable as well.

Hi, Sri, Forward Linear Prediction will increase resolution in the same way as zero filling does, but it avoids wiggles, which is an artifact you will get in the frequency domain spectrum when you zero fill an FID which is truncated and does not decay smoothly to zero (the FT of a step function is a sine type function which appears 'wiggly' at the foot of the peaks and will make peak picking more difficult and accurate integration impossible). As for the values you may use, you can multiply several times the number of points in the FID and then you combine this also with zero filling (once the FID has been linear predicted to a smooth decay to zero) - take a look at: http://mestrelab.com/resources/covariance-nmr/ (there is a section there with an example of LP which will help you understand a little better and which will show you the parameters we used in that example).

As for backward LP, this is typically used to correct corruption in the first few points of the FID, which is a common problem which will result in baseline distortions in the frequency domain (rolling baselines). Therefore, when you backward LP, you should predict the number of points which are corrupted. I am not sure what you can do in VNMRJ, but hopefully you can do the same as in Mnova, which is to visualize the FID, changing the scale to points rather than ppm, and then in that view you can see how many points are corrupted and backward predict them. If you don't have this capability in VNMRJ, try to increase the number of points you are backward predicting from 5, to 10, 15, etc. Typically, you would not need to predict any more than 20 at most.

Hi,

If you are using BioPack within VNMRJ you can easily set up LP parameters by using the Processing panels for t1 and t2. There is a button for automatic setting of LP parameters (based on the values of ni/ni2-assuming you've finished the data collection). This button will extend the data set a factor of two and also set the weighting functions for the extended data set. If you either are still acquiring the data, or wish to have more extended data, you can fill in entry boxes for how many increments you do have, and how many you would like to have in the extended data set. Then the "AutoLP" button will set the LP and weighting functions automatically.

If you are not using a BioPack sequence, you can still use BioPack to process data in the same manner. Just recall the data into an experiment with BioPack active. The processing panels will still work since they use standard VNMRJ parameters and commmands.

Regards, George Gray Agilent Technologies

Hello, not very easy to understand the question, but if you are asking - how many points to predict? Not too many, perhaps increase the total data size by 50%.

Linear prediction does not add any new information to the spectrum, only makes lines look sharper at the expense of adding artifacts.