At the core of Non Uniform Sampling and reconstructing of the resultant spectra is selecting which points to acquire and which to skip. Work in our lab has established the Poisson Gap Sampling Method (Hyberts et al, JACS, 2010) results in superior schedules with fewer artifacts during Forward Maximum Entropy reconstruction. Empirical evidence from our lab (unpublished) suggests this is also true for the Iterative Soft Threshold algorithm.
You can create your own schedules on your own computer using our Poisson Gap Schedule Java applet with instructions on this page. However, using the fields to the left on this page you can create a Poisson Gap Schedule right in the browser.
First, you want to select how many dimensions your spectrum will have. This number includes the directly acquired dimension. So for a typical 3D Triple Resoannce experiment, you will select 3D. Sampling Density helps select how many points you will ultimately acquire. It is the percentage of the entire matrix which is defined below. Obstained points is a specific amojnt of points you may wish to collect. If you put a number in here, it will calculate how much of a percentage of the matrix it is and put that into the Sampling Density field. Lastly we have a tolerance factor. Sometimes it is hard to get exactly the number of points you want so it is best to add some tolerance. 0.1-1% is good.
This field is very straight forward. Enter the number of points you want for an indirect dimension. The number you put here will be spanned by schedule points in a column matching the dimension number. That is, if you put 32 in the 1st indirect dimension, numbers from 0 to 31 will appear in the first column. If you put 64 in the 2nd indirect dimension, you will get nunbers spanning 0 to 63 in the second column. Its very important you have your order correctly, but the correct order depends on how you pulse sequence works. If you are doing a triple resonance experiment on a bruker machine, you almost certainly want 1st indirect dimension to be the number of complex points for 15N and the 2nd indirect dimension to be the number of complex points for 13C. Also keep in mind that we are talking about complex points here. The TD values in Bruker for exmaple are talking about total points (2x complex points) so putting 32 15N points and 64 13C points is like collecting a spectrum of 64 points in 15N and 128 points in 13C in the topspin software.
Here we can add a few details that control the type of output. Firstly, the Sinusoidal Weight factor. This controls where the sampling tends to concentrate. In simplest terms, a weighting of 2 (recommended) puts most samples towards the front of the schedule were you expect most signal (before evolution decay) and even in the case of constant time experiments, a weighting of 2 results in more points being collected in an area that is not so 'window functioned' to zero. Again, we suggest you use 2. A value of 1 will weight samples to the front and back of the sampling range. A value of 0 has no weighting.
See Value is used to initialize a random number generator. If you use 0 you will get a different schedule every time. This is the default and you probably want to do this.
Start Schedule at: allows you to have a schedule format where points start at '0' or '1'. Some pulseprograms/spectrometers use different standards. Most commonly, schedules should start at zero to 0 is the default.
Lastly, Random Output allows you to randomize the order of the samples so collection is randomized. The first complex point '0' or hypercomplex point '0 0' or '0 0 0' will always and should always be collected first. However there is no reason to collect points in order after that. In fact collecting randomly means you can check your spectrum by doing a quick reconstruction on partially acquired data and get the full resolution after only a few points. We suggest beginners collect in order at first and see how the hypercomplex points look until you are familiar with acquisition. Then collect randomly. Ordered collection is the default.
Old Browser Work-arounds: