Bruker data and TopSpin conventions

NMRTools reads and writes Bruker raw and processed data, and is designed to give the same spectra as TopSpin. This page describes what loadfid does to raw data, how the digital filter is handled, and the conventions that differ between the two programs.

What loadfid does

On loading a fid or ser file:

• indirect dimensions acquired with States, States-TPPI or echo/anti-echo are recombined into an extra multicomplex order, so a 2D experiment has Multicomplex{Float64,2,4} elements; the scheme used is recorded in each dimension's :quadrature metadata;
• non-phase-sensitive dimensions (relaxation delays, CEST offsets, …) become X2Dim/X3Dim;
• the digital filter's build-up points stay in the FID, on a time axis that starts before t = 0 (see below);
• the t = 0 point of each Fourier dimension is scaled by 0.5 (change this with the firstpoint keyword), except in the direct dimension of baseopt data;
• the data are scaled by 2^NC, as TopSpin scales them, so that intensities after ft, which applies no normalisation, match TopSpin's.

savefid undoes each of these steps, so loadfid on a written experiment returns the same data.

Digital filtering and group delay

Bruker spectrometers filter and decimate the oversampled signal digitally. The filter's response builds up over the first GRPDLY points of the stored FID, so time zero falls GRPDLY points into the data. The filter's response is symmetric, so those build-up points carry part of the signal at negative time.

loadfid keeps the build-up points: the direct dimension's time axis starts at -GRPDLY dwell times, and the delay is recorded in its :grpdly metadata. The processing functions respect this:

• apodize starts the window at t = 0 and leaves the build-up points unwindowed;
• zerofill appends zeros after the end of the FID;
• linearpredict predicts from the FID from t = 0;
• ft removes the delay with a linear phase ramp after the Fourier transform, which moves the build-up points to negative time (the end of the zero-filled array).

This is how TopSpin and NMRPipe process the data, and it gives the same spectrum as TopSpin. Discarding the build-up points instead truncates the filter's response and produces a "smile": a baseline that falls away at both edges of the spectrum.

How the delay is found depends on the acquisition parameters:

Baseopt

In baseopt mode the spectrometer aligns acquisition with time zero (the pulse programme's acqt0) and corrects the initial points of the FID itself, so no first-order phase correction should be needed and the t = 0 point already has its half weight.

FIDs modified by cryoproc1d

A DSPFVS outside 10–23 is not a firmware version Bruker defines. TopSpin's cryoproc1d, used to process X-nucleus spectra from cryoprobes, sets DSPFVS = 0 after rewriting the FID with its start removed: the filter build-up and some points after time zero (the original is saved in another experiment number, recorded in audita.txt). loadfid compares TD in acqu, which keeps the original value, with TD in acqus to find how many points after time zero are missing, and starts the time axis at that time, so that ft restores the correct time origin and the lines can be phased together. A warning reports the number of missing points.

Their absence leaves a ripple around strong lines. TopSpin rebuilds them by backward linear prediction with thousands of coefficients; linearpredict(fid, T1Dim; direction=:backward, npoints=n), or autoprocess(fid; rebuild=true), rebuilds them with a low-order prediction, which removes the ripple but can distort the baseline when the FID contains fast-decaying signals such as probe background.

Phase conventions

phase(spec, dim, p0, p1) applies the phase θ = p0 + p1·n/N to point n (counting from zero at the left-hand, highest-ppm edge), in degrees. The sign and pivot match TopSpin's PHC0 and PHC1, so values can be copied between the two. Each call adds to the dimension's :phase metadata, which holds the total correction applied so far as (p0=…, p1=…) with p0 referred to the first point.

Sign convention

NMRTools spectra are the complex conjugate of TopSpin's in every imaginary unit: the real parts agree and the imaginary parts have the opposite sign. phase uses TopSpin's sign for the phase angle, which is why PHC0 and PHC1 carry over unchanged.

When imaginary data are read from or written to Bruker processed files, they are converted between the two conventions: loadnmr(...; allcomponents=true) conjugates the imaginary components it reads, and savenmr conjugates those it writes. Each program's files therefore follow its own convention. For 2D data, 2ir holds the coefficient of im1 (the direct dimension) and 2ri that of im2.

Intensities

loadfid scales raw data by 2^NC and ft applies no normalisation, so spectra processed in NMRTools have the same intensities as TopSpin's (to about 1 part in 10⁸ for the 1r of a 1D). savenmr writes 32-bit integer data and sets NC_proc so that TopSpin reads the same intensities.