Processing 1D NMR data#

Processing, validation, and qualification#

  • Processing means applying operations to the data. For example, apply exponential apodization and then an FFT to a time-domain NDDataset. When working step by step, you choose the operations and their order.

  • Validation is the set of checks made when you call Experiment.process(). The plugin checks whether it can run the requested workflow on that dataset, using properties such as its domain, dimensions and acquisition encoding. Individual core operations also have their own input checks.

  • Qualification is evidence from testing a specific acquisition and processing configuration. For example, the TopSpin EFP replay below was compared with a TopSpin reference for the documented Bruker DQD profile. That comparison applies to the settings covered by the reference tests. A successful call to Experiment.process() tells you the input passed the runtime checks.

Processing a FID#

Read the Bruker FID with digital-filter correction disabled to inspect the recorded time-domain signal. The plot shows its real component.

[1]:
import spectrochempy as scp
from spectrochempy_nmr import autophase

path = "nmrdata/bruker/tests/nmr/topspin_1d/1/fid"
raw_fid = scp.nmr.read(path, remove_digital_filter=False)
_ = raw_fid.plot(title="Bruker FID as recorded")
../_images/userguide_processing_1d_1_1.png

The digital filter delays the signal by 76 samples (about 304 µs for this acquisition). Plot the first 480 points to see the delayed onset and subsequent decay. With remove_digital_filter=True, the reader compensates for the delay and the signal appears near the start of the corrected time axis. The magnitude shows the onset independently of the phase of the complex FID.

[2]:
fid = scp.nmr.read(path, remove_digital_filter=True)

_ = scp.plot_multiple(
    [abs(raw_fid[:480]), abs(fid[:480])],
    method="pen",
    labels=["As recorded", "Delay corrected"],
    legend="upper right",
    xlabel="Time / µs",
    ylabel="FID magnitude / count",
)
../_images/userguide_processing_1d_3_0.png

Wrap the corrected FID in Experiment and select explicit processing parameters.

[3]:
experiment = scp.nmr.Experiment(fid)
result = experiment.process(apodization="em", lb=5.0, phase=None)

See Apodization in NMR for the em, gm and sp windows. Use result.history to inspect the steps recorded on the spectrum. Use phase="manual" with explicit angles. With core 1.2.1 and phase=None, the transform leaves the phase untouched. The development version of the plugin also supports phase="metadata" to apply angles attached to the FID.

To inspect intermediate results, apply the core operations to the NDDataset exposed by experiment.dataset. You can call them as dataset methods or as functions such as scp.em(experiment.dataset, lb=5.0). Calling plot() on the complex spectrum displays its real component by default. Set show_complex=True to add the imaginary component as a dashed line.

[4]:
windowed = experiment.dataset.em(lb=5.0)
padded = windowed.zf_size(size=max(32768, windowed.shape[-1]))
spectrum = padded.fft()
unphased_axes = spectrum.plot(show_complex=True, title="After Fourier transformation")
unphased_axes.autoscale(axis="y")
../_images/userguide_processing_1d_7_0.png

Apply phase correction to the complex spectrum, then plot both components over the same ppm range. The zero-order angle phc0 is constant across the spectrum, while the first-order angle phc1 varies linearly with frequency. The angles below are in degrees and apply to this acquisition. Compare the shape of the main resonance before and after correction.

[5]:
phased = spectrum.pk(phc0=52.44, phc1=-16.84)
phased_axes = phased.plot(show_complex=True, title="After phase correction")
phased_axes.autoscale(axis="y")
../_images/userguide_processing_1d_9_0.png

Both approaches preserve the original FID by default. A supported complex JEOL 1D FID read with scp.nmr.read_jeol() can use either approach. For example, on the epicatechin JDF reference, Experiment.process() produces a spectrum referenced in ppm, and the stepwise core workflow above produces a frequency axis in Hz.

For a 1D acquisition outside the scope of Experiment.process(), you can still apply individual operations when the data meet their input requirements. For example, use a suitable time-domain axis for apodization and FFT.

An exponential window trades some resolution for sensitivity. Zero filling sets the spacing of the frequency bins. See Analyzing NMR spectra for baseline correction, peak picking, integration and fitting of a phased spectrum.

A direct .fft() transforms the samples supplied to it. For a Bruker FID you can explicitly correct the filter at import with remove_digital_filter=True. With the development plugin, use experiment.process(digital_filter="time") to correct a supported FID as part of processing. For a normalized FID that has not been scaled by scan count and receiver gain, prepare_fid() can also apply DC-offset and digital-filter corrections before the FFT.

A pseudo-2D series contains ordinary 1D spectra that can be processed one spectrum at a time. For acquisitions with two Fourier dimensions, see Processing 2D NMR data.

See Core operations used in NMR for the core functions used in NMR.

Estimate phase and correct the baseline#

The development version of the plugin provides autophase() for a complex spectrum after the FFT. It requires a version of nmrglue-ng with phase estimation support.

[6]:
diagnostic = autophase(spectrum, order=0, apply=False)
result = autophase(spectrum, order=0, baseline_options={"model": "asls", "lamb": 1e9})

Inspect result.estimate for the correction angles and fit diagnostics. result.phased is the complex spectrum. If you request baseline correction, result.baseline and result.corrected are real datasets. Use result.phased for further phase adjustments. AsLS is suited to positive peaks. For signed peaks, fit a polynomial on known signal-free regions.

ACME assumes positive absorption peaks. Pass a boolean exclude array to omit artifacts from phase scoring while retaining their data in the result. For isolated peaks of known signs, choose method="dispa" and provide references=[(start, stop, sign), ...] with half-open index windows. Use well-separated references for order one. The optional baseline_mask selects signal-free indices to fit a provisional complex linear background for phase scoring. The final real baseline is a separate fit. A successful solver status still calls for inspecting the result.

Replay TopSpin EFP on a Bruker DQD FID#

Use the raw FID and the processing status file (procs) from a completed TopSpin efp run in the same experiment:

[ ]:
fid = scp.nmr.read_topspin(
    "experiment/1/fid", remove_digital_filter=False, remove_dc_offset=False
)
spectrum = scp.nmr.Experiment(fid).process(replay="topspin")

Replay uses the recorded exponential window, FFT size and phase. It applies digital-filter compensation on the final FFT grid, including fractional group delay, and uses the processing ppm axis. Intensities retain the FID reader’s scan and receiver gain normalization, so comparison to TopSpin may require an overall amplitude factor.

The supported profile is a full-spectrum DQD 1D efp run with PKNL enabled, FCOR=1, and baseline correction, linear prediction and axis reversal disabled.

Fractional PKNL correction for a Bruker QSIM FID#

Read the original FID with both import corrections disabled. The plugin applies the digital-filter correction after the FFT. If you already corrected the filter during import, reread the original FID before using PKNL.

[ ]:
fid = scp.nmr.read_topspin(
    "experiment/80/fid", remove_digital_filter=False, remove_dc_offset=False
)
spectrum = scp.nmr.Experiment(fid).process(
    apodization="em", lb=50.0, size=8192, digital_filter="pknl"
)

The result uses the processing ppm reference. To apply measured phase angles at the first frequency bin, pass phase="manual" with phc0 and phc1 in degrees. The cross-polarization example uses one pair of angles from a representative FID for the entire series.