Documentation
Preparing NMR data
Referencing, phasing, baseline, solvent and field information that change whether a Spectra shortlist is scientifically usable.
Spectra can only reason about the spectrum you submit. Garbage in still produces a shortlist. The shortlist will look confident and still be wrong.
Referencing
Incorrect chemical-shift referencing moves every comparison. Check TMS, residual solvent, or the laboratory’s validated reference before upload. If 1H and 13C were referenced independently, say so in the case notes.
Phasing
Mis-phased multiplets distort integrals and apparent shifts. Phase in your usual processing tool (TopSpin, Mnova, Delta, or another validated processor) before asking Spectra to rank structures.
Baseline
Rolling baselines create false peaks and hide weak impurities. Correct obvious baseline errors first. Spectra is not advertised as a replacement for processing.
Solvent
The live example uses DMSO-d6. Always state the solvent. Residual solvent and water signals are otherwise easy to treat as analyte peaks.
Field strength
State the spectrometer frequency. The public example uses 600 MHz. Predicted versus measured residuals are not comparable across an undeclared field.
Nucleus
Submit the nucleus you actually measured. Do not relabel a DEPT or edited experiment as a fully assigned 13C without saying so.
Sample context
Note whether the sample is isolated, a mother-liquor, a spiked standard, or an extract. Mixture status changes how a one-compound ranking should be read.
Common quality problems
- Clipped windows that omit aliphatic or aromatic regions
- Severe overlap presented as a pure compound
- Unstated impurities from NMR tubes or grease
- Digital resolution too low for the claimed discrimination
- 13C not acquired when the 1H alone cannot separate analogues
Next: running an analysis.