Data Acquisition, Integration and Quantitation
Data Acquisition, Integration and Quantitation
This domain tests the chain from detector signal to reported number: sampling rate, integration, calibration, and the integrity of the record.
Acquisition settings
- Set data rate from the narrowest peak, not the run length. Aim for at least 15–20 points across the peak; more points add noise but describe peak shape more truthfully.
- Match detector time constant to data rate. Too slow a response filters the peak and costs resolution and plate count.
- Wavelength trades sensitivity for selectivity: absorbance maximum for trace work; an interference-free wavelength for complex matrices.
Integration
- Key parameters are peak width, threshold or slope sensitivity, and baseline treatment. Set them from observed peak widths, then hold them fixed for the batch.
- Manual integration is acceptable only under an SOP with justification and audit trail. Shoulder tangents must be applied consistently across the batch.
Calibration and quantitation
- Choose the regression from the data: linear through zero, linear with intercept, or quadratic. Weighting (1/x, 1/x²) is standard when the range spans more than about two orders of magnitude.
- LOD and LOQ commonly use S/N of 3 and 10, or the standard deviation of the response divided by the slope. Report accuracy and precision at the LOQ.
- Inspect residual plots, not only R²; R² = 0.999 can still conceal curvature.
Worked example
The narrowest peak has a baseline width of 0.08 min. Twenty points across it need 0.08/20 = 0.004 min per point = 0.24 s, so roughly 4 Hz. Set 10 Hz to leave margin for a column change that sharpens peaks, provided file size and disk capacity are acceptable.
Traps
Changing integration parameters mid-batch; extrapolating a calibration beyond the validated range; unweighted regression over a 1000-fold range; using peak height where a shoulder exists.