What's the best way to measure baseline noise?

[gdisaac]

We are establishing a process for determining the low level sensiutivity of LC UV detectors as part of the instrument qualification process for a pharmaceutical QC application. Now I realize that this is really an anal;yte-specific and wavelength-specific attribute, but it is being requested by our lab QA compliance group. Anyway, we've decided to do a simple S/N test, but there is some disagreement about whether to use the peak to peak noise (my choice) or RMS noise. I prefer peak to peak to avoid the smoothing effect of the RMS noise. My question is whether there is an industry accepted way to measure baseline noise - which approach is most common?

Thanks!

[tom jupille]

ASTM has standard practice(s) for testing detectors which include a detailed procedure for noise measurement (essentially peak-to-peak). E-1657 (for variable-wavelength detectors) is a good example.

I believe the USP procedure is similar, but I'm not 100% sure.

[HW Mueller]

When I first saw this question I thought: Not again! But then I remembered that, especially recently, I saw some researchers leave out quite a few of the more extreme peaks in their "peak to peak" determination. I myself usually pick a particularily smooth section of the baseline. So there is quite a bit of fudging done, apparently. I don´t see how statistical type of methods help too much in "standardizing" here. What should help is a close description of what was done to determine the noise.... and a fair amount of scepticism toward published figures.

[Bruce Hamilton]

Hi,

Noise measurement can depend a lot on the method and instrument, as well as what the client/regulator may want.

Is it a gradient method, if so UV noise may higher at periods of steep gradient changes, so try to ensure analytes don't appear there, or about 0.5 min immediately after ( for Agilent 1100 ).

Work out where all the peaks of interest elute, and look at noise in that region, not just the main peak. Choose your poison, peak to peak is visually easier, RMS may specified in some methods, I've seen both,
but it depends on the regulator or client, as well as what your data system delivers.

Once you choose, be consistent for all other methods and instruments, unless specificly required - it's good to have a "default" practice.

Look at noise in the samples and the blanks, at beginning and end of a sequence, they should be similar. I've seen noise measured in "blanks" and external standards that were much smoother than the samples, yet same LOQ/LOD was reported for samples.

Where possible, LOQ/LOD usually should to be confirmed by noise measurement after spiking a sample, rather than just dilution of a sample main peak to 0.x% with mobile phase, as some samples baseline noise may decrease with dilution, especially if sample solvent is not same as initial solvent. Depends on your method validation protocols, they should check that noise is similar for all sample concentrations and dilutions, as well as blanks, but I've seen some that don't.

If it's a UV detector method with high aqueous buffers or UV-absording solvents, ensure that baseline noise is checked regularly, as batch to batch variations of mobile phase components can also affect baseline noise.

I also suspect that you'll find there are regulatory guidances for your specific situation. Noise determination ( and the consequent LOQ/LOD )in HPLC methods tend to exercise auditors.

Bruce Hamilton