Sampling rate

[pstaunton]

How do you figure what's the best sampling rate for a method.

Can someone explain to me how too low or high a sampling rate can have on your method?

[rb6banjo]

You want to sample fast enough to accurately capture the instrument response to a particular analyte but not so fast that you "over capture" it. When you do too good of a job (too high of a sampling rate) you're going to also capture a lot of noise. For instance on my flame detectors, I usually go at about 20 Hz. In a mass spectrometer, it's a little different thought process as it's more about the number of scans per second.

[Jasn]

Hi,

it all depends on the width of your peaks and if you want to do quantitative or qualitative interpretation.
For robust quantitative interpretation of a peak you typically need 12-20 measurement points for each peak.
For qualitative work 7 might do.

So to know the 'ideal' sampling rate you need to know the approximate width of the smallest peak in your chromatogram (mostly the first peak that elutes) in seconds. Divide the points you want to have over that peak by the peak width.

Example:
Peak width : 3 seconds
You want : 12 points
-> Sampling rate = 12points/3sec = 4Hz

As rb6banjo said. Oversampling makes that there is more noise, so you lose signal/noise. Undersampling does not give a peak area that is reproducible.

[pstaunton]

Thanks for replies. I'm currently running on a method that has a sampling rate of 50.I'm assuming this is in hertz.

So would this be considered over sampling?

[Jasn]

If I were you i'd try out a few settings(Maybe 10, 20 and 50?). Some detectors/software average datapoints when you lower the sampling rate. A sampling rate that is too low could broaden your peaks with loss of resolution. Depending on the application, there is always a 'sweet spot'. 50 is probably overkill though. A typical GC peak is about 3seconds wide, so with a sampling rate of 50Hz you'd have about 150 datapoints across your peak. Electrical noise becomes aparent when the rate is too high. You can increase s/n (and LOD) with a factor approximately equal to the square of the factor reduction in sampling rate. So decreasing sampling rate from 50 to 10 could theoretically increase LOD by a factor square(50/10) = +-2.25.

[rb6banjo]

I agree with Jasn. Try a few different settings and see how it changes. You definitely don't want to undersample. If your peaks are not smooth on the leading and trailing sides, you are probably over sampling.