S/N calculation in GCMSsolution for Shimadzu

[sd12]

Dear All,

I seem to be having problems with the signal-to-noise calculations in the GCMSsolutions software for Shimadzu (we have a QP2010)...

For a start, the software will only allow me to do the calculations for single m/z values and I need to look at S/N in TIC too. As the help files show how the software calculates these values, I decided to recreate them in Excel so I could calculate my ratios for single m/z and TIC data alike. Unfortunately, the results I'm getting are significantly different from what the software gets and no amount of fudging or fiddling seems to do the trick! I've had my calculations double- and triple-checked by other folk (including a mathematician) who all say they are correct, follow the methods specified by Shimadzu (and use the correct dataset!). In some cases, even just an inspection of the exported raw data (i.e. without calculations) is sufficient to see it's impossible to get the values obtained by the software using their methods on said data (e.g. how can the RMS of a dataset be 9 less than the lowest value?!).

My questions, therefore, are:

1) Has anyone else had this problem, and if so how did you solve it/get round it?

2) Which is the more likely scenario:
a) there's an error or glitch in the software;
b) the software uses a completely different method from that which it claims to use;
c) the software uses different data in its calculations from that which it exports?
If scenario c) above is true, then I hate to think what I'm going to find when I start plotting calibration curves and the like!!

3) Can anyone suggest a reference which contains methods for calculating signal and noise values from raw chromatography data, please? (I'm understandably sceptical about the Shimadzu literature now!)

Thank you for taking the time to read this, and apologies for its length! I will be most grateful for any help you give me on this subject.

Many thanks in advance,
sd12

[macgyver]

Hi,

The SN calculator can only calculate using single m/z values.

The desciption given in the help is a bit vague, but I think I've figured out how it calculates the RMS value. I had been assuming (probably as you had) that the RMS was calculated from the noise values calculated using the 1.6x to 3x std. dev. calculations from the mean. It finally dawned on me that the RMS calculation is independent from these calculations.

The calculation is pretty simple:
(Max intensity of signal - Average intensity of noise region)/Standard deviation of noise region

The help does allude to this a bit:
"Particularly for RMS, the value of "Sigma" is defined as the noise level."
and
"For RMS with its noise calculated only from standard deviation, pickup points, or pickup ration for that matter is irrelevant."

I'm no statistician, so don't know if this is an appropriate way of calculating S/N. Instrument manufacturers tend to define S/N different ways - usually to try and beat each others specs, so the RMS method may not be appropriate for your samples.

[sd12]

Thank you for your reply, macgyver, but I was already calculating the RMS value independently of the sigma type calculations. Neither the sigma-based nor the RMS calculations return values like the ones given by the software.

For example, as you say, they define the signal level as:

max intensity of the peak detection time range - average intensity of the noise region

For one of my datasets,
max intensity of the peak detection time range = 40746;
average intensity of the noise region = 254;
40746 - 254 = 40492 yet the software quotes a signal level value of 40110!?

As for the noise level, the software quotes a value of 181 (for the RMS calculation), yet the root-mean-squared of the exported noise data is 255.5; the mean being 254, as I've just said. (Incidentally, the standard deviation of my noise regions intensities is only 28.96, so it's definitely not that either!). Alluding to the sigma-type calculations, (max noise intensity - min noise intensity)/2 = 58.5, while (max noise intensity + min noise intensity)/2 = 248.5. None are anywhere close to the elusive 181!

I've even tried using the other method for calculating noise given in the help section entitled "QA/QC Calculation Methods" where you take the distance between two parallel lines above and below the chromatogram points, but that gives me a value of 92.25. I've noticed that 2 x 92.25 = 184.5, which is closer to the 181 quoted than any of my other attempts, though there's nothing in the help files which says I need to double this value so I fear I may be clutching at straws here...

If anyone can shed any light on this problem, or even suggest a good reference for the definitions/calculations surrounding chromatography S/N ratios, I'd be really grateful as I'm totally out of ideas now!

Cheers,
sd12

[macgyver]

Hmmm, I can't explain why you're getting different results. I managed to get comparable results to the calculator using Excel. Make sure you are using the same noise region specified in the calculator parameters menu item.

If you still can't get the same results it might be worth contacting your local Shimadzu office and get the latest software update. There was some changes made to S/N calculations a few versions back. The latest version is 2.53 SU3 (which is the version I'm using).

[sd12]

Thanks macgyver, I wonder if it is maybe the version I have then (2.50 SU1) (I'm definitely using the correct noise region!). It's good to know that the latest version definitely works, though, and from what you said earlier the help files seem to give the same equations, so maybe I can just use those afterall.

Thanks again!
sd12