Range for S/N ratio

[anrao_2]

I understand that for LOQ the S/N ratio should be 10 and for LOD the S/N ratio should be 3 or 2. However, I also understand that it is not always possible to reach the exact figure of 10 or 3 (or 2). Is there any range prescribed for the S/N ratios for LOQ and LOD. Say for instance can we say a S/N of "10+/-3" (ie., 7 to 13) is acceptable for LOQ and similarly a S/N of 3 +/-2 (1 to 5) is acceptable for LOD. Is there any standard reference where this issue is addressed. Kindly advise.

[orcicdejan]

How do you mean, "to reach exact figure of 10"? By preparing different low-concentration standards and determining their S/N, until you hit "10"? In that case, you should better plot C vs S/N and determine from the curve the C which gives S/N = 10.

Anyway, that "10" figure is not "10.0000". You don't have to hit exact value. However, I believe you shouldn't go [much] below it (same for using S/N = 1 as LoD).

BTW, there a couple of ways for determining the LoQ/LoD, either instrument limits or method limits. If your sample preparation has a bad precision, actual limits will be much higher than those determined by S/N.

[anrao_2]

Thanks for your reply. I understand what you mean by saying that S/N need not be exactly 10.0000. But how variation is allowed? Is it mentioned any place in the literature?

[orcicdejan]

I guess it depends on number of samples analyzed, degrees of freedom and other statistical parameters.

The best I could find is:

2, 3, 5 or t (Student's t value for n-1 degrees of freedom, where n is the number of blanks analyzed, and 99 % confidence) as typical criterion for LoQ, and

5, 6 or 10 for LoD.

[lmh]

My take on this is: The LOQ is exactly that: a limit. You should not quantify below the lower limit of quantification, but there's nothing to stop you quantifying above the lower limit! So if your personal lower limit actually gives a S/N of 15, it's perfectly safe to quote it as a lower limit. You are just being a little excessively safe.

However, there are much better ways to determine a LOQ than S/N ratio. Methods based on the standard deviation of a low range calibration curve (determined in the same way as your samples) are probably more realistic.

[ksharp]

When determining LOD and LOQ, run a series of increasingly dilute samples and measure S/N. You can plot the results and calculate the concentration at which S/N = 3 or 10.

Or you can play it safe and set the LOD or LOQ at the last concentration where S/N > 3 or 10, and use that as your practical LOD / LOQ.

[tom jupille]

See my rant in the signal/noise ratio thread

viewtopic.php?t=10960

[bily200677]

For the ratio of S/N，i didn't find out the guidline with specifid range ! the LOQ and LOD is a inaccurate value , we should meet the requiement which S/N should be lager than 3 or 10. During our validation work , we usually control the range of S/N between 10~13 for LOQ and 3~5 for LOD!

[grzesiek]

there are no ranges for these
condition change so does your s/n
precision and accuracy is what counts
these s/n are only estimates

[shahis77]

There is no range the values are fixed

[lmh]

I'm getting a little frightened. See the other thread. But to summarise:

(1) Limits of quantifaction and detection are genuine, accurately assessable values, but they should, ideally, depend on what you are doing. If you need to get a measurement +/- 5%, your limit of quantification will be more exacting than if you only need +/- 20%.

These limits are points where the random errors of measurement (which are themselves measurable, and known) create a chance of detecting false positives, or missing false negatives (LOD), or increase the error in the measurement to a point that is deemed unacceptable in the context of the measurement (LOQ).

(2) S/N ratio is one of many means to estimate a LOD or LOQ. The S/N ratio you need isn't set in stone, as it depends on the context. But there is a rigorous statistical background to this, it's not just magic numbers plucked from the sky.

(3) It's possible to work happily without the foggiest understanding of all this, but there again, it's possible to skate on thin ice (for a while).

[grzesiek]

all this s/n discussion is somewhat strange - the reality is you want your results to be accurate and precise
what you measure is response of the instrument usually it is peak area and all s/n discussion and culculations is based on peak and baseline height
adding to confusion hardware and all the conditions that can affect the baseline I would suggest to not rely too much on s/n ratio - it can be helpful i agree but you need to validate your LOQ sooner or later and this mean accuracy and precision of the results (which you derive from peak area most of the time)
I found that peak area is more robust to changing conditions than s/n ratio so i advise using the s/n ratio carefully and not to become to dependend on it