LOD & LOQ Determinations for Gas Analysis on FTIR

[JWF239]

Hello,

So I am doing gas analysis on my FTIR with a 10 M path length cell and am at the point of doing LODs. I am trying to determine the best method for doing it. I wanted to toss some ideas around before I came to a conclusion on it. I would like to use signal to noise ratio as opposed to linear regression but I was unsure of what signal to actually use. Is it the signal of my lowest standard? Does it actually matter? Since the higher the concentration, the lower the standard deviation should be.

[HPLC chemist]

In the past I have determined the LOQ and LOD statistically. But it is usually confirmed in the pharmaceutical arena as 3 times the standard deviation (SD).

[JWF239]

In the past I have determined the LOQ and LOD statistically. But it is usually confirmed in the pharmaceutical arena as 3 times the standard deviation (SD).

Thanks for the reply, I am wondering the best way to go about this though. For instance, 3 times the standard deviation of what? The blank? a gas with a concentration at my MDL? My lowest calibration point? I have certified gases at certain levels, I don't have a way to "spike" the gas samples other than by varying the pressure to simulate other concentrations but this has its limits.

For example:

For carbon monoxide my lowest standard is 5.217 ppm
Using 5 samples and 4 degrees of freedom with 99.5% confidence I get
Mean = 5.2486
Std Dev = 0.053
T value = 5.84
MDL ( std dev*T Value) = 0.3095
LOQ (10* Std Dev) = 0.53
S/N (mean/Std Dev) = 99

So using my checks
1: (10*MDL>Conc) 3.095 is not greater than 5.217
2: (MDL<Conc) 0.3095 is less than 5.217 so this works
3: (MDL<requirement) My detection limit requirement is 0.5 ppm so this works
4: S/N is 99

So it is clear this is too high of a concentration to be used to determine my LODs from check 1, and the S/N seems to high, but what S/N signal would be acceptable?

Then when I use the blank it is
mean = -0.10056
Std Dev = 0.0396
DoF = 8
T Val = 3.5
So MDL = 0.1386
LOQ = 0.396
but the S/N is -2.5

Also the checks don't really work out using a blank because MDL cannot be less than the concentration when the concentration is 0.


Using linear regression my formula is Conc = 258.529 * (peak area) - 0.108

and LOD = 3* (Std Dev of Low Range Conc)/Slope so LOD = 0.00046

But this seems super low, also I don't know if using the concentration of a blank is acceptable as a "low range concentration".

The way I like best and gives me the values that make the most sense is using the blank data:

Signal = -0.10056
Noise = 0.0396
slope = 258.529

and

LOD = S/N / Slope of line * 3 = 0.0295 ppm

and

LOQ = S/N / Slope * 10 = 0.0982 ppm

Using a T score at 99.7% confidence and a Z score of 3 gives me

LOD = 0.049373 PPM

and

LOQ = 0.1546 PPM

This way allows me to use the blank scans for my LODs because I can't really get anymore concentrations of gases other than what I have used for my curves. Is this a valid way of doing LODs?