Requesting Advice on Calibration/Fit method.
Posted: Fri Dec 31, 2010 11:41 pm
Let me Start from the first so that i can explain everything clearly. I work on combustion and i use a Shimadzu QP2010 with a FID to quantify stable hydrocarbon species in an Ethylene Diffusion flame. I use a quartz probe to extract samples and i pass them through the GC to quantify species. In order to quantify we initially purchased a custom gas mixture which had 100 ppm mixture of all the 10 species that we are interested in.
We pass the standard gas into the GC and use a simple single point fit to obtain quantitative information of the 10 species in the unknown samples extracted from the flame. But we ran in to a problem, Since the measured/expected concentrations of all the species were few orders of magnitude off the 100 ppm standard, we felt that we were extrapolating the 100ppm signal too much to get concentration values that were a few orders of magnitude different (eg. measured ethylene was 100000 ppm). A second mixture was then made which had a mixture of all the 10 species as before but with different concentration ranges that fell within the expected range.
The plan was then to use both the standard mixtures in a multi (2) point calibration technique, that should hopefully give better results.
The instrument has 3 different curve fit options to create the calibration curves, and all of them gave different answers, so its was difficult to say which method was correct. We then conducted a simple experiment using two different single point calibration curves, one was based on the 100 ppm mixture and the second was based on the newer mixture. When we pass the standard gases as unknowns and quantify them using the two different curved we get two different concentrations as one expects.
Following this three more calibration curves were made with the three curve fit options using the two standard mixtures (Linear, Mean RF and Exponential). When we passed the standard gases as unknown the first two calibration curves (Linear, Mean RF) were incorrectly quantifying the concentration but the exponential fit worked perfectly in identifying both the 100 ppm mixture and the different concentrations in the newer mixture.
But the problem comes when i pass an unknown sample from the experiment, get it quantified using the exponential fit, it seems to incorrectly evaluate the concentrations and they are different from the results i get when i use Linear or Mean RF fitted calibration curve on the same sample. I am really confused as to what i am doing wrong with respect to the method of my creation of calibration curves. I right now am not confident with any of my calibration curves. Am i missing some thing?
PS: The old mixture had 100 ppm of Ethylene, Acetylne, Ethane, Benzene, Toluene, 1 butene, 1,3 Butadiene, hexane, propane and cyclopentane. (Measured concentraion varied from 100000 to 2 ppm).
The newer mixture has 100000ppm ethylene, 10000 acetylene, 1000 benzene, 2000 1,3 butdiene, 10 1 butene, 5 cyclopentane, 30 ethane, 5 hexane, 5 propane, and 10 toluene. These values were chosen based on the peak concentration measured using the 100 ppm mixture.
Sorry for the long post, thanks in advance.
We pass the standard gas into the GC and use a simple single point fit to obtain quantitative information of the 10 species in the unknown samples extracted from the flame. But we ran in to a problem, Since the measured/expected concentrations of all the species were few orders of magnitude off the 100 ppm standard, we felt that we were extrapolating the 100ppm signal too much to get concentration values that were a few orders of magnitude different (eg. measured ethylene was 100000 ppm). A second mixture was then made which had a mixture of all the 10 species as before but with different concentration ranges that fell within the expected range.
The plan was then to use both the standard mixtures in a multi (2) point calibration technique, that should hopefully give better results.
The instrument has 3 different curve fit options to create the calibration curves, and all of them gave different answers, so its was difficult to say which method was correct. We then conducted a simple experiment using two different single point calibration curves, one was based on the 100 ppm mixture and the second was based on the newer mixture. When we pass the standard gases as unknowns and quantify them using the two different curved we get two different concentrations as one expects.
Following this three more calibration curves were made with the three curve fit options using the two standard mixtures (Linear, Mean RF and Exponential). When we passed the standard gases as unknown the first two calibration curves (Linear, Mean RF) were incorrectly quantifying the concentration but the exponential fit worked perfectly in identifying both the 100 ppm mixture and the different concentrations in the newer mixture.
But the problem comes when i pass an unknown sample from the experiment, get it quantified using the exponential fit, it seems to incorrectly evaluate the concentrations and they are different from the results i get when i use Linear or Mean RF fitted calibration curve on the same sample. I am really confused as to what i am doing wrong with respect to the method of my creation of calibration curves. I right now am not confident with any of my calibration curves. Am i missing some thing?
PS: The old mixture had 100 ppm of Ethylene, Acetylne, Ethane, Benzene, Toluene, 1 butene, 1,3 Butadiene, hexane, propane and cyclopentane. (Measured concentraion varied from 100000 to 2 ppm).
The newer mixture has 100000ppm ethylene, 10000 acetylene, 1000 benzene, 2000 1,3 butdiene, 10 1 butene, 5 cyclopentane, 30 ethane, 5 hexane, 5 propane, and 10 toluene. These values were chosen based on the peak concentration measured using the 100 ppm mixture.
Sorry for the long post, thanks in advance.