Troubleshooting Callibrationcurve far away from Zeropoint

[SpiritAnalytics]

Hello Everybody

I work in a Distillery analysing our products and incoming raw Materials with an GC using FID detection.

Recently i recallibrated the System and afterwards an anomaly appeared which i cannot make sense of.
I have checked and double checked my Steps, calculations and Settings but found nothing wrong.
So i turn to You in hope of getting an idea whats going wrong here.

Usually i force the calibration Curve thru the Zeropoint. It makes Sense to me that Zero Concentration should equal Zero Detector Signal above Baseline. This has until now worked nicely for me as the callibration Curve did match up with the zeropoint with minimal deviation.

In my latest Callibration one of 34 Substances in my Standard came out different. The 36 Datapoints (6x6) show a linear Line but the Slope in no way fits the Zeropoint.

I made some Pictures so you can see my Problem:





With this i have two Options and none of them is really satisfying:
1. Ignore the Zeropoint and work with The Linear callibration Curve providet by the Data. This results in negativ Values for lower Concentrations. I can hardly send out Analysis-Reports containing negative Concentrations as this is plain false.
2. Force Zero and have R-Squarred at 0,79 when it should be at least 0,99. This callibration Curve does not really have any relation to real Concentrations and is thereby unusable.

Basically this Callibration is not usable and i dont know what caused it.
I have done two Callibrations recently and both came up with this problem so it aint a one time Thing either.

In my earlier Callibrations this Problem did not occur, therefor the error whereever it lies must have come with this new Callibration and it only happens for one out of 34 Substances.
Furthermore i have run Blankruns before and after the Calibrations which show nothing but a Baseline.

Im at my wits end and i hope one of You see´s where the Problem lies or has encountered this himself in the past and can help me get it fixed.

P.S.: i read the Sticky "Embedding Chromatograms and Reports" but when i resize my pics to 400x400 you cant read or see anything as it all get a pixelmess. Hope you can life with the big pics.

Thanks for your efforts,

SpiritAnalytics

[tkubowicz]

Hello

With this i have two Options and none of them is really satisfying:
1. Ignore the Zeropoint and work with The Linear callibration Curve providet by the Data. This results in negativ Values for lower Concentrations. I can hardly send out Analysis-Reports containing negative Concentrations as this is plain false.

SpiritAnalytics

Your analyte response (and given concentration) should be above (higher than) your lowest calibration standard. So just extend your calibration range (add lower concentration) and you can avoid negative concentration.
I think that your curve has really scattered value on each level...Are you sure there is nothing wrong with injection precision? (or LC sampler)

Regards

Tomasz Kubowicz

[James_Ball]

This happens if the sensitivity to the analyte is lower than normal. It is taking a larger amount of analyte to be on the column before it is detectable, but beyond that point the increase in response versus concentration is linear. You just have a positive offset to the calibration. You will have to determine what concentration is equivalent to the zero detector response and that will be the lowest detectable amount possible, and anything below that amount will be unknown.

If there are active sites in the sample pathway that adsorb up to a given amount of analyte this will happen. Is works as if you subtract a given amount from each standard before injecting it. If you had standards of 2,5,10 and you subtracted 1 from each before injection you would now have concentrations of 1,4,9 but the system is still thinking they represent 2,5,10 so the line will have a positive offset when plotted.

[CharapitsaS]

Hi,

presented above problems, probably, have the following origin.
Your origin/starting “water + ethanol” solvent contains an appreciable amount of methyl acetate.
It is this quantity that must be correctly taken into account when calculating the prepared standard solutions.

For example, ignoring an appreciable amount of methanol in the original solvent when preparing the standard solutions leads to the following graphs.



In our practice, when preparing standard solutions in accordance with the Commission Regulation (EC) No 2870/2000, we must determine impurities in the initially water-ethanol solution and then take into account when calculating the concentrations of volatile compouns in the prepared standard solutions for calibration.



Allowance for methanol impurities in the original solvent removes apparent contradictions in the calibration curve.



There is the following effective way “Ethanol as Internal Standard” for direct determination of volatile compounds in spirit products
https://www.youtube.com/channel/UCXgL2c ... lW1oxOGqtQ

Best regards,
Siarhei

[Peter Apps]

"It makes Sense to me that Zero Concentration should equal Zero Detector Signal above Baseline."

Is fine as far as it goes. Problems arise when a non-zero concentration also gives a zero signal.

Peter

[SpiritAnalytics]

@ CharapitsaS you were right

@ tkubowicz the scattered values come from the temperature program were using which gives us a rough baseline in that area but is neccesary for sepperation.

Sorry for the late answer but work had me busy.

The Culprit has been found!

Our Solvent Ethanol was contaminated with the Substance in Question. Thereby the Concentrations in each Calibration-Solution was not what i had calculated but what was calculated plus a Value that was equal in all Solutions. This explains why the Calibration Curve was off.

To clarify the Blankruns i made before my first post where runs without any injection. Just to check if my System is Contaminated somehow. After the post was made, i made a run with our Solvent solution water-Ethanol mixture which clearly showed a Peak for the Substance in Question.

Hopefully this Thread can help other who may encounter the same Problem in the Future.



Good bye

SpiritAnalytics

[Peter Apps]

Thanks for the feedback. The problem you describe is one of the reasons why forcing through zero is a bad idea - you were forcing a non-zero quantity to generate a zero result.

Peter