How to determine the Analysis Accuracy?

Chromatography Forum: GC Archives: How to determine the Analysis Accuracy?
Top of pagePrevious messageNext messageBottom of pageLink to this message  By Athar Wasim on Friday, May 30, 2003 - 04:48 am:

How can we determine the analysis accuracy of any given component (in PPM or percentage units) in process sample when GC accuracy given by GC manufacturer is 2 % and Calibration Standard (the GC is calibrated with) accuracy given by STD manufacturer is for example 1 %, It is square root of sum of both accuracies?
My second question is that what affect different in MESH sizes makes in column effeciency etc?

Athar Wasim
Analyzer Tech

Top of pagePrevious messageNext messageBottom of pageLink to this message  By Ralph on Tuesday, June 3, 2003 - 06:16 am:


I'll try the second question first

The smaller the mesh NUMBER then the larger are the mesh holes, so the larger the particle size. The larger the particle size then the smaller is the surface area. So, assuming that column dimensions remain the same then larger particles will give less opportunity than smaller particles for the sample molecules to interact with the surface i.e. reduced efficiency. So smaller particles are more efficient but this will lead to increased flow resistance and if the flow drops too much then the column can become less efficient. There is an optimum value for the two. In practice you select the particle size and column dimensions to suit the separation required in the time required.

As for the first question - my statistics are poor but I can make the following observations for what they are worth.

Your GC cannot be 2% accurate but it could have a repeatability on the result of +/- 1% within a working day. This will depend largely on the repeatability of your injections - are they manual or by autosampler?

If the supplied standard is say 97%+/- 0.5%, has it been analysed on a better GC system?

If the supplied standard is say 1000ppm then it could be anywhere between 950 and 1050ppm. Add to that the +/-1% variation in your GC area results means that your standard area could be equivalent to a range between 950+/-95 and 1050+/-105 giving a range of 855 - 1155ppm. But your sample will also have a +/-1% variation in the GC areas which will affect the calculation and should now be added in. If your standard is used as an internal standard then both experience the same variation so you only need to allow for the variation once.

So from all that (!) using an external standard the variation is plus/minus (0.5+1+1)= +/-2.5% and for an internal std the variation is (0.5+1)= +/-1.5%

BUT the measurement has variations which are the sum of ALL THE ERRORS in the analysis so also bear in mind that if you prepare a dilution of the standard how repeatable is your preparation? The same applies to your sample. What if someone else prepares the samples and standards?

If I've got it wrong I apologise and hope that someone who has more experience in this can reply.



Top of pagePrevious messageNext messageBottom of pageLink to this message  By Anonymous on Tuesday, June 3, 2003 - 01:56 pm:

Ralph et al

Good answer, Ralph. Mr. Wasim is an analyzer technician and certain details are different in that realm, however.

When this GC analyzer is said to be 2% accurate, that means as it samples and measures a single sample (or standard) over a 24 hour period (the measurement it was designed to measure, carbon dioxide: 1-5% in propane, for example) the measurement will vary within ±1% of the highest level within the range specified. In this example, 2% CO2 in a propane balance will be measured with a range of ±1% of 5%, or ± 0.05%. In a single day this analyzer will report the CO2 concentration within a window of 1.95% to 2.05% based on an external standard calibration.

The calibration standard Mr. Wasim uses has an accuracy of 1%. If he were describing the above example, then he would calibrate in the middle of the measurement range and have his standard contain 2.5% CO2. The standard is guarenteed to be within 1% of the specified concentration, or 99% to 101% of 2.5%, ie. 2.475% to 2.525%.

At any given time of day, the analyzer will give a reading with some error of measurement.

It will measure a CO2 concentration in a sample which may lie within a range of 1 to 5% based on an external standard whose concentration may be in error by ± 1% (2.475-2.525%) and whose single calibration may be in error by an additional ± 0.05%.

The best practise is to initially perform several calibration runs (5-8) get an average value to minimize calibration error, and do occasional calibrations daily or hourly as desired.

The errors can work against each other and cancel each other out partially or NOT.

Often exact determinations are not required and a trend analysis gives the most benefit to the operator.

I don't even want to discuss the non-linearity of analytes measured by TCD detectors. Enough !!!!

So Mr. Wasim, WHY do you need to know this? LOL


best wishes,

from Rod, a former analyzer jockey

Top of pagePrevious messageNext messageBottom of pageLink to this message  By Ralph on Wednesday, June 4, 2003 - 04:34 am:

Many thanks for that Rod


Top of pagePrevious messageNext messageBottom of pageLink to this message  By Athar Wasim on Thursday, June 5, 2003 - 04:14 am:

Mr. Rod and Ralph
Thank you Sirs, I am really grateful for your descriptive answeres.
Background of my question is as follows,

I am working in Analyzer Section of a Petrochemical plant.
We are weekly calibrating our Product critical on-line analyzers with calibration cylinders provided / certified by calibration cylinder manufacture. Each cal cylinder comes with certificate. The certificate provides components value and its accuracy printed against each given value. For each Calibration (external standard with six minimum runs to see repeatability) we fill and maintain a form called ‘Calibration Record Certificate’.
This form is part of our company’s approved SOP (standard operating procedure) and is auditable.
In this columned form we record,
Calibration cylinder serial no, cal cylinder expiry date, Component name, Cal cylinder value provided and certified by cal cylinders manufacturer, Analyzer reading on Cal cylinder, Error (accuracy) and remarks
Calibration Acceptable error as per our SOP is 2% in percentage range and 5% in PPM range (I don’t know the source of this error limits). We do follow this criteria and update calibration cylinder values mostly if the error is beyond the limits that are 2 % in percentage and 5% in PPM range.
Mr. Rod’s example can be quite close to our application in which we measure CO2 range 0 to 5 % (it is normally 2 % in sample) balance in Ethane + Ethylene. The calibration cylinder value is 2.5 % (± 2 %).
During Calibration (external standard) if CO2 averagely reads 2.53 which is 1.2 % of the cal value then we don’t update new calibration value to eliminate the difference and if it reads 2.56 % which is 2.4 % of the value than we update the cal value (in other word then we calibrate the analyzer).

(In PPM case, if CO2 cal cylinder value is 8 PPM (range 0 to 10 PPM) and on calibration run the Analyzer / GC reads 8.3 PPM which 3.75% of the value then we don’t update new Cal value (Response Factors) but if Analyzer / GC reads 8.5 PPM which is 6.25 % of the value then we update the cal value (RF etc).)
We are ISO certified, A few years back, In one ISO audit (by DNV) The auditor wanted to know that how we have come up with 2 % and % 5 numbers, and if it is true than how can you determine that your error (accuracy) is less than 2 % or 5 % when you have different accuracy working in your system, like accuracy of your Analyzer reading given by GC manufacturer and accuracy of your calibration cylinder given by cylinder manufacture.
Our all answers were not traceable to any international STD. And still today our operation staff asks us the same question.

Top of pagePrevious messageNext messageBottom of pageLink to this message  By Anonymous on Thursday, June 5, 2003 - 06:01 am:

Mr. Wasim wrote:

"We are ISO certified."

"Our all answers were not traceable to any international STD."

With tongue in cheek, this is the source as well as the solution to your inquiry.

"And still today our operation staff asks us the same question."

The Process analyzer is not engineered to measure the quality of the final product, but to control and monitor the process making the final product.

The 2% and 5% parameters are values determined by experienced chemical engineers to be practical limits when re-calibrating these control units.

As the unit ages and the parts wear, changes in response are normal. An engineer has to decide when a change in response justifies a recalibration.

This is based on experience and the economic return. How often do you pay someone to recalibrate the speedometer on your car as the tires wear down and decreases your actual speed even though the speedometer still reads 100km per hour? When it is off 2km per hour or 10km per hour from your actual speed?

Being ISO, document what you do and do religiously what your documents say you do.

And don't let an inspector tell you how to run your business, but run your business as you document for ISO.

Best wishes,


Add a Message

This is a private posting area. A valid username and password combination is required to post messages to this discussion.