Hello,

the problem is the light source which is aging rather fast. If you are lucky the useful life of the lamp is a 4-digit number of hours.

Giving the lamp time to stabilize does on the other side promote aging. Regarding the short life, it is wise to calibrate really frequent - daily :)

If you are operating the instrument continuously for an extended period of time, it may even make sense to calibrate every 5 or 10 operating hours.

Your microbalance is different: handled properly it does not age due to use.

Dear Rafael,

First don't forget that you measure in absorbance unit which has to be constant even if you can have a very small decrease between days.

Second, I propose you to plan an experiment so that you can have an idea of this need  :

You analyse different calibration samples during a week. At the end you develop a calibration curve with all these samples. In the last day you also develop a calibration curve with your usual procedure. Additionnally you prepare few known samples as a blind test. At the end, you use the two calibration curves to predict. You just have to compare results in order to answer your question.

Ludovic.

Hello Rafael

It depend on the light source life time, the materials you put your samples there If the manufacture change the cuvette materials ( they will change the Lot No. also) you should to recalibrate 

but you should take care about dust it make a big variation in the results if there is lenses or the lamp pulp (some lamps there is a caution don't touch the pulp)  or the cuvette are dirty so you should first clean 

Mina

I believe that you must have your SOPs in place that direct and mandates the calibration of each of the instrument that you use on daily basis must be routinely carried out. One may feel this job cumbersome or tiring, but to avoid hassles at a later point of time, it is better to calibrate the instrument before you start your day.

All the best!

Dear Rafael,

Light source aspects: The more traditional spectrophotometers use a Deuterium Lamp for the UV continuum, and a tungsten-halogen lamp for the VIS/NIR range. It is important to know that the intensity of a Deuterium lamp slightly varies per time that you start it up. This is an inherent property of such lamp types. Furthermore the quartz window ages (solarization). The near UVA and blue range of a Tungsten-Halogen lamp are very sensitive to tiny variations in lamp current. Both lamp types also age. With the appearance of continuum sources, stability and lifetime are quickly improving.

Calibration standard aspects: If one uses a certain dilution series for calibrations over a prolonged time, then the aging of those samples directly affects stability.

Spectrometer aspects: A double beam instrument (or a single beam that can subtract two consecutive measurements) offers improved stability.

Modern solid state detectors can be kept stable a lot easier (and age slower) than photo-multipliers.

Where do you TAKE your instrument? I saw that you are in the Agricultural discipline. Anyhow, in general, spectrometers with dual dispersion do much better than single grating instruments. But taking a dual slit mechanically scanning spectrophotometer into the field is quite impractical. Solid state duas dispersion spectrophotometers without moving parts, using e.g. an Echelle grating, are much more robust.

How is the "sending-in and uptake" path of your light? Lousy sample holder mechanics can be a big trouble. But also other aspects are important: In our Van 't Hoff Shared Innovation Program on Medical Photonics we do wideband (350 - 1830nm) contact spectrometry during surgery with very high accuracy. We designed our sterile fiber probes so that the mechanical surgical handling hardly affects the spectral transmission.

The best book I ever bought about this topic is:

Reliable Spectroradiometry

Henry J. Kostkowski

Library of Congress

Catalog Card Number 97-91743

1997

ISBN 0-9657713-0-X

Note: Only 1000 of these were printed, I got mine from Henry, don't know how easy you can buy it nowadays. But it is a GREAT book.

NIST in the USA and PTB in Germany are good in these topics. Added some stuff from the time I co-operated with PTB from my lecture archive as PDF.

Lamp and detector stability: these factors are not relevant to the question, unless it's the cause of excessive noise. The reason is that a spectrophotometer takes the ratio between the power transmitted for the sample vs a blank. This is built-in to double beam instruments, otherwise you do it by interchanging samples and blanks.

I think the point of your question may be that, since the specific absorbance is a physical constant of the analyte, you should never need to calibrate, provided the instrument(s) is both validated ('qualified' in the pharmaceutical jargon) and calibrated (nowadays with a bought-in filter or reference cell).

We can assume that the path length of a cell is close enough to nominal to be considered exact, and that the cell material is chosen to be transparent at the wavelength used (apart from surface reflections).

Consequently, for many purposes you may be able to determine the specific absorbance by following a written protocol (like a method validation protocol), with uncertainty analysis. An exception would be a derivatisation method where reaction yield and impurities due to the reagent might vary.

Having said all that, in practice we (or the regulatory authority) may prefer to calibrate "on the day". The uncertainty analysis may show that that is more precise and accurate. It also forms part of the system suitability test; you need to prove that both the analyst and the instrument are performing as expected.

The usual approach is to set up a method validation protocol (if possible >1 analyst and > 1 instrument), according to the rules of your "quality" system. If that demonstrates that the calibration curve is always a straight line through the origin, you can use single point calibration, generally in duplicate at least. The two points are separate weighings; serial dilutions don't give independent samples. The difference between the two standards gives a spot check on compliance with the precision that was determined during validation.

Bear in mind that the primitive signal (extinction = 1- transmittance) is an exponential function of the concentration. Don't go beyond 1.5 AU, whatever the instrument specifications may say; there are several physical and instrumental causes of non-linearity.

Finally, it's likely that the instrument data system does not use the most appropriate algorithm for the calculations: see my RG publication 'Calibration uncertainty in pharmaceutical analysis: replicate single-point calibration revisited'.

Dear Rafael,

The rate of calibration depends on the purpose of the measurement. For routine photometric assays, variation of parameters in a time elapsed from calibration to the measurement will be small enough to give rise to an appreciable increase in the total measurement error. (Note that the uncertainty in common analytical measurement including sample preparation is usually at the level of ten percent and more.) Then, the everyday calibration seems to be superfluous simply because the effort you make will be unjustified for the purpose. In any case, the experimental examination of the effect of daily or every-other-day calibration (usually made within validation studies) will provide the answer to the question.

It is often expedient to construct a "pooled" calibration line based on the data from that examination. This should be clearly stated in the written analytical procedure to follow. The one- or two point calibration check, also stipulated in the laboratory quality system, may be used to confirm that the “old” calibration remains valid.

I really appreciate the answers I've received so far. Together they bring constitute a surprisingly answer to what I had though of as a simple question.

Dear Dr. Rafael Andres Garcia, the first time when you build the calibration curve (regression line), I suggest you to calculate the standard deviation of the angular coefficient (slope). The next day, it is enough only checking if the new slope is within the limits of the standard deviation of the slope of the calibration curve of the previous day. If not, build a new calibration curve.

Best regards, Elcio Oliveira.

Sorry to come late to this "party."  Along the lines suggested by some others, since you have apparently already collected lots of day-to-day calibration data for your analyte, just plot the slopes and interecepts with uncertainties vs day and see if you have (1) a statistically significant trend, and more importantly, (2) a practically significant trend.  If not, over a number of days, use all the calibration data collectively to obtain a much more precise calibration -- meaning your unknown's uncertainty will be determined almost entirely by its measurements, and also, that you can use the normal distribution for confidence limits (since t-dist becomes normal for many points).

Of course, it is always wise to check calibration periodically, to ensure that something hasn't gone wrong ... especially if you get a surprising result for a new unknown sample.

Please, take a look at ISO Guide 35:2006, item B.5 Stability study and shelf life.

Best regards,

Hello,

does anyone know if it is possible to use an old calibration curve for the analysis of the same type of sample ? and how ?

The instrument is : AnalytikJena AAS 800D contrAA

thank you