The function of a spectrometer is to produce analyses. The fact that spectrometers consist of components such as an energy source and stand, optics, PMT, electronics and computer and that these in turn consist of sub-components and individual parts is often only of secondary interest to users. In a comparable way,a car-driver(only)wants to drive a car and a television viewer(only)wants to view television.

    If faults have occurred in a spectrometer which has been commissioned in good order. these faults must be removed. It is then necessary to test that overall operation of the anectrometer is satistactory by measuring the RSD of the recalibration samples.

     If faults have occurred in a spectrometer which has been commissioned in good or-

der. these faults must be removed. It is then necessary to test that overall operation of the anectrometer is satistactory by measuring the RSD of the recalibration samples.

     If the user has not made a clear statement of defects(eg. values are fluctuating exces-

sively), an effort must be made to specify them. A user unaccustomed to the terminology(and a service technician)may understand "fluctuating” as meaning that:

      1. measurement accuracy on one and the same sample is not good, that is, for individual or for all analytes;

      2. measurements on one and the same sample at different times(on 1 day, in 1 week etc.)

      3.spectrochemical analyses,in some cases or in all cases, do not agree with chemical analyses.

     There are different reasons for each of these three variants.What is good or bad re-

quires clear definition.

      If no definite fault is reported, and no definite defect is found on the spectrometer, its efficiency must be tested in the form of the RSD, for which recalibration samples must be measured.

      If the actual values are found to be within the nominal values a stability test should be carried out. This long-term test should include a pause of 1 h and the start and/or the end of a night. After these pauses 3.10 measurements are taken.

      If the stability test result is within the nominal limits, the calibration must be checked

with some check or reference samples. In so doing, so far as possible, samples should be used which were also used for the calibration.

       Note: Watch the RSD of references for the corresponding base, namely at least one reference line for each optics system. The RSD should be 0.5-1%.A check on the BEC can also be useful.

      If these tests lead to detection of a defect and to work being done on the spectrometer confirmed by measuring some recalibration samples. If these tests do not lead to detection of the fault, the user's criticism usually relates to non-agreement between the spectrochemical and wet chemistry or other values. If it is a question of systematic differ-

ences (although the reference samples agree), agreement can often be obtained, with the user's consent,by a small displacement of the recalibration HP. This should be noted in the service report and signed by the user.

     Under no circumstances should spectrometer-specific parameters such as the electrode gap, discharge parameters, pre-sparking and measurement times, interference equivalent etc.be altered. If this should prove necessary in special cases (if the user wants additional calibration for special work eg. weg. wire, free-cutting steels),separate analysis programmes should be established. Whenever the analysis range is to be altered or extended the operator who carried out commissioning should first be contacted.