CA1321486C - Monochromator drift compensation - Google Patents

Abstract

ABSTRACT OF THE DISCLOSURE

Iterative compensation of drift of peak positions of spectral lines in a spectral monochromator including a grating, a detector of spectral fractions of a spectral band, a stepper motor for varying relative orientation of the grating and the detector, and a computer. A series of computer-defined spectral windows each encompasses one spectral band and has a nominal spectral position and an initial spectral center. Each window is scanned such as to determine a peak spectral position. Calculations are made for a spectral offset of the peak position from the initial center for each corresponding window, an average of the offsets for the peaks as a linear function of window position, and a revised spectral center for each window equal to the initial center plus the average offset for the window position determined from the linear function. Each window is shifted correspondingly.
The step of successively scanning through each window is repeated such as to determine a new peak position for each corresponding band, whereby each new peak position is maintained near the spectral center of each corresponding window.

Description

~ PATENT
132~486 ID 3809 MONOC~ROMATOR DRIFT COMPENSATIVN

This invention relates to spectral monochromator~ and particularly to a method and a computerized system for compensating for drift of peak positions of spectral lines in a monochromator.

BAC~GROUND OF THE INVENTION

~arious types of optical spectrometers are in use for such purposes as atomic emission spectroscopy, atomic absorption spectroscopy and astronomy. A complete syste~ ge~erally consists of a source of radiation, a spectrometer for separating and detecting individual spectral components~ and a data station ~or processing the inormation from the spectrometer. The radiation source, for example, may be a system for injecting a test sample into an inductively coupled plasma where t~e atomic species in the sample are excited to radiate characteristiG atomic emission.
As another example9 a sample is evaporated in a graphite furnace where the gaseous sample absorbs certain fre~uencies of the incident radiation to provi~e ato~i~.absorption lines.
Similarly, astronomical sources provide atomic emission and 2û absorption lines.

The type of spectrometer of particular interest herein involves sequential measurement utilizing a monochromator in which a grating or prism is rotated to direct a narrow portion of the spectrum to a slit and a detectorO The angle is adjusted to 25 correspQnd to the different emission (or absorption) lines of the elements A A single detector i5 used, either a solid state , i, .,, ' , ,, ' ~ ~ ", '' ' ' ' , : ' ` ~2~
II)o3B09 detec~or or a photomultiplIar tubeO The measurement process involves rotation of the gratinq with meas~remen~s at each of a series of selected locations corresponding to grating angles appropriate to the atomic emission line~, S Sophi~ticated monochrom~tor~, particulasly of the t~pe used for quantitative analysis of a~omic elemen~s in sample~ injected through an induc~io~ coupled plas~ are con~rolled by microprocessor~ and personal ~omp~ter~. Such a sy~te~ is typlified by a Model Plasma 40* emission spectrometer sold by The Perkîn-Elmer Corporation, Norwalk, Connecticut, and described in U.S. Patent No. 4,779,216, issued October 18, 1988 (Collins) assign~d to the assignee of ~he pre~ent appllcation. A s~pper motoe orient~ a grating wit:h respec~ to the ~lit of the de~ector to locat~ any ~elect~d portion of the spectrum for measurement of th~ i~tensity of that portion. A
dedicated microproces~or provide~ ~ ~uitable ignal to he motor for ~el~c~ive orientatio~ in r~l~tio~ to wav~lengthO The ~icroprocessor al~o rec~lve~ th~ i~t~n~ity ~ignal from th~
detector, and provide~ d~ta in th~ form of ~pectr~l intensity vs spectral po~itionO Xn pr~ct~c~ a scanning ~ign~l i8 provided ~o ~he motor to sequen~ially ~can th~ ~pectru~ in a ~erie~ of step~.

In ordeE to ~llow a reason~ly f~t ~cany ~ign21~ to ~he motor ar~ ~uch ~ to sc~n in ~pectr~l window8 which a~ ju~t wid~
enough to encoDpa~s ea¢h o~ the selected sp~ctral band~ with some m~rgin. The ~otor sc~n~ t~rough ~ tep~ in ~ ~indow~ and then ~oves quickly to t~ next w$ndow b~for~ ~c~nning in ~tep~ again, and on to the next window, ~te.~ ~o~ th~ whol~ ~er~ o~ spec ral b~nd3.

Th~ aforem~ntioned Collin8 re~erenc~ di~clo~ a syste~ for initially calibrat~ng th@ window po~itions for th~ de~ired * Trade-mark `

11 321~6 I~-3809 spectral lines in a monochromator to compensate for mechanical imperfections in i~s diffraction grating and grating drive assembly~ The disclosed system employs a two-~tage interactive procedure. Each stage involves measuring position errors of lines for a standard element and fitting these errors to a ~uadratic polynominal by the least squares m~thod, as a function of window position. An iterative~ sel-consistent, discrete Fourier transform is used for the determination of multiple positioning correction termsO When the Fourier calculations are completed, the results of the calibration procedure are presented by the system to the analyst for acceptance. If accepted, the positioning error of the primary calibration line is measured, stored and used by the system to establish a zero centered distribution o~ positioning errors each time the monochromator is reinitialized.

~owever, the above-described calibra~ion may be insufficient, especially for long-term operation~. ~emperature changes cause minute distortions in the spectrometer resulting in drifting of the positions o~ the spectral bands with respect to their 20 windows. Without compensation, over a period of time a peak may drift through an edge of its window resulting in erroneous, undetected data. As temperature controlled rooms are often impractical or insufficient, a common method for minimi~ing drift is to temperature control the monochromator with a built-in heater and optionally a thermostat. Temperature control system has been found to be less than satisfactory because it is difficult to control all components uniformly without adding substantial cvst and complexity to the apparatus~

.... ., . .. . , - . . . . . . . . . - , . , . . , . - ..

. ` '.' ``;, ' ~`. ' ~ - . ;~. ''.' `- , . . - '.` ., ':, '.' ;,' , ~ . ' '. .`,' -:
.
' ~ ~ 2 3L 4 ~ 6 ID-3809 SUMMARY OF ~HE INVENTION

Therefore an object of ~he present invention is to provide con~inuously updated compensation in a spectral monochromator for instrum~nt drift, such as may be due to temperature variations~
5 A further obJect is to provide such compensation at relatively modest cost. Another object is to provide such compensation by means of computer data proces~ing.

The foregoing and other objects are achieved in a spectral monochromator including radiation means for generating elemental 10 radiation for atomic element~, optical dispersion means receptive of the elemental radiation for producing a series of spectral bands each having a peak spectral posi~ion associated with maximum radiation intensityr and detector means selectively receptive of each spectral raction of each spectral band ~or 15 producing an intensity signal representative of the intensi~y of the spectral fraction being received. Scanning means are responsive o a scanning signal or varyin~ relative orienta~ion of the dispersion means and the dete¢tor means such that the serie~ of spactral bands is scanned across the detector means in 20 spectral positions corresponding to successive spectral frac~ions. A computer is receptive of the intensity signal for t yenerating the s~anning signal and calculatin~ spectral po~ition of each peak.

According to an embodiment of the present invention a method 25 comprises7 sequentially, establishing a set of spectral windows each encompassing one of a serie~ of selected spectral bands and ~avlng a nominal spectral position and an initial spectral center t successively scanning through each window suçh as to determine a pea~ spectral position for each corresponding band, 30 calculating a spectral ofset of the peak position from the . ~ .
': :
-1 3 2 1 1~ ~ 6 ID-3809 initial center for each corresponding window, calculating an average o~ the oset~ for the peaks, calcula~ing a revised spectral center for each window equal to the initial center plus the corresponding average offset for the nominal window position, and shifting ~ach window so as to center on its corresponding revised center. The step of successively scanning through e~ch window is repeated such as to determine a new peak position for each coxresponding band. Thus each new peak position is maintained near the spectral center of each corresp~nding window.
The method is especialLy suitable when successively repeating the sequence o steps on successive samples.

Preferably the step o~ calculating an average of the offsets comprises calculating a linear function of offsets vs. nominal window position, and determining the average offset for each }5 nominal window position from the linear functionn According to a further embodiment, the monochromator further comprises a compensatin~ system for compensating for drift of peak positions of spectral lines, comprising window means for establishin~ a series of spectral window~ each encompassing one of a series of selected spectral bands and having a nominal spectral position and an initial spectral centerl first scanning means for successively scanning through each window such as to ~etermine the peak spectral position for the corresponding band, first of fset means for calculating a spectral offset of the peak position from the initial center for each window, second offsat means ~or calculating an average of the ofsets for the peaks, center means for calculating a revised spectral center for each window equal to the initial center plus the corresponding average offset for th~ window position, shiting ~eans for shifting each window so as to have its revised center, and second scanning means for repeating the step of successiv~ly scanning through .

13 2 ~ 4 8 6 XD-3~09 each window such as to determine a new peak position for each corresponding band. Each new peak position is thereby maintained near the spectral center of each corresponding window.

BRIEF DESCRIPTION OF TH~ DRAWI~GS

FIG. 1 is a schematic drawing of a spectral monochromator for carrying out the invention~

FIG. 2 is a graphical plot of a spectral band processed according to the inven~ion.

FIG. 3 is a flow diagram of a program for establishiny an initial 10 spectral position of a window according to the invention.

FI~. 4 is a ~ow diagram of a program for compensating for drift according to the invention.

FIGS. 5-6 are flow diagram for groups of steps indicated in FIGr 4 ~IG. 7 is a graphical plot of results of a run made according to the invent:lon.

DETAILED DESCRIPTION OF THE IN~ENTION

A spectrographic system for carrying out the present invention is shown sch~matically in FIG. 1, which, for purposes of illustration herein, is of a type comparable to the aforementioned Perkin-Elmer Plasma 40 emiss~on spectrometer, There are~ broadly, three components, na~ely, a sourc~ of radiation 10, an optical spectrometer 12~ and a data station 14 .

1321~86 ID-3809 Radiation source 10 produces infrared, visible andior ultraviolet radiation generally characteristic of atomic elements. The source may be, or example, an inductively coupled plasma 11, driven by a power supply 13, into which a sample of test material . 5 (sometimes known as an~ly~e) is injected by a sample aspiratsr 15~ Alternatively the source may be a grlaphite furnace or the like operating to provide emission lines or absorption lines of atomic elements, or extraterrestrial with light being collec~ed by an astronomical telescope.

Sample aspirator 15 injects into the plasma a mixture of a sample (e.g. unknown) material for ionization and analysis, and (optional) selected reerence atomic elements as desired for calibration and/or standardi~ation. Thus an input beam 1~ from the source is characteristic of a plurality of atomic elements including the sample and references (if any~

The ~ubsequent components, namely optical syste~ 12 and data station 14, provide a quantitative measurement of the atomic elemen~s associated with source 10. Optical sy~tem 12 is of the conventional or desired type that produces a display of spectral lines, and in the present example is of the grating type.

Light 16 from source 10 reflects from a first toroid mirror 17 passes through an entrance pupil 18, reflects rom a second toroid mirror 19, and passes through an entrance slit 21. Ray~
22 are then reflected by a concave collimator 23 to a reflective grating 20. Di~persed rays 2~ reflected in a spectral pattern rom the grating are passed to a concave spherical re~lector 25 ~; which focuses the ray~ selectively through an exit slit 26 to a detector 27~ Th~ detector i~ preferably a photomultiplier tu~e ~0 (PMT) for maximum sensitivity, although a solid state detector may be used.

. :

ID ;~3809 ~2~

A stepper motor 28 is operatively connected to rotate the grating by way of a conventional drive mechanism 2g of g~ars~ pulleys and belts or ~he lik~ such as disclosed in the aforementioned Collin~
reference. The stepper motor thu~ orierllt3 grating 20 with S respec~c ~o ~lit 26 o~ th~ detector means to locate any selected portion of the spectrum for measuremen~ of the intenslty o that portionO

A dedicated microproce~so~ unit ~2 (C3?Uj provid~ a ~uitable signal through an input/output board 43 and a driver ~4 'co motor 10 28 ~or selectiv~ orientation. Microproces~or 42 also receives an intensity ~ignal from d~tector 27 ~ er an analog/d~g.Lt~l conver~cer 45 and a second input/output board 4~, and provideR
output dat~ in th~ form of ~p~ctra~ t~nsity fo~ ~a~h ~p~ctral po~ition~ Th~ latt~r po~ition i~ fundamentally wav~length (or 15 frequency) bu~e for purpo~e~ of interllal eoD~puta ion~ and control is conveniently tho st~p po~ition og th~ ~oto~ n ~tandard prac~cic~ a scanning ignal i8 pro~ridedl 'co th~ ~oto~ ~uch a~ to se~u~ntially scall por'cion~ o~ th~ 3pectrum.

To provid~ for ~urther control and al~o to iEIple~ent th~ pre~nt 20 invention as de~cr~b~d h@reinb~low;, ~ ~econd data process~FIg unit 48 ~u~h a~ a por~onal co~pu'cor (PC) is~ pro~r~ d for fllrther proce~ing o~ th~ infor~tios~l on intensity and posi~ion. The PC
commulaic~te~ with microproce~o~ 42 v~a an IE~B~8 cabl~ ~g. PC
~8~ CP~ ~2~ and ~h~ lnte~connecti~g so~ nts ~lth ~pectromet~
25 12, collectiv~ly cons~ ut~ d~ta ~t~cio~a 14. D~ta ~tatiosa 1~
alternatively may b~ incorpor~t~d into a singl~ ¢omputer or part of a cent~ yst~m to ~ffeet the r~u~red d~t~ p~oce~sing, and i~ broadly t~med ~computet~ o~ ~computer ~e~n~ hereln and in th~ claims.

* Trade-mark 132.~8~ ~D-3809 The spectral ~lines" from the atomic elements of a sample being analyzed are each actually in the form of a bell shaped band 50, as illustrated in FIG. 2. This band is shown drawn on a horizontal position axis P with t~e step I)ositions 52 of the stepper motor with widths marked thereon c:orresponding to the spectral width o~ slit 2~ ~FIG. 1). The detector measures intensity displayed on the vertical axis I of each spec~ral fraction corresponding to a step pocition t as indicated by the stepped intensity line 5~ in FIG, 2. The purpose of the monochromator being to provide quantitative data on line intensities and, correspondingly, atomic elements, the intensity of the peak 56 associated with ma~imum radiation inten~ity is also det~rmined. The peak intensity and spectral po~ition 58 of the peak are computed conventionally in the computer by three point fitting of a parabola to the step 60 with highest intensity and the two steps 62,62' on either side o~ the masimumD The actual position of peak 56 thus may be expressed within a fraction of a step position, ~or ~he purposes of further calculations presented hereinbelow~

In order to provide for a fast scan, signals from computer 1~ to motor 28 (FIG. 1) are such as to scan only in spectral ~windows~
which are selected to encompass with some margin each of the bands that are selected for measurement. A typi~al window is shown in FIGo 2 as defined between the two edges 64,66 shown as 25 broken vertical lines, Thus microprocessor 42 is prograrrlmed ~co provide scanning signals for the motor to scan through all s~eps 52 in a window, and then move rapidly to the next selected window where the scanning steps are made again, and on to the next window, etc., for the whole series of spectral bands in a run.
30 With each step being e.g. 0.022 nm or an ultr~violet grating, a window should be between about 8 and 70 step~ wide/ for example 30 steps. The computer program establishing the windows and ;, . ; : ,.

~ ~ 21~ ~ 6 ID-3809 signalling the motor accordingly is, for example, the same as used in the aforemen~ioned Collins reference.

Because of the precision nature of the optical spectrometer, temperature changes cause minute distortions in the apparatus resulting in drifting of the positions of the spectral bands. In other words, a stepper motor setting that aligns a peak within a window in one run may not exactly do so in the next rusl; in other words, ~he peaks drift with respect to their windows. Without compensation, over a period of time a peak 56 may drlft: through 10 an edge 6~ or 6fi, and out of the window, resulting in unmeasured intensities and erroneous data, and this is the problem to which the present invention is directed. More specifically t.he invention is directed to maintaining each peak near a spectral center 68 (FIG. 2) of the corresponding window.

15 As an additional conventional step, ~or the purpose of determining inten~ities a radiation run is preliminarily made across each window without any elements bein~ introduced. This baekground is subtracted from certain measured intensities to provide a net in~ensity indicated hereinafter, It will be appreciated that, for less accuracy or very low background, this step may not be necessary.

A flow diagram illustrating the initial (standardization~ step~
is presented in FIG. 3 for a run in which the initial positions of corresponding windows are ~elected and deined by running 25 standards. This seguence is conventional as in the Perkin-Elmer Plasma 40, but i5 presented herein for elarity.

Referring to FIG. 3, the desired speetral bands or analysis are selected (72), and drift 6truetures and offsets are initialized (733~ ~ s~andard sample is aspirated into the plasma (7~) for 1 3 2 ~ ID-3809 ionization, and counters ~re initialized ~75). E~timated preliminary peak positions for the selected spectral bands of at least one standard source containing known atomic elemen~s with spectral lines for each of the selected bands are utilized for defining preliminary spectral centers of t:he windows (76). ~his run may be made with wider than normal windows defined to assure finding the peaks initially, for example twice as wide.

Offset corrections are made to the window center at this poin~
~77). Initially these are zero7 Centering offset is determined as part of the standardi2ation o~ FIG. 3; drift offset may apply at this point in a restandardization only if determined subsequently according to the present invention and then a restandardiza~ion run is made. The window position is redeined to its new center (781, and the spectral band is scanned for its maximum radiation inte~sity and position (79)O A new offset is computed ~80) as the old offset (after previous run if any; zero if no prior run) plus ~he difference (e.g~ in motor steps) between the peak location a~d the curre~ window center. The subsequent steps ~81~ are directed to calculating intensity, detPrmination of a highest intensi~y for several runs (Wreplicates~ on standards, counting, moving to the next window, and running with further standards. After the last standard run~
the standardized Hcenter offset~ for each window (at 80) is established as an initial center utilized as a basis for the further steps described below.

The method and means steps for carrying out sample runs which make drift offset corrections according to the invention are aepicted in the fl~w diagram of FIG. 4. Elements of this and subsequent diagrams also represent schematically portions o the system, including the computer programi that carry out the invention, iOe. the means for ~arrying out the steps.

~-3809 ~L 3 ~
After initializing offsets to zero (90~ (the first time) radiation is provided from a sample analyte (91), typically an unknown for which a quantitative atomic analysis i5 bein determinedO Counters are initiali~ed (92,93). The center of the 5 window is first defined with its ini~ialized position (94), an~
then corrected ~95) for the center offset (described above) and a drift offset which is as described below and is zero the first run. A window is scanned and the spectral pos~tion and inten~ity of a peak, if any, is measured and stored (96). The peak error ~offset) is prepared and screened ~97~, as described further below, and some peaks may be rejected; i.2. not used in a correction calculation. A counter is incremented (g8) and the next window band is scanned t99~ 94). For a next replicate another counter is incremented (~OO)~

An average of offsets is determined next (101~. Generally there is some scatter among the values for the various drift offsets, and an average value was deemed to be necessaryO A simple average may be quite satisfactory~ ~owever, it was further determined that use of a simple average (sum of offsets divided 2G by the number of peaks) may not provide suitable correction across the entire spectrum, fre~uently being too high or too low at the ends of the spectru~. -It was then discovered that there tends to be a general increa~ein offset with window position when the temperature is rising, and a ~imilar decrease in off~et with window position ~hen the temperature i8 dropping, at least with the Plas~a 40 instrument being usedO (Other instruments could have opposite trends, but the principle~ and application of the invention herein still should apply.) Thereore t according to a preferred method step of the present invention, and continuin~ with reference to FIG.

,.

1321~ ID-3809 4, an average offset is determined from a simple function of offsets vs. nominal window position ~101). It generally is satisfactory that the simple functiorl be a linear function. For better accuracy i~ may be desirable as a refinement to add a small quadratic term to the linear unction to account for some apparent curvature. ~owever, any complex function accurately describing the scattered values would defeat the purpose of providing an average offset value or each window. The phrase ~average of the offsets~ as used herein and in the claims means either a simple average or an average determined from a simple function.

The simple function , e.g. an equation for a line, is readily determined by the computer using conventional algorithms ~or the least squares method. Th~re should be at least two offset~
available from a run to determine the function. The average offset for a window is determined by en~ering its nominal position into the equation. A further test (described ~elow) is made to determine whe~her to even use the function to update offset~ (102)o It is convenient and sufficiently accurate to utili~e nominal position for the window for calculating the function to determine average o~fsets, such as the preliminary center or a boundary of the window. The number of total step scanning positions across a spectrum may approach 3~0,000 or more, so small differences in defini~g nominal window position for the linear function are inconseqllential.

~pon selection of the same or another sample (103), the program returns to step 91 or 93r and the next window is selected (94).
A revised spectr~l center for each window is calculated ~95) as equal to its initial center position plus the corresponding .: , : ' ` ` ~ "' ' :
, ,, :, : ~, : ' ,:

1321~
updated average drift offset from step 101 of the prior run, as well as the "center offset" determined frum the standardization runs. The window is shifted by the computer so as to have it center aligned with its revised center. Another run 1s made, with the same or other sample, and the sequences of FIG. 4 for determining offsets and adjusting window positions are repeated for successively continuing runs (106~ on other samples. It may be seen that any drift due to temperature changes or the like are compensated with continuous updates. The inven~ion is particulary useful for many repeated runs made automat:ically over an extended period of time, and each peak position is generally maintained near the spectral center of each window.

Drift offset preparation (97) including screening of the peaks is effected in or~er to account for situations where the data available for determining offsets may have no or questionable validity. This screening for peak shape and intensity is made readily in ~he computer by subroutines with conventional conditional statements which determine if values are greater or les~ than predetermined thresholds. Referring to FIGS. 5A, 5B
and 5C each of a set of subroutines help~ determine whether each offset is sufficiently valid for use in calculating the function.
Peaks are selected or rejected for purposes of offset calculations, not necessarily for the anaiytical anaiysis based on intensities. The æubroutines are carried out for each windvw.

FIG. 5A summarizes the program flow for drift correction (offse~) preparation. Determination is made for good shape (110~
(detailed below for FIG. 5~, and position error (offset~ is computed (111). Position error is also squared and the result is stored (112) for future use. Determination is made next or good intensity (113) (as detailed below for FIGo 5C~ If shape and intensity are both ~ood (11~), th~ offset value is stored and the 13 2 ~ ID-380S
program is passed back to point 97 in FIG,. 4; if not, the program returns to point 97 with the offset value marked as unusable (116), for proceeding to the next steps. A "force correction" is utilized later in the program.

FIG. 5B illustrates ~he conditional statements for deter~ining good peak shape, tak~n from point 110 (FI~. 5~)~

One si~uation is interference from one or more other peak~ in the window~ If a peak is not at a window edge (120), a three point fit to a parabola ~123), by a conventional algorithm comprising determining for each window a parabolic line computed from the parabolic equation Y=AX2~BX+C whe.re A, ~ and C are parabolic parameters, X represents successive spectral positions of spectral fractions in the corre~ponding window, Y represents the intensity of each corresponding spectral fraction. The computation of the parabolic line is made by fitting intensities for three points consisting of the peak and ~wo spectral fractions adjacen~ to and on either side of the peak (60, 62~ 62 of FIG. 2). A curvature parameter is defined as the value of parameter A, which should be a negative number~ (A negative value for curvature indicates the upwardly pointing status of a peak.) Exact peak location and intensity are determined from ~he parabola (123) in the conventional manner.

If the peak is only one ~tep away from an edge (12~), the computer program a~ks whether the curvature is sufficient, vis.
les~ than a first (neg~tive) criterion value (125); a curvaturs criterion of -95 is suitable.. If ~yesn, th~ program proceeds through a fsrce correction set (126) (for future use) to a noise test (130). If "noa/ further refinement is desirable and a five point parabola is fitted (127~ with the peak spectral fraction0 and each of the two spectral fractions on either side of the ., . ', ' ' ,, , ' ,: ' ' ~ ~ 2 ~ D-380g peak. Curvature parameter A is recalculated, and the sum of the squares of the errors of these five points from the parabola is calculated~ ~his sum is divided by ~he curvature to define an error ratio (114).

S Sets of preferable acceptance criteria for curvature and error ratio have been determined as set forth as examples in the Table herein; ~wo such sets ~128,129~ are depicted in FIG~ 5 for illustration. An offset is used for calculating the simple function only if the combined criteria in the set are met. As a minimum requirement, the c~rvature should be less than -7S and the error ratio should be greater than -1000.

Table Combin~d Cri~eria for Acceptanc~ o~ Offset Curvatu~e Error Ratio C-~3 ~any) -43 to -30 > ~50 ~2g.9 to -2~ 20 -20.9 t~ -11 >-45 -10.9 to -3 >-2 If the curvature and error ratio criteria are met, so~ewhere in the Table, and if the (negative~ curva~ure is greater than a urther threshold (131), e~gO greater than -35 (i.e. a rela ively flat peak~, then a signal/noise ratio is checked (132). Signal is peak intensity less background. ~oise may be defined several ways~ A suitable way is to u5e the intensity of the nex~ highest peak in the window or, lacking same, the lowest intensity in the window; background is subtracted from either. The peak passes :132~4~6 the test if the ratio is greater than a selected value such as 3.5.

The next test is another noise criterion ~130), which may be any conven~1onal or desired test~ In the present example, the three S poin~ parabola il23) determines maximum intensity YMAX; the highest intensity of the ~hree spec~ral fractions fitted is selected a~ Y2; the spectral intensity for the righ~hand spectral fraction adjacent to the bighest inten~ity is Y3; and the value of curvature A is utilized. If YM~X>l.lY2 or if ~(Y2 -Y3~ <-0.2, the peak is deemed to be too noisy and is rejectecl from theoffset error calculation.

A further problem to account for is a peak lying at or beyond the edge of a window~ Thi~ may occur after a series of runs with samples having no element~1 spectral band for the window and~
thereore, no offset correction for a while. It is desirable to pull a line back into the window even at the expense oP some accuracy for the linear function calculation. ~his pull-back is effected with a subroutine to recognize a tail o a peak off the edge, and, i so, apply two tests. The subroutine queries whether a peak intensity is measured abutti~g a window edge (120). If ~yes", then the first test (121) is that the intensity of the spectral fraction at the edge be greater than a pre~etermined threshold such as 0.5% and preferably greater than

2~ of a high intensity standard (described below)~ The second test (122) is that ther~ should be at least two and preferably four additional ~pectral fractions adjacent the abutting fraction having successively decreasing inten~ities. A ~no~ for either test rejects the peak. Thu~ peak shapæ in subroutine yroup 110 is determined to be good ~133) so the peak off can be used, or bad (134) so the peak cannot be used, and the flow is passed back to the main program.

~ -: .. - : ;:

~ 3 ~
The next screening situation is peak intensi~y. Referring to FIG. 5C, ~aken from poin~ 113 (FIG. 5A), if a standard intensity exists (135), a net standard intensity IA is determined as a high intensity of a selected standard i~ high concentration, less background (136); and a net sample peak intensity IB is determined as the peak intensity, less backgrou~d (137). A
relative threshold intensity value IB~IA should be exceeded ~138) before an offset may be utilized for calculating the average of offsets; otherwise the peak is rejected. An ab~olute minimum may be selected, but the system is more versatile if correlated with a set of standard peaks. For example, under routine calibration, several standards with different intensities are run for all selected æpectral bands and windows. It is considered to be desirable to determine the threshold intensity as equal to a predetermined threshold percentage of a selected standard sample peak intensity, the selected peak being the highest of all the standard~ run for the window. A threshold of, for example 0.99~
was fo~nd to be suitablea Thus intensity is either good (139~ or bad (139G) (peak reject~d), and the program is retu~ned to point 113 (FIG. 5A). If t~ere is no standard intensity available for ~he peak ~135), the ratio is set to zero (1353) and the intensity will be treated as bad and rejected (139').

Further screening is shown in the flow sheet of FIG. 6~ which is depicted at point 102 in FIG. 4, for testing data including the linear function itself for suitability. An un~uitable function is aborted with no ofset correction being made for the next run.

Correl~tion coefficient, conventionally defined from a computation for the fit to the simple function, eOg. the leas~
squares linear fit, is calculated ~14&). ~ormally the correlation coefficient s~ould be great~r than a first minimum ~":.''.,; . ~' 13~ 4~6 ID-3809 selected value such as 0.25~ for example greater than 0.31. This criterion is applied so long as the offset correction i~
successively increasing or successively decrea~ing such as due to consistent}y increasing or decreasing temp~erature. Yet another desira~le criterion is the simple average of offsets for a run (total of ofset ~alues divided by their number (1~1), as distinguished from the average determined from the simple function of window position).

Data indicating a temperature reYersal could be suspect r detected by a reversal in the sign of the simple average of offslets from the previous run (142). A related correction of window position is likely to be small anyhow~ Thus if th~ simple average did change sign, a mînimum correlation coeficient or other similar merit value o the function may be selected at a higher level such as twice the normal level, e.g~ 0.62. If the higher correlation coefficient is not exceeded, no offset correction is made to the window~ for the nest run. Typical cor~ela ion coefficients were found to bP generally around û~,9.

Returning to the flow diagram of FIG. 6, a determination is made whether the sign of simple average AV has changed ~1~2), if so, the correla~ion coefficient ~CC) is tested for being above the higher level criterion (eSg,. 0.62). The simple average is also tested to determine that the absolute value of the si~ple average i8 greater than the absolute value for the previous runO If the 2S sign has changed and if AV and CC exceed the criteria, an abort switch i~ set at ~no~ 3); otherwi~e at ~yes" (14~).

The simple average AV is agai~ tested for its threshold criterion (145~ ~e.g. 0.6) and, if ~ye~, the program proceeds ~o a query on a~ort set (146)o If AV is less than the threshold~ the program queries (1~7) if the force corre~ion is set to "yes~

. , .......... , , .................. , . - . .: , . ......... ..

.

13 2 1 ~ ~ ~
(126 of FIG. 5B); if ~yes" the program proceeds to the abort set query (1~6). This retains a peak if it is; one step from an edge.

If force correction (147) is not set to ~yes", the correlation coefficient is tested (14$) to exceed its normal lower threshold (e.g. 0~31). If ~yes~ the program sim}larly proceeds t:o the abort set query (146)o If "no~ the run data for drift is aborted (1~9~, i.e., not utilized for a drift correc~ion, and the program returns to point 102 (FIG. 4)O The run is similarly aborted (149) and returned if the abort set (1~6) is no~ set O~l ~non. If abort is set on ~no~ 6), it is determined whether there is more than one peak ~150) available for the function; if not, the run is aborted (149).

If there are sufficient peaks (150), standard deviations are determined (151~. If ~n o~fset is more than a selected multiple of its standard deviation, such as two and preferably three standard deviations away from ~he corresponding average offset (de~ermined from the linear function) (152), it is considered an outlier and the offset is rejec ed. The ~ample function is recalculated without outliers (153~ and the average of offsets is updated (15~).

As indicated above, the invention is particulary useful for many repeated runs made automatical~y over an e~tended period of time, with each peak position being maintained nea~ the spectral center o~ each windowO FIG. 7 illustrates the success ~f ~he present invention in compensating for temperature change~s and drit~
in particular the resul'cs of compensations made during a typical unattended eigh~een hour overnight, using the Lant~anum 398.B52nm line. The solid line 160 indicates room temperature~ The dotted line lb2 indicates uncorrected positions of the selected peakO
The dashed lines 16i~ indicate the corrected peak positionsO

~' , '' .

~ 3 2 ~
If the correction had not been used, the intensities recorded for Lanthanum 398.852nm would have been inaccurate for 17% of the instrument run (that is, during all times at which the peak position exceeded the edge of the scan window), Instead, accurate intensi~ies were recorded for 100~ o the eiyhteen hour overnight instrument run.

While the invention has been described above in detail with reference to specific embodiments~ various changes and modifications which fall within ~he spirit of the invention and scope of the appended ~laims will become apparent to ~hose skilled in this art. The invention is therefore only intended to be limited by the appended claims or their equivalents.

_

Claims (21)

What is Claimed is:

1. A method of continuously compensating for drift of peak positions of spectral lines in a spectral monochromator including radiation means for generating elemental radiation for atomic elements, optical dispersion means receptive of the elemental radiation for producing a series of spectral bands each having a peak spectral position associated with maximum radiation intensity, detector means selectively receptive of a spectral fraction of a spectral band for producing an intensity signal representative of the intensity of the spectral fraction being received, scanning means responsive of a scanning signal for varying relative orientation of the dispersion means and the detector means such that the series of spectral bands is scanned across the detector means in spectral positions corresponding to successive spectral fractions, and computer means receptive of the intensity signal for generating the scanning signal and calculating spectral position of each peak; the method comprising the steps of, in sequence:
Establishing a series of spectral windows each encompassing one of a series of selected spectral bands and having a nominal spectral position and an initial spectral center, successively scanning through each of the windows such as to determine a peak spectral position for a corresponding band, calculating a spectral offset of the peak position from the initial center for each corresponding window, calculating an average of the offsets for the positions, calculating a revised spectral center for each window equal to the initial center plus the corresponding average offset for the nominal window position, shifting each window so as to center on its corresponding revised center, and repeating the step of successively scanning through each of the windows such as to determine a new peak position for each corresponding band, whereby each new peak position is maintained near the spectral center of each corresponding window.

2. A method according to Claim 1 wherein the dispersion means includes a dispersion member and the scanning means includes a stepper motor receptive of the scanning signal and operatively connected to the dispersion member.

3. A method according to Claim 1 further comprising successively repeating in sequence the steps of calculating a spectral offset of the peak position from the initial center for each window, calculating an average of the offsets, calculating a revised spectral center, shifting each window, and repeating the step of successively scanning through each of the windows, such as to determine a further new peak position for each corresponding band, whereby each further new peak position is maintained near the spectral center of each corresponding window for successive spectral scans.

4. A method according to Claim 1 further comprising the step of utilizing each offset for calculating the average of offsets only if such offset meets a predetermined criterion.

5. A method according to Claim 4 wherein the step of utilizing each offset comprises measuring a peak intensity for each peak and utilizing each offset for calculating the average of offsets only if such peak intensity associated with such offset exceeds a predetermined threshold intensity.

6. A method according to Claim 5 wherein the threshold intensity for each peak is determined as equal to a predetermined threshold percentage of a standard sample peak intensity .

7. A method according to Claim 4 wherein the step of utilizing each offset comprises calculating a curvature parameter for each peak, and utilizing the corresponding offset for calculating the simple function only if the corresponding curvature parameter meets a predetermined curvature criterion.

8. A method according to Claim 7 further comprises determining for each window a parabolic line computed from the parabolic equation Y=AX2+BX+C where A, B and C are parabolic parameters, X
represents successive spectral positions of spectral fractions in the corresponding window, Y represents the intensity of each corresponding spectral fractions the computation of the parabolic line is made by fitting intensities for at least three intensity points consisting of the peak and adjacent spectral fractions adjacent to and on either side of the peak, and wherein the curvature parameter is defined as parameter A, and wherein the curvature parameter is less than a predetermined curvature value.

9. A method according to Claim 8 wherein each of the intensity points has an intensity error from the parabola, and the method further comprises calculating an error ratio defined as the sum of the squares of the intensity errors divided by the curvature, and utilizing the corresponding offset for calculating the simple function only if the corresponding error ratio exceeds a predetermined error ratio.

10. A method according to Claim 4 wherein the step of utilizing each offset comprises detecting each peak that is in a spectral fraction abutting an edge of a window, measuring a peak intensity for each such abutting spectral fraction, measuring corresponding intensities of at least two spectral fractions adjacent the abutting spectral fraction, and utilizing a corresponding offset for calculating the simple function only if the abutting intensity exceeds a predetermined threshold intensity and if the adjacent spectral fractions have successively decreasing intensities away from the peak abutting fraction.

11. A method according to Claim 1 wherein the step of calculating an average of the offsets comprises computing simple function of nominal window position.

120 A method according to Claim 11 wherein the step of computing a simple function comprises computing a linear function of offsets vs. nominal window position, and the average offset for each nominal window position is determined from the linear function.

13. A method according to Claim 11 further comprising calculating a standard deviation for the simple function, rejecting offsets that exceed a preselected multiple of the standard deviation, and recalculating the simple function without the rejected offsets prior to shifting each window.

14. A method according to Claim 11 further comprising additional steps of calculating a merit value for the simple function and shifting each window only if the merit value meets a predetermined criterion.

15. A method according to Claim 14 wherein the additional steps comprise calculating a correlation coefficient for the simple function, and shifting each window only if the correlation coefficient exceeds a predetermined correlation coefficient.

16. In a spectral monochromator including radiation means for generating elemental radiation for atomic elements, optical dispersion means receptive of the elemental radiation for producing a series of spectral bands each having a peak spectral position associated with maximum radiation intensity, detector means selectively receptive of a spectral fraction of a spectral band for producing an intensity signal representative of the intensity of the spectral fraction being received, scanning means responsive of a scanning signal for varying relative orientation of the dispersion means and the detector means such that the series of spectral bands is scanned across the detector means in spectral positions corresponding to successive spectral fractions, and computer means receptive of the intensity signal for generating the scanning signal and calculating spectral position of each peak; a compensating system for continuously compensating for drift of peak positions of spectral lines, comprising:

window means for establishing a series of spectral windows each encompassing one of a series of selected spectral bands and having a nominal spectral position and an initial spectral center, first scanning means for successively scanning through each of the windows such as to determine the peak spectral position for a corresponding band, first offset means for calculating a spectral offset of the peak position from the initial center for each window, second offset means for calculating an average of the offsets for the peaks, center means for calculating a revised spectral center for each window equal to the initial center plus the corresponding average offset for the window position, shifting means for shifting each window so as to have its revised center, and second scanning means for repeating the step of successively scanning through each of the windows such as to determine a new peak position for each corresponding band, whereby each new peak position is maintained near the spectral center of each corresponding window.

17. A compensating system according to Claim 16 further comprising screening means for utilizing each offset for calculating the average of offsets only if such offset meets a predetermined peak criterion.

18. A compensating system according to Claim 16 wherein the second offset means comprises function means for computing a simple function of nominal window position.

19. A compensating system according to Claim 18 further comprising merit means for calculating a merit value for the simple function and shifting each window only if the merit valve meets a predetermined criterion.

20. A compensating system according to Claim 18 wherein the function means comprises means for calculating a linear function of offsets vs. nominal window position, and means for determining the average offset for each nominal window position from the linear function.

21. A compensating system according to Claim 16 wherein the dispersion means includes a dispersion member and the scanning means includes a stepper motor receptive of the scanning signal and operatively connected to the dispersion member.