DE10238356A1 - Quantitive gas absorption spectrometer for power station combustion chambers uses Fourier transform processing with sampling at less than wavelength modulation frequency. - Google Patents

Abstract

A quantitive gas absorption spectrometer measures the transmission of radiation from a narrow spectral width laser diode light source through a sample with wider absorption line and whose wavelength is sampled at a rate (M1) less than the applied wavelength modulation frequency (M2) and filtered using a Fourier transform and comb filter around M2 with interval M1 with demodulation of transmitted channels.

Description

Territory of invention

The invention relates to the quantitative spectroscopic detection of at least one absorber in a fixed, liquid or gaseous Fabric sample of variable composition, the absorber being spectrally resolved, i.e. H. structured, if possible isolated absorption signatures (e.g. vibration bands or rotation lines) having.

There are numerous procedures and Devices with different light sources, light converters and evaluation methods become known and described in the literature, that enable such quantitative analysis [1–6].

Because of their high spectral resolution Tunable lasers are often used to detect gases. In particular Diode lasers are suitable because they are very narrow-band and fast and are continuously tunable and with inexpensive Semiconductor detectors can be combined as light converters.

In addition to sampling-based, so mentioned extractive analysis, in which a substance sample, for example chemical process taken and under (in terms of pressure, temperature etc.) controlled conditions in an absorption cell is is for technical applications especially the so-called sample-free in-situ spectroscopy (in-situ = on site) of increasing importance [2, 4, 5]. The substance within the process, e.g. B. in Combustion chamber of a power plant, analyzed by absorption spectrometry.

To do this, it is essential to provide all of the evidence Absorber outgoing signal influencing interference from the absorption signal through mathematical or electronic compensation or filtering to remove.

For the spectroscopic determination of the absorber concentration, the measuring light is applied to the absorber and the emission wavelength of the light source is scanned periodically and as linearly as possible over the absorption line to be detected. This scanning process can take place, for example, in the form of a symmetrical or asymmetrical triangular function with the scanning frequency M1. With pure wavelength modulation after the absorption path - due to the target species - an intensity modulation is obtained, the intensity of which is based on the Lambert-Beer law I (λ, t) = I 0 (λ) · exp (–S (T) · ϕ (λ - λ 0 ) · L · L) (1) is linked to the concentration (N) of the absorber and the absorption length (L). The concentration N is usually the size we are looking for. I ₀ (λ) stands for the incident optical intensity.

The absorption line itself is characterized by two parameters: the so-called line thickness S (T) and the line shape function ϕ (λ – λ ₀ ). S represents a measure of the spectrally integrated "area" of the absorption line and is dependent on the temperature (T) via the occupation of the initial level.

ϕ, on the other hand, only describes the line profile centered on the line center λ ₀ and is therefore standardized to one in the area. The maximum value (peak) and the temporal course of the intensity modulation (from now on absorption profile or profile) is determined by the spectral shape (?) Of the absorption signature (from now on absorption line or line) and by the dynamic tuning properties of the laser.

Generally one assumes that the voting process is independent the tuning speed (i.e. the modulation frequency M1) is linear, so that the shape of the absorption profile (in the time domain) with the Shape of the absorption line (in the frequency domain) is identical.

In the case of an in-situ measurement, the optical power P (λ, t) recorded at the detector after passing through the measuring section can be represented with an expanded form of Lambert-Beer law: P (λ, t) = P 0 (Λ) exp (-S (T) · φ (-λ O ) · L · L) Tr (t) + E (t) (2) As in the simple form of Lambert-Beer law, P ₀ (λ) stands for the radiated optical power. Tr (t) describes the transmission of the in-situ measuring section without the absorber to be detected and E (t) the false light detected by the detector - not from the measuring light source.

A key component in quantification of the absorption signal is the evaluation method. Two of the most important Methods are direct absorption spectroscopy, DA, [1, 2] and wavelength modulation spectroscopy, WMS, [3-6].

Direct absorption spectroscopy, THERE

For example, in [2] the DA explained. At the DA the laser is used e.g. B. with the frequency M1 periodically over the Line scanned.

The absorption signal is recorded and possibly low-pass filtered, but without the absorption profile Modification of the spectral content to change. The area of the Absorption line is made according to the Lambert Beer law certainly. The area is a measure of concentration N of the absorber.

Has proven to be particularly advantageous the determination of the area below the absorption line, for example by means of curve fitting (linear, non-linear or recursive, e.g. with the Levenberg-Marquardt algorithm). The from the measured value (area) determined relative mixture proportion (ppmV) of the absorber is at this procedure despite a pressure-dependent broadening of the absorption line independently from pressure. So this method is also in variable pressure systems used.

In addition, the absolute absorber density N (particles per volume) determined directly without the need for calibration, i. H. a Calibration of the sensor response using a premixed concentration standard serving test gases for the empirical determination of possibly even time-variable device constants. Cause regular calibration processes for conventional analyzers considerable costs for Personnel, investments (calibration devices) or consumables (e.g. calibration gases). A calibration free So the process has one for the industry very important cost advantage.

[1, 2] also explains how the resolution of the method can be further increased by noise suppression being achieved by averaging the absorption profiles (so-called scan integration) successively. If you average a high number (n) of absorption profiles, the resolution (smallest detectable absorption) can be reduced by a factor √ in this case of statistical disturbances n be increased.

Especially for in-situ measurements, in which the influence of the absorption signal due to the fluctuating boundary conditions (temperature, pressure, sample composition, convection, etc.) can be expected, it is important to evaluate the interference before evaluating the raw signals correct it efficiently. These disorders arise e.g. B. by broadband, ie spectrally little variable, for. B. dust or ash-induced absorptions (Tr), ie by changes in the basic transmission of the measuring section. Another fault can arise from false light (E), e.g. B. by glowing particles (see explanations in [2] or 7 ).

To effectively correct these disorders or suppress, it is important to adjust the tuning speed. H. the modulation frequency M1 to choose so high that the interference (additive due to false light, multiplicative due to transmission fluctuations) "frozen" within an absorption profile, i. H. can be assumed to be constant can. This can be guaranteed especially with diode lasers, as these can offer the necessary high tuning speeds. typical Time scales for disorders are in the millisecond range. Typical values for the scan frequency M1 are therefore 1 kHz, i.e. H. it only takes about 1 ms to scan the line.

• – ihre Geschwindigkeit,
• – keine Notwendigkeit einer Kalibration,
• – Möglichkeit der Störungskorrektur am Messsignal selbst,
• – einfachster apparativer Aufbau, insbesondere ist kein Lock-in-Verstärker nötig.

The advantages of DA are

• - their speed,
• - no need for calibration,
• - Possibility of correcting interference on the measurement signal itself,
• - The simplest construction, in particular, no lock-in amplifier is necessary.

• – Die DA ist in ihrer Empfindlichkeit begrenzt, da sie alle Störsignale mit dem Detektor aufnimmt und kaum Selektionsmöglichkeiten zur Unterdrückung der Störsignale bietet. Sie besitzt eine deutlich kleinere Auflösung (in optischer Dichte = OD) bzw. Empfindlichkeit als die WMS, da sich das Rohsignal aus einem sehr hohen Offset und einem nur relativ kleinen Nutzsignal zusammensetzt. Dadurch wird auch der mögliche Dynamikbereich des Messverfahrens zu kleinen Absorptionen hin eingeschränkt
• – Die DA erlaubt zwar aufgrund ihrer Tauglichkeit für den Einsatz hoher Modulationsfrequenzen M1 die Korrektur additiver und multiplikativer Störungen, wie sie bei In-situ-Messungen unvermeidlich sind, die dazu erforderlichen hohen Modulationsfrequenzen führen jedoch zu nichtlinearem Abstimmverhalten von Diodenlasern. Dieses nichtlineare Abstimmverhalten bewirkt eine Deformation des Absorptionsprofils im Zeitbereich, somit systematische Fehler in der Auswertung, und daher eine reduzierte Präzision der Auswertung. Weitere Störungen entstehen, da bei nichtlinearem Abstimmverhalten die Fläche unter dem Absorptionsprofil von der zeitlichen Position innerhalb des Scans abhängt und somit bei spektralen Instabilitäten (Driften der Laserwellenlänge) deutliche systematische Fehler zu erwarten sind, was die Präzision des Verfahrens einschränkt oder deutlich höhere Kosten für eine bessere spektrale Stabilisierung verursacht. Dazu wurde zwar in [5] ein Verfahren zur Linearisierung des Abstimmverhaltens beschrieben, dieses ist jedoch sehr aufwendig, erfordert teure Signalgeneratoren mit der Möglichkeit zur frei programmierbaren Modulation und verursacht durch die nichtlineare Stromansteuerung eine bei der Auswertung sehr störende nichtlineare Amplitudenmodulation.

The devices and methods described in [1, 2, 5] based on direct absorption spectroscopy, DA, have the following disadvantages:

• - The sensitivity of the DA is limited because it records all interference signals with the detector and offers hardly any selection options for suppressing the interference signals. It has a significantly smaller resolution (in optical density = OD) or sensitivity than the WMS, since the raw signal is composed of a very high offset and a relatively small useful signal. This also limits the possible dynamic range of the measuring method towards small absorptions
• - Although the DA allows the correction of additive and multiplicative disturbances due to its suitability for the use of high modulation frequencies M1, as they are unavoidable with in-situ measurements, the high modulation frequencies required for this lead to non-linear tuning behavior of diode lasers. This nonlinear tuning behavior causes a deformation of the absorption profile in the time domain, thus systematic errors in the evaluation, and therefore a reduced precision of the evaluation. Further disturbances arise because with nonlinear tuning behavior the area under the absorption profile depends on the time position within the scan and thus in the case of spectral instabilities (drifting of the laser wavelength) significant systematic errors are to be expected, which limits the precision of the method or significantly higher costs for a causes better spectral stabilization. A method for linearizing the tuning behavior was described in [5], but it is very complex, requires expensive signal generators with the possibility of freely programmable modulation and, due to the non-linear current control, causes a non-linear amplitude modulation which is very disruptive during the evaluation.

Wavelength modulation spectroscopy, WMS

Another sensitive process The WMS represents recording of small absorptions Laser as in the case of the DA z. B. with a triangular function with the Frequency M1 periodically over an absorption line scanned. At the same time, it becomes clear with one faster sine modulation with frequency M2 with much smaller Wellenlängenhub acted as M1. Typical values for M2 are 100 kHz.

On the flank of an absorption line receives then a transmission signal modulated with M2. The laser wavelength is on the area around the maximum of the absorption matched, so you get a Signal at twice the frequency of M2 (so-called 2f signal), because a wavelength modulation on both sides of the maximum to an increase in transmission leads. This shows the 2nd with very small modulation amplitudes. Derivation of the absorption line.

Without a narrow-band absorber thanks to the linearity of the current-light characteristic, plunge into the detector signal in the beam path of the laser only odd harmonics of the scan frequency M1 and the Root note of the fast sine modulation M2. Through an absorption line however, the measuring section has a non-linear transfer function, so that also higher ones Harmonics of M2 can be found. This is preferred in the WMS the second harmonic (2f) - or fourth or sixth harmonic - from M2 as a measure of the through the target species-induced absorption is examined and recorded. For the Limit cases of small modulation depths take the respective harmonic Signals of order n the form of the respective nth derivative of the Absorption profile. I.e. the 2f signal has the shape of the second Derivation of the absorption profile according to time (or according to the wavelength in the case a linear tuning process) [7].

The amplitude of the 2f signal component or the second derivative from the transmission signal z. B. through Fourier transformation or use of a lock-in amplifier be determined. It is then possible that Amplitude of the 2nd derivative of the absorption profile as a function of center frequency or center wavelength scanned with M1.

Usually you get a function, like the one in 3 in the right column in the middle. The difference between minimum and maximum is proportional to the absorber concentration. The proportionality constant must be determined by calibration. As a rule, it is a function of the line shape, the pressure, the temperature and the collision partner, i.e. the composition of the sample.

In [3, 4] it is described how with Using an analog lock-in amplifier as Evaluation circuit the higher harmonics generated by the absorption line Frequency components in the photocurrent can be detected selectively.

In the case of the WMS, since only after periodic signals with frequencies M1, 2 * M1, 3 + M1, etc. searched is evaluated, the detector current is AC coupled, in the Usually with the help of a narrow-band lock-in amplifier for oppression of all DC components. The dynamic range of the electronics can thus be used much better be what along with the narrowband detection characteristic the lock-in amplifier a significantly higher one resolution than in the DA down to below 10 ^ –6 OD (i.e. 1 ppm light attenuation).

To further increase the resolution and The detection bandwidth can be reduced at the lock-in amplifier Integration time can be increased. This extends However, the measurement time is typically 100–1000 ms. Then it is no longer possible to completely record a spectrum within 1 ms so as not to go through the fluctuations in time get influenced results.

In [6] the application of the WMS in the case of an in-situ application described. One advantage of the WMS is that additive disorders (e.g. false light) outside the modulation or detection frequency and their harmonics very much well suppressed become.

However, this does not apply to multiplicative Effects such as transmission disorders. These also appear in the 2f signal. Have multiplicative disorders the effect that the idealized Fourier spectrum of the signal with the Fourier spectrum of the perturbation is folded. Falsify it disorders the 2f signal. As a rule, the ideal Fourier spectrum of the 2f signal due to interference widened. Usually it takes correction to correct.

As described in [4], this problem could be solved by capturing an independent signal, which is only influenced by the transmission disturbance, and subsequently using it to standardize the 2f-WMS signal. This correction signal is the offset of the 1f absorption profile, which is caused by the intensity modulation of the laser light that is unavoidable in diode lasers during the current-induced tuning process. This offset is proportional to the steepness of the current-light characteristic of the laser and is only influenced by the transmission of the measuring distance, so it can be used as a transmission sensor. To do this As described in [4], the raw 1f signal is separated from the resonant signal component caused by the target absorber, which is done in [4] by adapting the curve to the resonant profile. The time-dependent offset within the scan can then be used for a local transmission correction within the 2f signal.

• – nur eine eingeschränkte Unterdrückung multiplikativer Störungen erlauben;
• – nur ein langsames Scannen erlauben, d. h. relativ kleine Modulationsfrequenzen M1. Da aber schnelles Scannen eine Voraussetzung für die Korrektur von Störungen ist, wie sie in in-situ-Anwendungen auftreten, gibt es beträchtliche Schwierigkeiten und Komplikationen beim Einsatz der WMS für in-situ-Anwendungen unter realen Bedingungen. Das in [4] beschriebene Verfahren der Korrektur von Störungen funktioniert z. B. nur, solange die Transmissionsstörungen den resonanten Anteil des 1f-Signales nicht maskieren. Dies schränkt den Einsatz des Verfahrens aus [4] auf Fälle mit relativ schwachen Transmissionsstörungen ein.
• – eine hohe Abhängigkeit des Nutzsignals von äußeren Bedingungen (Druck, Temperatur, chemische Zusammensetzung) zeigen und somit unbedingt
• – eine Kalibration des Messsignals mit Prüfgasen oder anderen Normalen erfordern.
• – auf analog-elektronischen Lock-in-Verstärkern basieren, die driftanfällig sind, oder
• – mit kommerziellen digitalen Lock-in-Verstärkern ausgeführt werden, die teuer sind, meist keine Upgrades der Auswertung erlauben und somit unflexibel sind und in der Art der Mittelung nicht konfigurierbar sind.
• – Eine gemultiplexte Messung mehrerer Spezies ist wie in [6] dargestellt über mehrere Modulationsfrequenzen möglich, dazu sind jedoch mehrere Signalgeneratoren und auch mehrere Lock in-Verstärker erforderlich, was zu beträchtlichen Kosten führt.
• – Eine quasi-simultane, gemultiplexte Mehrspeziesmessung gemäß dem Zeitmultiplexing-Prinzip ist nicht mit WMS kombinierbar, da mit Lock-in-Verstärkern aufgrund der langsamen Scanfrequenzen nicht rasch genug zwischen den verschiedenen Spezies oder Methoden hin und her geschaltet werden kann.

The devices and methods described in [3, 4, 6] based on wavelength modulation spectroscopy (WMS) have disadvantages because they:

• - allow only limited suppression of multiplicative interference;
• - allow only slow scanning, ie relatively low modulation frequencies M1. However, since fast scanning is a prerequisite for correcting disturbances that occur in in-situ applications, there are considerable difficulties and complications when using the WMS for in-situ applications under real conditions. The method of correcting disturbances described in [4] works e.g. B. only as long as the transmission interference does not mask the resonant portion of the 1f signal. This limits the use of the method from [4] to cases with relatively weak transmission disorders.
• - show a high dependency of the useful signal on external conditions (pressure, temperature, chemical composition) and therefore absolutely
• - Require calibration of the measurement signal with test gases or other standards.
• - based on analog-electronic lock-in amplifiers that are prone to drift, or
• - Are carried out with commercial digital lock-in amplifiers, which are expensive, usually do not allow upgrades to the evaluation and are therefore inflexible and cannot be configured in the way of averaging.
• - A multiplexed measurement of several species is possible as shown in [6] over several modulation frequencies, but this requires several signal generators and also several lock-in amplifiers, which leads to considerable costs.
• - A quasi-simultaneous, multiplexed multi-species measurement according to the time-division multiplexing principle cannot be combined with WMS, since lock-in amplifiers cannot switch back and forth quickly enough between the different species or methods due to the slow scan frequencies.

The object of the invention is a highly sensitive quantitative determination of an absorber in one Gas or liquid sample enable under in situ conditions.

solution

This task is accomplished through the inventions with the characteristics of independent Expectations solved. advantageous Further developments of the inventions are characterized in the subclaims.

The task is solved by that contrary to the usual Derivative spectroscopy methods have a higher repetition frequency Modulation M1 (scan) is used, typically at 1 kHz, such as in the DA. This makes the spectrum of the signal to be detected fanned.

The detection of the derivative signals according to known methods with phase-sensitive amplifiers, so-called Lock-in amplifiers in such a case it is only possible that their detection bandwidths be enlarged, resulting in losses in signal quality leads.

Another method according to the invention to acquire the signals and to limit the effective detection bandwidth proposed. The invention takes advantage of the fact that the spectrum of the useful signal (transmission signal) is discrete in the case of strictly periodic repetition of the modulations. There are bands around M2 and z * M2 with sidebands at +/- M1, +/- 2 * M1, +/- 3 * M1, etc., where z is an integer. The spectrum around M2 or z * M2 is spread by fast scanning with M1.

The detection bandwidth can therefore by using a suitable combination of belt and comb filters be reduced. The bandwidth of each band of the comb filter can be chosen very narrow typically, according to the bandwidth of a lock-in amplifier, such as he for WMS is used, namely 1 Hz.

After filtering, the extraction takes place the derivative signals by demodulation, i.e. that is, when implemented in analog electronics, the signal of the respective band is included the signal of a local oscillator with the frequency of the central frequency the gang (M2 or z * M2) mixed (multiplied), resulting in results in a frequency component around 0 Hz in the mixed signal. Which band resulting at 0 Hz is rectified and filtered, i. H. the amount is formed and the signal is averaged.

Disorders, e.g. B. reduced transmission, remain as multipliers of the signals or their amplitudes, and although there are disturbances due to the convolution of all ideal spectra with the spectrum of the Interference by the narrow-band comb filtering the same multiplier for the 1f (M2) - and 2f (2 * M2) signal and for all sidebands.

The transmission signal always has a 1f signal, since the diode laser determines the strength of the Be drive current is coordinated. At the same time, however, the laser power also changes. But this is a modulation with 1f or M1. The 1f signal thus has a constant offset.

The 1f signal has three in total determining components: the line of the molecule, the offset through the modulation the laser power (which is known) and the interference, which is mathematically corrected can be. The interference can be eliminated the 1f signal can be eliminated. This can cause the interference proportionality be determined and applied to the evaluation of the 2f signal become. The multiplicative disorders can be eliminated. I.e. The 1f signal can be used to calibrate the 2f signal be used. So that can the 2f signals from Interference free become.

After that, from the 1f or the 2f signal concluded in a known manner on the absorber concentration become.

By changing the known modulation and detection schemes for the acquisition of derivative signals, the use of high frequencies for the Modulation M1 with the detection bandwidth remaining unchanged at the same time and a transmission correction via the 1f signal can both additive and multiplicative interference signals are effectively suppressed.

This is the method according to the invention sensitive and in-situ compatible.

With the Fourier transformation of the a 1f or zf spectrum results from the origin of the shifted spectrum with real and imaginary part, corresponding to two lock-in amplifier signals. It is advisable to choose the phase of the Fourier transformation in such a way that the real part becomes maximum.

A comb filter can e.g. B. digitally by phase-rigid averaging of several (scan) periods of the detected signal realize. To do this, first of all the fast modulation frequency (M2) is a multiple of the slow one Modulation frequency (M1) so that a whole number of fast Modulations fit into a period of slow modulation and there is an in-phase averaging. Then the signal into individual sections according to a period of slow Modulation M1 disassembled. About these Sections are averaged. When averaging, only the frequencies are left that are an integer multiple of M1. All others average Out. This results in comb filtering.

Is not averaged in phase, are the comb filter frequencies not offset by a multiple of M1 next to M2 or z * M2.

Furthermore, the scheme of comb filtering can be varied. It is the use of interlocking comb filters possible, whereby with a suitable choice of the modulation frequencies M1 and M2 also overlapping harmonic channels can be separated and so higher Frequencies for M1 can be used without an increase the frequency M2 becomes necessary.

If the modulation frequency M2 z. B. chosen so that they are an integer plus a half-fold multiple of the frequency M1 is (M2 = (n + 0.5) * M1, n = 1, 2, 3,. , .), the discrete contributions of neighboring harmonic channels entangled in the spectrum on and can through parallel, alternating and non-alternating averaging be separated.

There is often a need for several chemical ones To detect species at the same time. A method is used in [5] described on the basis of the WMS, in which two different lasers - one for each a species - with different modulation frequencies M2 are operated. The Laser beams are overlaid by the absorption zone and with only one detector for both wavelength demonstrated.

The respective absorption signals can nevertheless separated based on their different modulation frequencies can be evaluated by synchronizing them with two suitably Lock-in amplifiers detected and be separated. The species are thus simultaneously in the same Volume detected. This entanglement and separation of several Signals are also called multiplexing in English. In the above The case is modulation frequency multiplexing.

The present invention provides the possibility of doing something similar to be achieved by not averaging in phase. In one in such a case, the comb filter frequencies are not a multiple offset from M1 next to M2 or z * M2.

This opens up a way the combs deliberately postpone by averaging one phase (Exp (i * φ)) is inserted multiplicatively between adjacent periods. Corresponding to a phase φ 180 ° z. B. leads to a displacement of the combs by 1/2 M1, e.g. B. 500 Hz.

Are used to prove different Absorber different lasers with different modulation frequencies used so can these are separated from one another in the evaluation by deliberately moving the combs become. This allows also lasers (per species) can be differentiated in the modulation frequency M2 are only 500 Hz apart (at M2 = 100 kHz), i.e. 100 and 100.5 kHz.

This allows multiple absorbers to be combined with one Detector can be detected at the same time.

Usually the 1f or 2f bands are themselves about half as wide as the frequency f or M2, i.e. z. B. 50 kHz. With the method according to the invention is it also possible to separate many species from one another by means of targeted frequency bands.

Instead of shifting the comb filter, the spectrum of the signal can also be shifted by the modulation already during generation with successive modulation periods M1 (scan) M2 is successively shifted by a certain phase.

A similar to modulation frequency multiplexing The method is time division multiplexing, in which the radiation sources alternately switched on one after the other, scanned over the absorption line and be switched off again. Because of the quick modulation of diode lasers can do this switching in kHz clock and faster take place, so that one speaks of a quasi-simultaneous measurement.

In combination with moving the combs of the comb filter can this way a large number can be detected by different absorbers.

Because the dynamics of the 1f and 2f signals is very different, it is advantageous if the transmission signal with the help of at least one separate bandpass filter around the frequency M2 and / or z * M2 is filtered.

The calibration freedom of the direct Absorption spectroscopy is used in that either the signals obtained are equivalent to direct absorption spectroscopy Signal is extracted or the inventive method by time multiplexing is connected with direct absorption spectroscopy. It will then from interval to interval the additional modulation with the Frequency M2 switched on or off, in rapid change, e.g. B. in every other interval.

Every second interval then offers the possibility, the signals in the usual Way of evaluating the DA (see, for example, [1]). The other intervals are evaluated according to the invention, the results of the DA evaluation can be used, e.g. B. for calibration.

The freedom from calibration is then obtained the DA and the accuracy of the WMS. This is the method according to the invention sensitive and suitable in-situ.

If a diode laser, as for the execution of the inventive method necessary, scanned quickly, hangs the wavelength of the laser is non-linear of the operating current and thus of the time from. This depends on it together that the light output of the diode laser is the operating current follows directly, the wavelength but not. Would like to one for the measuring principle is a particularly advantageous temporal triangular function for reaching the wavelength there are delays at the turning point.

These difficulties can be overcome avoid when the operating current of the diode laser at the reversal point is changed in the form of a step function such that the diode laser its wavelength changes as linearly as possible with time, starting with the time of the jump. By the current jump, the power of the diode laser remains linear at the time.

The invention offers the apparatus The advantage that lock-in amplifiers are completely dispensed with can be. You can to be replaced by means for sampling the transmission, e.g. B. through an oscilloscope with an integrated analog-to-digital converter. The sampling rate is thereby at least twice M2. Typically with 10 * M2 worked. This results in a typical sampling rate of 10 ^ 6 samples per second.

The invention is described below of embodiments explained in more detail the are shown schematically in the figures. Same reference numbers in the individual figures denote the same elements. In detail shows:

1 a schematic representation of a structure for performing the method;

2 a schematic spectrum of the detected signal;

3 schematically a possible sequence of electronic data processing when performing the method;

4 an example signal;

5 an example signal;

6 an example signal;

7 a transmission signal in direct absorption spectroscopy;

8th a representation of the dynamic tuning behavior of a current-tuned diode laser;

9 the modulation scheme for linearizing the wavelength tuning of a diode laser; and

10 the influence of the step height of the operating current on the linearity of the tuning for a given laser.

A possible structure for carrying out the method is in 1 shown. Suitable electronics ( 1 ) generates z. B. sawtooth or triangular modulation function M1 (scan), the frequency of which is typically in the kHz range, and a z. B. sinusoidal modulation function M2 (can be switched off), the frequency of which is typically around 100 kHz. The sum of the two phase-locked functions is used to control the tunable light source ( 2 ), e.g. B. a diode laser.

The wavelength-modulated light strikes a detector after passing through the sample ( 3 ) whose output signal is amplified ( 4 ) and filtered ( 5 ) becomes. The spectral components of the Suppressed signals that do not contribute to the actual useful signal. Both low-frequency signal components, which essentially result from the detector being illuminated by non-absorbed light, and high-frequency signal components, which lead to errors in the subsequent digitization (anti-aliasing), are suppressed here.

An optional combination of a phase locked local oscillator ( 6 ), Mixer ( 7 ), Amplifier ( 8th ) and filter ( 9 ) can be used to shift the signal to lower frequencies, which means that possibly cheaper elements can be used for the following stages and / or higher modulation frequencies.

An analog / digital converter ( 10 ) is used to record the signal. A phase-rigid averaging parallel to modulation M1 (scan) serves for comb filtering of the signal. With the help of electronic data processing, individual harmonic channels of the signal are digitally isolated and demodulated.

Used to explain the mode of operation 2 , in which the schematic spectrum of the detected signal in the derivative spectrocopy (WMS) is shown. It is composed of the harmonic multiples of the fast modulation frequency M2 and the information-bearing sidebands through the periodic, slower tuning of the light source with the frequency M1. The disturbed spectrum results from the convolution with the spectrum of the multiplicative disturbances.

The interference-free detector signal contains the high-frequency modulation frequency M2 and in the presence of one nonlinearity on the light path (e.g. absorption line) whose multiples n · M2 with decreasing amplitudes.

If the light source is additionally on the Modulation M1 (scan) tuned, the different frequencies by change of non-linearity amplitude modulated. In the spectrum, therefore, join every harmonic Frequency n · M2 Modulation sidebands on, their widths from the tuning speed and their shapes from the profile of non-linearity depend. With periodically repeated tuning, these modulation sidebands expose themselves discrete frequencies at a distance from the slow modulation frequency M1 together.

Will the signal in a known manner detected with the aid of a phase-sensitive amplifier (lock-in amplifier) adjust its bandwidth so that the modulation sidebands are completely within of this channel can be recorded. To a cheap one Signal-to-noise ratio too have to achieve this detection bandwidth and correspondingly the frequency of the slow ones Modulation M1 selected to be very small become.

This is the crucial disadvantage of derivative spectroscopy. While the non-multiplicative interference by choosing high modulation frequencies for the fast modulation M2 suppressed can be results for the disturbances multiplicative no advantage. They go completely independent of the fast modulation frequency M2 in the signal because the disturbed spectrum arises from the convolution of the undisturbed Spectrum with the spectrum of multiplicative interference results. For each discrete frequency that occurs, an additional, continuous fault sideband on. There is an overlapping Sideband system, that completely covers the useful signal. The interference affects the bandwidth of the amplifier directly through on the signal and can cannot be distinguished from the useful signal.

In contrast, the modulation frequency M1 (scan) is selected large the sideband spectrum fanned out. The frequency M1 can be chosen to be large enough that adjacent fault sidebands no longer overlap. This eliminates the noise at all discrete frequencies that the undisturbed Allocate spectrum, almost identical. Comb filtering of the signal hardly any high-frequency noise components occur within a "comb prong" from adjacent interference sidebands. With correspondingly long averaging times, only the very influence low frequency interference the signal and act over approximately a voting period like a multiplicative constant.

In 3 the sequence of electronic data processing is shown with an example signal. The band and comb filtered signal is shown in the top left. The comb structure no longer shows the discretized representation of the spectrum, since each frequency point represents a channel of the comb. The individual harmonic channels are isolated and the respective spectrum is shifted by the channel center frequency and demodulated by reverse transformation. The direct component of the harmonic channel n = 1 (dashed) indicates the transmission averaged over the measurement period, with which the signals can be corrected.

Phase sensitive demodulation For example, the signal can be performed using the following procedure. A discrete Fourier transform (DFT) provides the spectral representation of the averaged signal. In this illustration, due to the Digitization only represents discrete frequencies belonging to the comb filter, so that the comb structure doesn't appear obvious.

The complete amplitude and phase information of the detected signal can with the help of cosine and sine transformations. It's mathematically equivalent perform a complex valued transformation, taking the amplitudes the occurring negative frequencies are rejected or set to zero become.

A single, modulated harmonic channel, ie a frequency centered around a frequency n · M2 window of limited width is digitally isolated. The demodulation itself is carried out by inverse transformation into the time domain after the spectrum has been shifted in such a way that the central frequency of the selected harmonic channel becomes zero.

The shift has the same Effect like dividing the signal in the time domain by a complex value harmonic function whose frequency is determined by the central frequency of the Channel is given.

The result of the mathematical procedure provides the temporal development of the modulated channel by amount and phase. The extracted signal ideally differs only by a constant factor from the undisturbed signal that the middle Transmission during of the averaging period.

The further evaluation of the signal is shown below.

Hierbei bezeichnet g"(ν) die Übertragungsfunktion und ihre Ableitungen n-ter Ordnung, ν ist der Mittelwert des Arguments und A und ω beschreiben die Modulationsamplitude und -frequenz der Modulation M2.The properties of the irradiated medium can be represented by a transmission function that is dependent on the wavelength or the number of waves. If the wavelength or wavenumber is modulated sinusoidally, the result can be written according to the binomial formulas in the following, known manner:

Here g "(ν) denotes the transfer function and its derivatives of the nth order, ν is the mean value of the argument and A and ω describe the modulation amplitude and frequency of the modulation M2.

Im Falle kleiner Modulationsamplituden, sind die Funktionen H_n proportional zur n-ten Ableitung der Übertragungsfunktion.The summation can be sorted according to the harmonic terms that occur:

In the case of small modulation amplitudes, the functions H _{n are} proportional to the nth derivative of the transfer function.

Hier stellt I(t) das Modulationssignal dar, im speziellen Beispiel den Betriebsstrom. Î ist die Amplitude der Modulation M2, k_ν ist der dynamische Abstimmkoeffizient der Wellenlänge oder Wellenzahl ν der Lichtquelle und φ beschreibt die Phasenverschiebung der Wellenlängenabstimmung gegenüber der Modulation M2.Neglecting additive disturbances, the time-dependent, detected signal is the product of the emitted light output and the transfer function of the irradiated medium. The latter can be broken down into two factors for wavelength-dependent and multiplicative interference properties. If the light source is modulated, for example, via the operating current, the following equations result:

Here I (t) represents the modulation signal, in the specific example the operating current. Î is the amplitude of the modulation M2, k _ν is the dynamic tuning coefficient of the wavelength or wavenumber ν of the light source and φ describes the phase shift of the wavelength tuning compared to the modulation M2.

Diese Gleichung lässt sich wiederum nach harmonischen Termen sortieren:

Die verschiedenen Summanden beschreiben die verschiedenen harmonischen Kanäle. Bei genügend kleinen Modulationsamplituden für die Modulation M2 ist die letzte Summe vernachlässigbar. Nach Anwendung des erfindungsgemäßen Verfahrens lässt sich der Term für die multiplikativen Störungen Tr(t) durch eine Konstante Tr ersetzen. Ist die Phasenverschiebung φ bekannt, oder wird sie aus den Signalen ermittelt, können die übrigen Summanden durch Auswertung der gewonnenen Signale zugeordnet und zur Bestimmung der gesuchten Absorption benutzt werden. Hierbei ist insbesondere zu bemerken, dass bei vorhandener Modulation der Ausgangsleistung der Lichtquelle der Faktor Tr aus den Ergebnissen der harmonischen Kanäle n=1 und n=2 eindeutig bestimmbar ist. Des Weiteren enthalten die Kanäle n=0 und n=1 bei kleinen Amplituden der Modulation M2 jeweils ein zur direkten Absorptionsspektroskopie äquivalentes Signal, da in dem Fall der H0-Term die Übertragungsfunktion g widerspiegelt.It is assumed for simplification that a possible modulation of the output power is the mod lation signal follows instantaneously and a linear dependency with the slope kp applies for the completely noise-free signal:

This equation can be sorted according to harmonic terms:

The different summands describe the different harmonic channels. If the modulation amplitudes are sufficiently small for the modulation M2, the last sum is negligible. After applying the method according to the invention, the term for the multiplicative disturbances Tr (t) can be replaced by a constant Tr. If the phase shift φ is known, or if it is determined from the signals, the remaining summands can be assigned by evaluating the signals obtained and used to determine the absorption sought. It should be noted here in particular that if the output power of the light source is modulated, the factor Tr can be clearly determined from the results of the harmonic channels n = 1 and n = 2. Furthermore, the channels n = 0 and n = 1 each contain a signal equivalent to direct absorption spectroscopy at small amplitudes of the modulation M2, since in the case the H0 term reflects the transfer function g.

Die Winkel α₁ = 6.2° und β₁ = –26.9° sind für das in den 5 und 6 gezeigte Beispiel durch die absoluten Phasen der Amplituden- bzw. Frequenzmodulation bestimmt. Die Phasenverschiebung φ ergibt sich aus der Differenz α₁–β₁.In order to be able to correctly separate the signal components, the following transformation is applied to the real and imaginary parts:

The angles α ₁ = 6.2 ° and β ₁ = –26.9 ° are for that in the 5 and 6 shown example determined by the absolute phases of the amplitude or frequency modulation. The phase shift φ results from the difference α ₁ –β ₁ .

The phases of the two main components of the 2ω signal - another name for the 2f signal - are identified by calculating the difference between the signal at the center of the line (maximum value) and the mean value. The result gives the angle α ₂ = 156.7 °. The angle β ₂ is given by α ₂ –φ = 123.6 °. In principle, α _{2 should be} equal to β ₁ . For the 2ω signal, however, the overall phase shift caused by the electronics deviates from that of the 1ω signal. The mean of the signal is not zero because there is a slight ubiquitous non-linearity in the output power modulation of the diode laser.

The 4 . 5 and 6 show example signals obtained with a diode laser as the light source in the presence of a single absorption line.

The calibration freedom of the direct Absorption spectroscopy is guaranteed using the following method. Both the degree of transmission and the proportion of the background light can be from the height of a modulation period if both over this time interval as can be viewed constantly and the form of the modulation in the absence of specific absorption is known. The simplest procedure is maximum and minimum of the detected signal during a modulation period and by scaling with the Extremely the laser modulation function the detectable light output to determine at the laser threshold. The calculated value gives the Background performance again.

The time dependency of the laser operating current is described by the mean current I0, the modulation amplitude ΔI and any amplitude-normalized modulation function f (t): I (t) = I 0 + ΔI · f (t) (15) Above the laser threshold Ith, the following applies to the laser output power Pout with a linear characteristic and the transmission factor ε: P out (t) = ε (I 0 - I th + ΔI · f (t)) (16) The transfer factor ε is also called curve steepness. It indicates how much the laser power changes when the operating current changes. In other words, ε is the derivative of the laser power according to the current power.

With the transmittance Tr assumed to be constant and the background (false light) E also constant, the following applies to the detected power: P (t) = Tr · ε (I 0 - I th + ΔI · f (t)) + E = P 0 + ∆Pf (t) (17) P0 and ΔP are the measurable quantities that e.g. B. can be determined from the extremes of the detected power. It follows: Tr · ε = P I and E = P 0 - P I (I 0 - I th ) (18) I.e. the amplitude modulation of the laser is used as a sensor for the transmission.

The background portion can be derive from a simple proportionality. The proportionality factor (hereinafter modulation parameter) can be easily determined. This method leaves can be used without any problems if the performance characteristic of the laser and the transfer function of the optical elements are linear. The extremes can then be used for evaluation if they are unaffected by variable, specific absorptions by the sample.

If, due to the high modulation frequency M1, the signals are distorted due to the non-linear tuning behavior of the light source, and if this can be traced back to a relaxation process, the distortions can be avoided by inserting a phase-appropriate jump at the reversal points of the modulation function M1 is, ie the addition of a further rectangular modulation M3. In this way, piecewise linear tuning can be achieved, with any modulation of the output power also remaining piecewise linear. 9 illustrates the modulation scheme for linearization of wavelength tuning.

wobei App die Spitze-zu-Spitze-Amplitude (Peak-Peak) der rein dreiecksförmigen Modulation bezeichnet. Die Relaxationszeit ist mit τ bezeichnet, und T_½ gibt die halbe Modulationsperiode wieder.The height of the jump ΔI is to be selected as follows:

where App denotes the peak-to-peak amplitude (peak-peak) of the purely triangular modulation. The relaxation time is denoted by τ and T _½ represents half the modulation period.

8th shows the dynamic tuning behavior of a current-tuned diode laser, determined using a 10 cm air etalon. The modulation depth was Ipp = 17.2 mA. The mirror distance of the etalon was 10 cm. The spectral distance between successive interference maxima (free spectral range FSR) is therefore 0.05 cm ^ -1 or 1.5 GHz. As a solid line is in 9 the static tuning coefficient determined with a wavemeter. In the lower part of the 9 the dynamic tuning rate of the above-mentioned laser is plotted as a function of the modulation frequency.

10 shows the influence of the jump height on the linearity of the tuning. The Laser Sharp LT 016 0-14 was used as in 8th at 5 kHz modulation frequency.

There are numerous within the scope of the invention Modifications and developments of the described exemplary embodiments realizable.

So parts of digital filtering through analog comb or Band filter to be replaced. The demodulation of the signals can be analog respectively. To improve the signal-to-noise ratio can the signal paths are often used in parallel and each path assigned to one or more harmonic channels.

The advantages of the procedure exist in that firstly the known advantages of derivative spectroscopy can be used. Which includes the high sensitivity through oppression additive interference signal components (e.g. noise from the light source, detector noise, electronic Noise, background light and electromagnetic interference) and selective amplification the actual useful signal (strength absorption) the inevitable accompanying signal from non-absorbed radiation from the light source used. This is effective through a low Detection bandwidth reached, which in the range of high frequencies lies. Second, the decisive advantage of direct absorption spectroscopy be used, which consists in the fact that by using high Modulation frequencies M1 (scan) multiply the effective bandwidth incoming disturbances can be drastically reduced and then the detected signals from the undisturbed signals distinguish only by a multiplicative, almost constant factor. This Factor can also determined by evaluating the first harmonic channel (1f channel) and used to correct other channels.

Third, the procedure can be followed Time division multiplexing with the method of direct absorption spectroscopy connect: Will successive periods of modulation M1 (Scan) alternately with and without additional, high-frequency modulation M2 generated, and the detected signal cycles are also alternating two carried out separately Averages fed, so can the signals are evaluated in parallel. It becomes quasi-simultaneous Operation of both methods achieved. That way, not only the high sensitivity of derivative spectroscopy, but at the same time freedom from calibration direct absorption spectroscopy.

List of literature cited

• [1] Reid, J., D.T. Cassidy: "High-Sensitivity Detection of Trace Gases Using Sweep Integration and Tunable Diode Lasers "; Appl. Opt. 21: 2527-2530. (1982).
• [2] V. Ebert, T. Fernholz, C. Giesemann, H. Teichert: "Diode laser-based sampling-free multi-component gas analysis in a 1000MWth gas power plant "; technical measurement, 2001, volume 68, issue 9, 406-414.
• [3] M. Kroll, J.A. McClintock, O. Ollinger; Appl. Phys. Lett. 51, 1465 (1987).
• [4] V. Ebert, K.-U. Pleban, J. Wolfrum: "In-situ oxygen monitoring using near-infrared diode Lasers and Wavelength Modulation Spectroscopy "; in Laser Applications to Chemical and Environmental Analysis, Technical Digest (Optical Society of America, Washington DC), 206-209 (1998).
• [5] V. Ebert, J. Fitzer, I. Gerstenberg, K.-U. Pleban, H. Pitz, J. Wolf rum, M. Jochem, J. Martin: "0n-line monitoring of water vapor with a fiber coupled NIR-diode laser spectrometer "; VDI reports 1366, 5th Intern. Symposium on Gas Analysis by Tunable Diode Lasers, 145-154 (1998).
• [6] D. B. Oh, M. E. Paige, D. S. Bomse: "Frequency Modulation Multiplexing for Simultaneous detection of multiple gases by use of wavelength modulation Spectroscopy with Diode Lasers "; Appl. Opt. 37, 2499 (1998).
• [7] J. Reid, D. Labrie: "Second Harmonic Detection with Tunable Diode Lasers - Comparison of Experiment and Theory "; Appl. Phys. B 26, 203-210 (1981).

Claims (13)

Method for quantitative spectroscopic determination at least one absorber in a solid, liquid or gaseous sample, the absorber having a spectrally resolved absorption line, with the following steps: a) the transmission of a radiation source is determined by the sample, the radiation source being a lesser has spectral bandwidth as the absorption line of those to be examined Sample; b) the wavelength the radiation source is repeated at a repetition rate M1 Absorption line of the absorber tuned; c) in addition the wavelength the radiation source with a frequency M2 which is greater as the frequency M1, modulated; d) the transmission signal with the help of at least one comb filter around the frequency M2 and / or z * M2 - or accordingly shifted frequencies - filtered, using the comb filter only frequency components of M2 or z * M2 - or shifted accordingly Frequencies - +/- a plurality of multiples of M1, where z is an integer is; and e) at least one channel of the transmission signal demodulated. A method according to claim 1, characterized in that the demodulation in step e) with the following sub-steps that can be implemented in analog electronics are carried out: e1) the transmission signal is filtered by means of at least one bandpass filter around the frequency M2 and / or z * M2; e2) the bandpass-filtered signal in this way is mixed with a local oscillator; and e3) the signal obtained is rectified and low-pass filtered. A method according to claim 1, characterized in that the Demodulation in step e) is carried out in the following substeps: e1) the comb filtered signal is Fourier transformed; e2) that resulting spectrum around z * M2 is selected, where z is an integer; e3) the selected spectrum is increased by –z * M2 shifted the frequency origin; and e4) the shifted spectrum is transformed inversely Fourier. Method according to the preceding claim, characterized in that that the phase of the Fourier transformation is selected in step h) in such a way that there is a maximum real part of the transformed signal. Method according to one of the preceding claims, characterized in that M2 is an integer multiple of M1 is; and that in step d) the comb filter is phase locked offset averaging is realized. Method according to one of the preceding claims, characterized in that M2 is not an integer multiple of M1 is; and that in the averaging a specific phase (exp (i * φ)) between adjacent periods are added multiplicatively. Method according to one of the preceding claims, characterized characterized that the transmission of various radiation sources through the sample to detect different absorbers in time is determined alternately. Method according to one of the preceding claims, characterized characterized in that the transmission signal using at least a bandpass filter around the frequency M2 and / or z * M2 filtered becomes. Method according to one of the preceding claims, characterized characterized that at predetermined intervals of voting the radiation source over the absorption line of the absorber the additional modulation according to step b) is omitted. Procedure for wavelength tuning a diode laser, characterized in that the operating current the diode laser is changed in the form of a step function, that the diode laser's wavelength coincides with the time of the jump starting as possible changes linearly with time. At least a device for quantitative spectroscopic determination an absorber in a solid, liquid or gaseous sample, wherein the absorber has a spectrally resolved absorption line which contains the following components:  a) a radiation source for exposure of the sample to radiation in the area of the absorption lines of the absorber, the radiation source has a smaller spectral bandwidth than the absorption line the sample to be examined; b) a radiation receiver for Detection of the transmitted radiation; c) Modulation means the wavelength the radiation source with a repetition rate M1 for periodic tuning the radiation source over the absorption line of the absorber; d) means for modulating the wavelength the radiation source with a frequency M2 which is greater as the frequency M1; e) at least one comb filter, where the comb filter only frequency components of M2 or z * M2 - or equivalent shifted frequencies - +/- a plurality of multiples of M1, where z is an integer is; and f) means for demodulating at least one channel of the transmission signal. Method for determining the transmission attenuation (Tr) of a beam as it passes through a sample, the power of the beam being able to be controlled by an operating current (I), comprising the following steps: a) the amplitude of the power of the beam is modulated with a predetermined modulation depth (ΔI) of the operating current (I); b) the transmitted power of the beam is determined; c) the depth (ΔP) of the modulation of the amplitude of the transmitted power of the beam is determined; d) the transmission attenuation is determined using the equation Tr · ε = ΔP / ΔI, where ε is the slope of the characteristic of the beam source. Method for determining the strength of the signal (E) detected by a detector of an absorption determination device, which signal is not caused by the absorption or transmission of a beam, wherein the power of the beam can be controlled by an operating current (I), with the following steps: a ) the amplitude of the power of the beam is modulated with a predetermined modulation depth (ΔI) of the operating current (I); b) the transmitted power of the beam is determined; c) the depth (ΔP) of the modulation of the amplitude of the transmitted power of the beam is determined; d) the mean transmitted power (P ₀ ) is determined; e) the threshold current (I _th ) of the radiation source is determined; f) the constant portion (I ₀ ) of the operating current is determined; and g) the strength (E) is determined using the equation E = P 0 - ΔP / ΔI (I 0 - I th ).