A gas sample separated from transformer oil is circulated through an NDIR gas sensor system which obtains an acetylene concentration by calculating a detected acetylene concentration obtained by an absorption biased (“AB”) NDIR acetylene gas sensor, calculating a detected carbon dioxide concentration obtained by an AB NDIR carbon dioxide gas sensor, calculating a detected water vapor concentration obtained by an AB NDIR water vapor NDIR gas sensor and then determining the acetylene concentration from the detected acetylene concentration through use of the detected carbon dioxide and water vapor concentrations to compensate for their interference.

DETAILED DESCRIPTION OF THE INVENTION
The present invention advances a methodology for the design of an NDIR gas sensor system which is capable of detecting sub-ppm levels of Acetylene gas in transformer oil in the presence of multiple fault gases and moisture. In order to achieve a successful methodology, a number of technical obstacles must be recognized, addressed and overcome.
The present invention addresses the drawback of NDIR gas sensors for detecting gases in the sub-ppm concentration ranges having very weak absorption bands located in the far infrared region of the electromagnetic spectrum (>12.0μ) and very unique ultra-narrow line shapes and overcomes these drawbacks by a novel NDIR design methodology. This design methodology comprises 1) matching the ultra-narrow line shape of the gas' absorption band with a carefully designed narrow band pass filter; 2) making sure that the out-of-band leakage of the filter is not more than one part in 103; 3) utilizing a MEMS infrared source with built-in collimating optics and a large area thermopile detector for S/N optimization; 4) using a sufficiently long path length waveguide sample chamber for achieving adequate modulation by the target gas and 5) including a sensitive CO2 and a sensitive Dew Point NDIR sensor for simultaneously measuring the concentration levels of both gases in order to compensate for any possible interferences by these gases to the measurement of the target gas.
The first step needed to advance a successful methodology is to design a narrow infrared band pass filter with a center wavelength (CWL) specified at 13.70μ coincident with the center wavelength of the absorption line of Acetylene and with an adequately narrow full width at half-maximum (FWHM). The FWHM of the specified narrow band pass filter must be narrow enough so as to afford adequate signal modulation for measuring concentration levels down to sub-ppm or ppb for Acetylene gas. In order to be able to detect Acetylene down to +/−0.5 ppm concentration level, a thermopile detector is used with an average D* around 1×108 cm Hz0.5 W−1 and the FWHM of the narrow band pass filter must not exceed 0.1μ.
With the required spectral properties for the narrow band pass filter determined, step two is to calculate the path length for the sample chamber required to be able to accurately measure concentration levels of Acetylene down to +/−0.5 ppm. Calculations showed that based upon the very weak absorption of Acetylene gas, the minimum path length required to get the job done is no less than ˜20 inches.
The third step is to recognize the possibility of interference effects on the measurement accuracy for Acetylene by other gases and then compensate for any such interference. After a careful study it was found that only two gases, namely carbon dioxide (CO2) and water vapor (H2O), could severely interfere with the measurement accuracy of Acetylene at 13.70μ. For CO2, it is due to the presence of a very strong absorption band at ˜15.0μ. Water vapor on the other hand has absorption bands almost everywhere in the infrared extending way into the far infrared wavelength region. Even though water vapor has several very strong absorption bands in the infrared, those that are present around 13.70μ are extremely weak. Due to the fact that the FWHM of the specified filter is only ˜0.1μ, it is not expected that water vapor will cause any interference problem. However, the state-of-the-art for making a narrow band pass interference filter in the far infrared today only guarantees spectral blocking outside of the designed pass band to be no better than 1:103. Thus, because of the omnipresence of water vapor absorption lines everywhere, the out-of-the-band absorption, i.e. for the wavelength region outside of the FWHM at 13.70μ, the interference effect of water vapor is still very significant. Furthermore, because of the fact that the amount of water vapor present in air or dissolved in the transformer oil could vary from a few to tens of mmHg (1 mmHg=1,316 ppm at Standard Temperature and Pressure [STP] conditions), its interference effects on the measurement accuracy of Acetylene cannot be ignored. In conclusion, after a careful analysis of the subject, the interference effects of both CO2 and water vapor must be carefully taken into consideration before the measurement accuracy of Acetylene down to sub-ppm levels can be confidently realized.
To take the interference effects of both CO2 and water vapor on the measurement accuracy of Acetylene into consideration is in theory not a complicated task. One practical and viable approach is to first ascertain quantitatively, based upon the narrow band pass filter to be used in the design of the Acetylene sensor, the amount of Acetylene that is equivalent to a certain known amounts of both CO2 and water vapor. The experimentally measured results indicated that under the same measurement conditions, the presence of 5,000 ppm of CO2 would be equivalent to the presence of 20 ppm of Acetylene and the presence of 8.0 mmHg of water vapor (H2O) would be equivalent to 40 ppm of Acetylene. Thus, in order to be able to guarantee the accuracy of +/−0.5 ppm of Acetylene, one must be able to accurately detect +/−5,000/40 or +/−125 ppm of CO2 and +/−8.0/80 mmHg or +/−0.1 mmHg of water vapor simultaneously while measuring Acetylene using the same sensor in order to adequately compensate for their presence as interfering gases.
As it turns out, it is possible to design an output stable NDIR CO2 gas sensor capable of detecting the gas with an accuracy of +/−100 ppm and with a reasonable response time (0-90%) commensurate with that required for measuring Acetylene which is 3-5 minutes. The challenge lies in the fact that an output stable NDIR dew point or water vapor sensor capable of detecting the gas with an accuracy of +/−0.1 mmHg simply cannot be found anywhere today. Using a conventional humidity sensor and converting its output to water vapor pressure by measuring also the temperature will not work because of two factors. First, humidity sensors are known to have output drifts that are difficult to determine over time. Second, presently available humidity sensors are not accurate enough to meet the required accuracy of +/−0.1 mmHg. Thus, the present invention has to embark on an additional innovative step to realize an NDIR gas sensor capable of measuring water vapor pressure with an accuracy of +/−0.1 mmHg and a response time (0-90%) of 3-5 minutes.
Finally in order to complete a well-thought-out measurement methodology, the best available infrared source and thermopile detector must be selected for the design of the sensor. At the same time the accompanying signal processing electronic system should also be designed to be detector-noise-limited. In other words, the overall noise of the sensor can only be limited by the detector and not by the infrared source, the electronic signal processing circuit or any part of the remaining system.