AbstractThe introduction of affordable NIR diode array spectrometers now make the measurement of several parameters that are critical to product value readily available. Over the last decade we have seen the introduction of several affordable spectrometers using silicon based sensors, in the 200nm to 1100nm range. This has spawned a host of new applications, including at line/ on line measurements for product quality and process control which were not affordable before. With the introduction of this NIR diode array spectrometer in a similar price range as its visible counterparts, we expect to see similar results. Since the product has just been introduced, not many applications have been developed. Preliminary results by Dr. Arnold Schumann, CREC-UFL show that this spectrometer may be used successfully for the determination of N content (protein) and water in citrus leaves. The spectrometer is currently undergoing application development for use in the corn, soybean and feed industries.This paper also presents the advantages of NIR diode array technology and highlights the performance characteristics of this affordably priced NIR spectrometer.Description of InstrumentThe getSpec NIR 0117 basic spectrometer is a fixed diffraction grating, post disperse device, with a 128 element InGaAs photodiode array. A polychromatic tungsten halogen light source is positioned over the sample and the dispersed light from the sample is collected by a collimating lens and fed back to the spectrometer via fibre optic cable. The measurement range of the spectrometer is 900nm to 1700nm (Fig 1).Presentation of Results for N (protein) and Moisture Analysis in Citrus LeavesThe data presented here is the result of a one-week instrument test by Dr. Arnold Schumann, CREC-UFL. Due to the short test period, these results are somewhat preliminary and conclusions should be considered tentative. They were obtained by Dr. Schumann primarily to test the performance of this particular model of a low-cost NIR spectrometer in performing a specific task - the measurement of protein-N in citrus leaf tissue.Tests were conducted with the instrument after a 45-min warm-up period. Briefly, the geometric alignment of the light - sample - collimator lens triangle must be optimised for maximum collection of dispersed light by the spectrometer. This is easily done interactively by adjusting the light platform while observing the sensor output graphically on the computer screen. The measurement integration time used was 0.008 s. The instrument was then referenced for background light/sensor correction using a white reflective Halon / Abrilon reference. The reference was re-sampled after every ten leaf samples.Results of the calibration and validation runs are presented in Figures 2-9. Figure 10 shows the typical precision which we currently can expect with a Kjeldahl - steam distillation unit. These duplicate N results were from various orange, tangerine and grapefruit leaf samples analysed in 2001. Because the Kjeldahl N is the primary method on which we rely for the NIR calibration, we should not expect the NIR calibrations for dry leaf N to be much better than R 2 =0.89,SE=0.101.Valencia orange dry leaf samples calibrated well (R 2 =0.89, SE=0.08) before outlier removal.Removal of five outliers improved the fit to R 2 =0.93, SE=0.068 (Figure 2). Removal of outliers is usually justified because of the expected variability introduced from Kjeldahl N results. The Valencia calibration was validated with an independent set of dry leaf samples from grapefruit and tangerine trees (Figure 3). The fit was less successful (R 2 =0.73, SE=0.268), and showed considerable bias (slope=0.63), but should be expected when comparing varieties. To correct the bias problem, either a larger global calibration should be developed with more varieties, or a separate calibration should be developed for each variety.Individual fresh Valencia orange leaves were calibrated for water content and N concentration on a fresh weight basis. Leaves were scanned on both upper and lower surfaces. After outlier removal, good, nearly identical calibrations were obtained for water content using both leaf surfaces (Figures 4-5). The outlier samples appearing on scatter plots of both leaf surfaces were the same, implying that there were not errors with the NIR spectra, but more likely that errors were introduced during the gravimetric water measurement. A comparison plot of NIR-predicted leaf water for both leaf surfaces showed the generally good agreement - even including the outliers. Validation of a similarly developed water calibration with tangerine leaves, using the independent Valencia samples, was good once the 12 outliers mentioned previously were removed (Figure 7).NIR-calibrations for fresh leaf N were less successful than dry leaf N or leaf water, probably for two reasons. First, the N concentration in fresh leaves is about half of that in dry leaves, and the abundant water can mask the N features in NIR spectra. Secondly, lab data used in these calibrations consisted of derived leaf N values (% FW), obtained from both Kjeldahl leaf N (% DM), and gravimetric water content. Thus the derived variable %N FW had at least two cumulative sources of error, more than any other calibrated variable. Nevertheless, the calibration was still satisfactory for tangerines (R 2 =0.89, SE=0.049) (Figure 8) after eliminating four outliers. However, validation of tangerine lower leaf N using the upper leaf N calibration equation was much less successful (R 2 =0.63, SE=0.105) after removal of some 19 outliers (Figure 9). Unfortunately leaf N was not determined for individual leaves of the Valencia orange samples, so that they could not be used for validation of tangerine fresh leaf N equations.The instrument performed very well during these tests, and these preliminary results were mostly as expected - very good calibrations for water (abundant concentrations, large NIR peaks), less well for dry leaf N, and worst for fresh leaf N. Validations still need to be improved before one could use this technique for routine measurement. The challenge to measure fresh leaf N may however still be achieved by careful tuning of some instrument operating conditions (e.g. integration time, scan averaging), sample preparation and presentation (e.g. desiccation of dry samples before analysis, flattening of leaves before scanning), and improving the precision of the primary data (Kjeldahl N, leaf water). If this NIR method were adopted for routine leaf analysis, the problem of outliers could be overcome by repeated measures (e.g. quadruplicate NIR scanning and automatically [with software] eliminating any outliers which deviate from the mean by > 2SE). Since each NIR measurement takes less than a second, the additional time required for repeated measures is very acceptable.
Advantages of Diode Array SpectroscopyThere are several advantages of diode array spectroscopy over conventional scanning spectrometers.The following is a list of advantages:
Fast! Fixed grating diode array spectrometers look at all wavelengths of the spectrum simultaneously. This allows us to take a complete spectrum in less than a millisecond. For most applications, 50 to 100 spectra/sec is a reasonable expectation.
Sensitive! Since we look at all the wavelengths simultaneously, one can integrate longer and not pay the penalty like a scanning system that would require separate integrations per wavelength interval.
No Moving Parts! Being a fixed grating device, there are no moving parts and so there is no variability in wavelength between different spectra. This is a great help for calibration transfer between instruments.
Fibre Optic Input! Allows the spectrometer to be located away from the sampling optics, if the environment is hostile. Allows the use of fibre optic probes.
Portable and Lightweight! Our NIR spectrometers can be configure on an ISA pc plug-in card (wt ~ 1lb) or a stand alone unit with USB interface (~ 2lbs).
Low Cost! This is truly a low priced NIR spectrometer with all the features and
performance of higher priced instruments. The instrument is temperature stabilized, allowing a max integration time of 1sec to 2 sec. For almost all applications, this is not a problem at all as the typical integration time is between 1 msec to 50 msecs.ApplicationsApart from being a very versatile laboratory instrument, the getSpec NIR 0117 basic spectrometer is undergoing laboratory and field tests for numerous applications, including the following:
Moisture content in corn
Moisture, Fat and Protein content in Soy Beans
In-Situ monitor for Thin Film deposition
Monitoring of Erbium Doped Amplifier outputWe expect the number of different applications to increase rapidly, as potential users become aware of the product.ConclusionThe getSpec NIR 0117 basic spectrometer, with its high performance specifications/features and affordable pricing, should make NIR spectroscopy available for numerous process and quality control applications. It also provides instrument makers and system integrators a very cost effective and viable platform on which to base their products. This device also provides researchers, especially at Universities with an affordable NIR spectrometer.
