| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Department of Food Science, Cornell University, Ithaca, NY 14853
Dairy Programs of the USDA Agricultural Marketing Service, Carrollton, TX 75006
2 Corresponding author: jl72{at}cornell.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: evaluation • mid-infrared • milk • precalibration
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
In 1989, we described the history of milk analysis for payment testing as it evolved from chemical methods to the Milko-tester and finally MIR (Barbano and Clark, 1989). At that time, we outlined the challenges that MIR-based testing presented and gave a general overview of plans to develop procedures to assure accurate and uniform milk testing that could be implemented throughout the dairy industry. The 3 major components in achieving this goal were 1) establishing and documenting the performance of chemical reference methods for calibration of the instruments, 2) developing uniform and meaningful procedures for assessing instrument performance and integrity prior to calibration, and 3) developing and validating instrument calibration procedures. This work was initiated and supported by the Test Procedures Committee of the USDA Federal Milk Market Administrators (FMMA), in collaboration with the authors at Cornell University.
All MIR analyzers are secondary testing instruments that require calibration by chemical reference methods. Thus, the first task was to identify appropriate reference methods for milk payment testing. Once methods were identified, they were reviewed, modified to improve performance, and clearly described. Interlaboratory collaborative studies were conducted according to the guidelines of AOAC and method performance was documented. Using this approach, reference methods for milk true protein (Kjeldahl protein N; methods 991.22, 991.23), fat (modified Mojonnier ether extraction; method 989.05), total solids (oven dry; methods 990.20, 990.19), and SNF (by difference; method 990.21) were adopted as AOAC final action methods (AOAC, 2000). A collaborative study for determination of milk anhydrous lactose using an enzymatic spectrophotometric procedure is currently underway.
The second component of achieving accurate tests with MIR involves ensuring proper instrument performance before calibration by means of what is commonly called precalibration evaluation. Many of the basic principles of precalibration are described both in the literature (Barbano and Clark, 1989; AOAC, 2000; IDF, 2000; Wehr and Frank, 2004) and in the instrument manufacturers’ manuals. The purpose of the present paper is to present a detailed account and summary of the results of the procedures developed and implemented by the USDA FMMA. Work on the final component in achieving optimum testing performance of MIR instruments, calibration and validation, is in progress.
|
PRINCIPLE OF INFRARED ANALYSIS |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
More recent MIR instrumentation employs FTIR technology that uses an interferometer to produce full spectral information within the midinfrared region (Agnet, 1998). However, most FTIR instruments can be set up to simulate filter instruments by selecting the appropriate primary and reference wavelengths, and bandwidth (fixed wavelength approach). Currently, only the fixed wavelength approach to MIR milk analysis (using either fixed filter or FTIR instrumentation) has been approved as an official method (AOAC, 2000) for use in payment testing in the United States.
|
PRECALIBRATION, CALIBRATION, AND VALIDATION |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
. At the end of precalibration, intercorrection factors are established.
|
Applying intercorrection factors to the uncorrected signal produces an intercorrected signal. Calibration involves using linear regression to establish a secondary slope and intercept to convert the intercorrected signal to a corrected signal (Figure 1
). This is done using a set of milk-based calibration samples (typically 10 to 14) whose composition has been established using chemical reference tests. Finally, calibration is verified by testing an independent set of validation milk samples and comparing the corrected signals to the chemical reference tests of these same milks (Figure 1).
The general equations for the application of intercorrections, secondary slope, and intercept are presented in Table 1. The scheme presented here is referred to as the fixed intercorrection calibration approach, because the intercorrection factors, once established, are, for all practical purposes, constant for an instrument (if the wavelengths are not altered and the primary slope of the uncorrected signal is kept constant). Calibration in this case consists solely of adjusting the final (secondary) slope and intercept. Note that the uncorrected signal is defined as the MIR reading before the application of intercorrection factors and the secondary slope and intercept. The uncorrected signal is essentially the linearized, gain-adjusted (described later on) instrument signal with the intercorrection factors set to 0 and the secondary slope and intercept set to 1 and 0, respectively.
|
|
PRECALIBRATION PROCEDURES |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The components of precalibration listed in Figure 1 are intended to be conducted in the order given and on a regular (e.g., monthly) basis. Although all the procedures can technically be applied to all models of instruments, the operator needs access to the uncorrected signal (with the ability to adjust linearity and gain separately) and intercorrection factors. We suggest that instrument manufacturers make these easily available to the operator of an MIR milk analyzer. The criteria used in the USDA FMMA for passing or failing a particular test are based on the manufacturer’s guidelines, the recommendations of the various standardization organizations (AOAC, 2000; IDF, 2000; Wehr and Frank, 2004), and our experience working with different instruments and laboratories.
For each of the precalibration procedures, it is critical that all the zeroing solutions and milk samples to be tested be maintained at the appropriate temperature. We recommend 41 ± 1°C, which is 1°C higher than current standards (AOAC, 2000; IDF, 2000). The reason for this is 2-fold. First, the efficiency of the instrument homogenizer is dependent on sample temperature and, within a reasonable range, higher temperatures result in better efficiency (Walstra and Jenness, 1984). However, this shortens the time that the milk can remain in a tempering water bath because increasing the temperature will accelerate milk deterioration and oiling off of fat. A temperature of 41°C was chosen as a reasonable compromise between the amount of time a good quality milk can stay in the tempering water bath (approximately 15 min after reaching 41°C) and the increase in performance gained by using higher temperatures. The second reason for choosing 41°C is based on the realities of high-volume testing where automatic conveyer belts are used to feed racks of samples to the instrument. Once the samples in the racks are removed from the tempering water bath and placed on the conveyer, they start to cool. This can result in a significant decrease in sample temperature especially for the last samples in the rack that are exposed to the ambient temperature the longest. In addition to maintaining proper sample temperature, the milk (formulated or not) used in all the precalibration procedures must be of good quality and thoroughly mixed before being presented to the instrument. Although seemingly trivial, these requirements cannot be overemphasized. Testing irregularities are often improperly attributed to instrument problems when in reality they are caused by temperature deviations, poor quality samples, or improper sample handling.
Throughout the following discussion, milk and zeroing solution are described as being presented to the instruments in vials. These are standard 47-mL volume plastic vials with lids, filled approximately three-fourths full (e.g., Capitol Vial, Fulton, NY). The zeroing solution recommended by the instrument manufacturer (typically distilled water containing 0.01% Triton X-100) is used for all zero setting and checks. All evaluations are done solely using the uncorrected readings (signals) from the appropriate component channel or channels under evaluation.
A final comment is that, in routine operation, some instruments can be set up to apply a purge correction to the signal. All the precalibration procedures require that this option be disabled when performing the evaluations.
Procedure 1: Flow System Checks
Flow system checks involve evaluating the flow system and homogenizer temperature. The specific procedures and pass/fail criteria for evaluating the flow system are instrument-dependent and the operator needs to refer to the instrument manual.
The block of the homogenizer has small empty holes that can be filled with thermal heat-sink compound. The temperature of the homogenizer is measured by inserting a temperature probe in one of the holes. The temperature should be between 40 ± 2°C. If not within this range, the temperature should be adjusted according to the instrument manufacturer’s instructions.
Procedure 2: Homogenization Efficiency
The purpose of checking instrument homogenization efficiency is to determine if the fat globules are being broken down into small enough sizes to give accurate test results and minimize light scattering. Large milk fat globules cause spectral effects that result in analytical error and poor repeatability (Smith et al., 1994, 1995). Homogenizers wear over time and need to be replaced or rebuilt periodically. The homogenization test allows the operator to monitor wear and schedule replacement before test results are adversely affected.
The "recycle" test is the most common method for evaluating homogenizer performance (AOAC, 2000; IDF, 2000; Wehr and Frank, 2004). This test consists of running a nonhomogenized milk through the instrument, recording the instrument readings from the fat channels, and collecting the milk as it exits the machine. The milk collected upon exit (first-pass milk) represents milk that has been homogenized by the instrument homogenizer. This first-pass milk is then run through the instrument a second time, and again the readings are recorded. Homogenization is considered adequate if the difference in readings between the first and second passes of the milk is <0.05%. The limitations of the recycle test have been previously discussed (Smith et al., 1993a, 1995). The recycle test is not used in the USDA FMMA program.
A more informative method is direct determination of the milk fat globule size distribution after whole milk is homogenized through the instrument. This is done using a laser light-scattering particle size analyzer capable of determining particle size distribution between 1 and 80 µm, as described by Smith et al. (1995). Fat globule size needs to be measured using forward light scattering (Smith et al., 1995), not side scattering. This is to minimize the contribution of casein micelles to light scattering, which would result in an overestimation of homogenization efficiency. The milk to be evaluated should be between 3 to 3.5% fat, unhomogenized, pasteurized (to inactivate milk lipase), preserved, and of good quality. Two vials of this milk are prepared and warmed to 41 ± 1°C. Between 3 to 4 instrument measurement uptake cycles are run using the first vial to flush out the flow system and fill it completely with the milk of interest. This also helps with temperature stabilization of the system by heating the uptake tubing before the homogenizer. The second vial is presented to the instrument and the milk from 3 measurement cycles (total volume 
5 to 15 mL) is collected from the cuvette by-pass line. Only the cuvette by-pass line should be used because the milk from the bleeder line is not homogenized and there is too little milk exiting the cuvette line. The instrument-homogenized milk is immediately cooled in ice and refrigerated before particle size analysis.
Results from the particle size analysis are evaluated by looking at the actual scan (from 1 to 80 µm) and the d(0.9) value (Figure 2). The actual scan of the instrument-homogenized milk should be unimodal and resemble a normal distribution curve. If additional peaks (other than trace) appear on the high side outside the main distribution (e.g., >10 µm), this usually indicates fat globule clustering or coalescence of free fat and may be symptomatic of poor homogenization or sample quality problems. The d(0.9), calculated on a volume basis, is the mean globule diameter below which 90% of the fat volume is contained. It is a more rugged parameter to evaluate than the volume mean diameter, d(4, 3), because it is not influenced by the occasional appearance of a large air bubble in the distribution. Homogenizers with a d(0.9) >1.70 µm are wearing out and should be replaced; homogenizers with a d(0.9) of >2 µm are unacceptable (Smith et al., 1995)
|
0.04.
Procedure 4: Initial Zeros/Zero Shift
The purpose of the zero shift test is to determine if the windows of the instrument’s cuvette are deteriorating. Regular commercially homogenized milk is used for this test, so as not to confound the interpretation with issues unique to raw milk. Four vials of homogenized milk and 4 or 5 of zeroing solution are warmed to 41 ± 1°C. The flow system is cleaned according the manufacturer’s instructions immediately before the test and the instrument is zeroed. After setting the zero, 3 measurements of zeroing solution from a new vial are taken and recorded for each channel (initial zeros). This is followed by 12 sample uptake cycles (do not use purge) of the homogenized milk (3 from each vial), in which the actual readings are unimportant. Three sample uptake cycles of zeroing solution from a single vial are then run to flush out the flow system to negate any influence of purging efficiency. Finally, another 3 measurements of zeroing solution from a new vial are made and recorded (final zeros).
The average of the 3 initial zero readings should be 0.00 ± 0.02 (initial zero check) for each channel. The difference between the average of the 3 initial zero reading and the average of the 3 final zero readings should be ± 0.02% on all channels (zero shift evaluation). Problems with cuvette wear often become evident on the protein or lactose channels before they appear on the fat channels and manifest themselves as a shift in readings due to rapid accumulation of milk solids on the windows of a freshly cleaned cuvette. Regardless, a consistent failure on any channel means the cuvette should be replaced. The evaluation of initial zeros and zero shift is formally conducted early on in the precalibration evaluation sequence, but the basic procedure is an integral part of most of the subsequent tests to assure that the zero is stable and maintained.
Procedure 5: Linearity
The purpose of linearity evaluation is to determine if linear changes in component concentration (wt/wt) are reflected in linear changes in the uncorrected signals. Linearity adjustment is necessary because the Beer-Lambert law is not always valid for low-resolution spectroscopy (Biggs et al., 1987). Nonlinearity has a negative impact on the standard deviation of the difference between instrument and chemical test results and can cause errors in the determination of the intercorrection factors (Biggs, 1979).
Smith et al. (1993b) described the limitations of the AOAC (2000) and IDF (1990, 2000) procedures for linearity evaluation and proposed an alternative designed to overcome them. Linearity is assessed using 2 sets of solutions: one for the fat channels and one for both the protein and lactose channels. The concentration ranges chosen for the linearity solutions are designed to encompass the range over which the final calibration will be used.
The fat linearity solutions are made using a 6% fat (approximate) stock material prepared by diluting commercial homogenized, pasteurized half-and-half (10 to 18% milk fat; Code of Federal Regulations, 2004) with pasteurized skim milk (0.2% milk fat). The stock material is diluted (wt/wt) with distilled water to achieve a series of 5 fat concentrations: 2, 3, 4, 5, and 6% fat. The fat chemistry for each solution is calculated from the weights used during preparation. Determination of the percentage fat in the stock solution by chemical analysis is not a requirement, although it is advised as a quality control tool.
For the protein/lactose linearity solutions, 6 solutions are prepared by diluting a can of evaporated skim milk with distilled water. In the absence of directly measuring true protein and lactose, the evaporated skim milk can be assumed to contain approximately 7% true protein and 11% anhydrous lactose. For true protein, the solutions range from approximately 2.8 to 3.9%. The corresponding range for anhydrous lactose is 4.3 to 6.0%. The chemistry for each solution is calculated from the weights obtained during preparation.
For each set of linearity solutions (either fat or protein/lactose), the instrument is zeroed and then 3 measurements of 41 ± 1°C zeroing solution are recorded (initial zeros). The linearity solutions (41 ± 1°C) are presented to the instrument, going from low to high concentration and taking 3 measurements per vial from the channel(s) under evaluation. Three measurement cycles of zeroing solution from a single vial are then run to flush out the flow system, followed by 3 recorded measurements of zeroing solution (final zeros).
The initial zeros and zero shift are calculated and evaluated as previous described. Failure to pass these tests invalidates the linearity evaluation. For each linearity solution, the first reading is not used to avoid possible distortions due to poor purging efficiency. The second and third readings are evaluated for repeatability and only accepted if the difference between the readings is 
0.04%. Acceptable readings are averaged. A nonzero forcing linear regression is conducted with actual chemistry (calculated from the dilutions of the stock solutions) as the dependent variable and the uncorrected signal as the independent variable. Mid-infrared predicted chemical values are calculated using the derived regression equation. Residual nonlinearity values, calculated by subtracting actual chemistry from MIR-predicted chemistry, are plotted against actual chemistry (Figure 3). Residuals of ± 0.02% within the calibration range are deemed acceptable. For examples of residual plot calculations and additional details on making up the linearity evaluation solutions for fat and protein/lactose, see Smith et al. (1993b). Residual nonlinearity will influence the setting of primary slope and intercorrection factors (Biggs, 1979) and will weaken the agreement between instrument predicted and actual chemistry values.
|
The procedure described below uses milk-based ingredients to formulate pairs of milks differing only in the concentration of the component of interest. The difference between the 2 uncorrected readings from the primary component channel divided by difference in actual chemistry is the primary slope. An important characteristic within each pair of milks is a similar background chemistry. This removes the influence of variation in the coabsorbance of secondary components on the primary signal.
Separate sets of primary slope solutions are prepared for fat, protein, and lactose, each consisting of a "low" and "high" solution. The solutions are formulated using homogenized (pasteurized) whole milk, pasteurized skim milk, low-heat nonfat dry milk powder (NDM), lactose monohydrate, and a 6% fat stock solution as ingredients. The 6% fat stock solution is prepared by diluting commercial homogenized, pasteurized half-and-half with pasteurized skim milk. Chemical reference tests are run on each of the ingredients (except lactose monohydrate) to determine the true protein, fat and total solids contents. Lactose is determined either by difference or by spectrophotometric enzymatic analysis. Lactose by difference is calculated as % anhydrous lactose = % total solids – (% fat + % true protein + 0.19 + % ash), where 0.19 is the NPN factor to convert true protein to CP and % ash is either directly measured (method 945.46; AOAC, 2000) or indirectly calculated using an updated version of the equation described by Lynch et al. (1990) where % ash = (% true protein x 0.0596) + 0.5379. The spectrophotometric enzymatic analysis of lactose is based on AOAC method 984.15 (AOAC, 2000), modified by weighing all volume additions. The ingredients used to formulate the solutions and their typical composition are presented in Table 2.
|
gives an example of the formulations and final component concentrations, using the composition of the ingredients listed in Table 2. Note that only the low and high solutions in Table 3 are used for evaluation of primary slope.
|
Primary slope is dependent on cuvette path length, the level of light intensity striking the detector, and the linearity of the uncorrected signal. Generally, as a cuvette is used over time, there is erosion of the windows and the path length of the cuvette increases. This will cause the primary slope to increase gradually for all component channels. Current practice in the USDA FMMA laboratories is to adjust primary slope back to 1.00 if it is 0.95 or 1.05 on any channel. However, large sudden changes in primary slope are typically caused by a change in the light source, detector, or cuvette. In these cases, the primary slope must be checked immediately after instrument repair and adjusted to 1.00 for all channels. This allows the optimized intercorrection factors already established for that instrument to continue to function properly.
The importance of maintaining primary slope is illustrated in the following example. Assume that the fat B uncorrected signal is 4.58 at a primary slope of 1.00. If the slope is changed to the upper limit of acceptability (1.05), then the signal increases to 4.81, and if it is changed to the lower limit (0.95) the signal becomes 4.35. Thus, a deviation in primary slope from 1.00 even within the narrow range of acceptable primary slope settings can change the uncorrected reading by as much as 0.23. If this were the only impact of variation in primary slope on the result, a compensating change in secondary slope would take care of it. However, adjusting secondary slope would not address the issue of the proper function of the intercorrection factors
Procedure 7: Milk Repeatability
Although seemingly counterintuitive, the milk repeatability test is conducted after linearity and primary slope have been evaluated and adjusted if necessary. The reason is that nonoptimized primary slope settings can either mask or cause poor repeatability. On a practical basis, an instrument that is not repeatable inherently gives unreliable results.
The milk repeatability test consists of 2 separate evaluations, the first using raw whole milk and the second using pasteurized, homogenized whole milk. If the instrument passes the raw milk evaluation, there is no need to run the homogenized milk. However, if the instrument fails the raw milk evaluation, running the homogenized milk helps the operator determine the cause of the failure. If the instrument fails the raw milk evaluation but passes the evaluation of the homogenized milk, then poor repeatability is most likely caused by improper sample mixing or splitting, homogenizer failure, or sample quality. If the instrument fails both the raw and homogenized milk evaluation, then the problem is more serious and it is likely that the instrument was close to the failure limit on the previously performed water repeatability test.
For each test, 6 vials of milk (raw or homogenized, depending on the test) and at least 4 of zeroing solution (for setting the zero and determining zero shift) are warmed to 41 ± 1°C. The instrument is zeroed, 3 measurements of zeroing solution are taken (initial zeros), followed by 18 measurement cycles of the milk (3 from each vial). Three measurement cycles of zeroing solution from a single vial are then run to flush out the flow system and another 3 measurements of zeroing solution from a new vial are made and recorded (final zeros).
The initial zeros and zero shift are evaluated as previously described and the range of the readings is calculated for each channel from the 2 through 18 measurements of the milk. The first measurement is not used to avoid possible distortions due to poor purging efficiency. The range of milk readings for each channel should be 0.04.
Procedure 8: Purging Efficiency
Evaluation of purging efficiency is a check on the amount of carryover from one sample vial to the next. The evaluation should initially be carried out using the normal mode of instrument operation: either manual or automatic. Automatic mixing and sampling poses additional carryover risk because there can be carryover from the mechanical stirrer used to mix the sample before measurement. Liquid adhering to the stirrer will be transferred from one sample to another. Because most instruments used for payment testing or herd evaluations make just one measurement per sample vial, the impact of carryover on actual test results can be significant. The influence of purging efficiency is made clear when going from zeroing solution (referred to as water) to milk (lowering the milk reading) or milk to zeroing solution (increasing the zero). Thus, the evaluation looks at these 2 situations.
To perform the test, 12 vials (of the same volume used for routine testing) are filled with zeroing solution and 10 vials are filled with pasteurized, homogenized whole milk. The solutions are warmed to 41 ± 1°C, the instrument is zeroed, and the initial zeros are recorded. The 22 vials are then presented to the instrument in the order of alternating pairs: water (W1), water (W2), milk (M1), milk (M2), water (W1), water (W2), etc. A single measurement is made for each vial and the readings from each channel are recorded. At the conclusion, 3 measurement cycles of zeroing solution from a single vial are made to obtain the final zeros.
As usual, the initial zeros and zero shift are evaluated. Purging efficiency is calculated as follows
 |
 |
Purging efficiency should be 99.0% for all channels for both water to milk and milk to water. If an instrument fails the purging efficiency test in automatic mode then the test should be run again in manual mode. If the instrument fails in automatic mode but passes in manual mode, then the most likely explanation is carryover from the stirrer (which should be repaired, replaced, or a different shape should be utilized).
The procedures of IDF (2000) and AOAC (2000) evaluate purging efficiency in a similar manner, but only consider the efficiency of water to milk, and for AOAC (2000), the evaluation is limited to the fat channels. Both the water to milk and milk to water evaluations are described in the 17th edition of Standard Methods (Wehr and Frank, 2004). The pass/fail criteria for all the standards organizations are at the 99% level.
Procedure 9: Intercorrection Factors
As discussed previously, intercorrection factors are necessary to compensate for the influence of water displacement and absorbance of secondary components at the wavelength of the component of interest. Intercorrection factors for each channel are a function of linearity, primary slope, and the center wavelength and bandwidth of both the primary and reference wavelengths. Once established, intercorrection factors should be relatively constant unless there is a major change in the instrument (e.g., filter or filter angle for filter-based instruments; central wavelength and bandwidth for FTIR instruments) or changes are made to primary slope or linearity. From this discussion, it should be apparent that it is critical to establish linearity and primary slope of the uncorrected signal before setting or evaluating intercorrection factors.
For each primary component channel, separate intercorrections must be established for each of the secondary components. For instance, when true protein is measured at its primary wavelength (6.465 µm), intercorrections are applied for the effects of lactose on protein and fat on protein. Thus, for instruments using the full complement of channels (fat A, fat B, protein, and lactose), there will be 2 intercorrection factors for each channel (8 total). The issue arises as to whether the fat A or fat B signal should be used to establish intercorrections for the effect of fat on protein and lactose. Historically, the fat A signal was used because that was the only fat filter available in early instrumentation. Most modern filter instruments, however, are equipped with both fat A and B filters and, of course, FTIR gives access to both wavelengths. All the instruments in the FMMA program use the fat B wavelength to establish the fat intercorrections for protein and lactose. We consider it preferable to fat A because the magnitude of the fat B signal is greater and thus has the potential to provide greater sensitivity.
Intercorrection factors are both set and evaluated by determining them experimentally and comparing them to previously established values. The solutions used for the evaluation are identical to those used for primary slope except that a "medium" solution is added to each set (Table 3). Note that the solutions (e.g., low, medium, and high fat) are used to determine the effect of the secondary component (e.g., fat) on the primary signals of the other 2 components (e.g., lactose and protein). The ingredients used to formulate the solutions and their typical composition is presented in Table 2, and an example of the formulations and final component concentrations is given in Table 3. Formulations are done by weight, background component composition is similar within each set, and actual composition is determined by chemical analysis, as was described for primary slope. Each set of intercorrection solutions are run on the instrument as previously described (i.e., solutions tempered to 41 ± 1°C, 3 readings per vial, initial zeros and zero shift evaluation between each set of solutions) and the average of the second and third readings, evaluated for repeatability, are used to calculate the intercorrection factors.
The intercorrection factor for the secondary component on the primary component determined using the low and high solutions (named after the secondary component) is calculated as follows:
 |
where PriChem = g/100 g of primary component determined by chemical analysis, PriIR = uncorrected signal of primary component, SecChem = g/100 g of secondary component determined by chemical analysis, low = low solution, and high = high solution. An example of the calculations for the effect of fat on protein is given in Table 4.
|
|
USDA FMMA PRECALIBRATION EVALUATION PROGRAM |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
. Of the 20 instruments listed, 14 are filter-based and 6 are FTIR-based set up to simulate traditional filter instruments. All the instruments operate using fat A, protein, lactose, and fat B wavelengths and most use the 4 traditional reference filter wavelengths. The Dairy Lab 2 uses the protein reference wavelength for all channels and the Bentley 2000 uses the fat B reference wavelength for the fat B channel and the fat A reference wavelength for the other channels. Milk-o-Scan 6000 instruments (Foss Electric, Hillerød, Denmark) in the FMMA system that operate in the traditional fixed filter calibration mode have not yet been included in the precalibration evaluation program because the operator does not have access to uncorrected signals, linearity adjustments, and intercorrection factors.
|
Under this program, the MIR instruments in the FMMA system have routinely met the passing criteria for all precalibration procedures. Failure of any one procedure results in action to correct the problem. Determining the cause of failure is facilitated by the association of specific tests with identifiable causes. This has shortened downtime and reduced maintenance service costs. Especially valuable are the homogenization checks because they not only allow the user to preemptively replace or rebuild a homogenizer, but they also save the user from unnecessarily incurring the cost of replacing a correctly functioning homogenizer.
Of special interest are the intercorrection factors. When the program was initiated in 1995, no criteria were set for evaluating intercorrection factors. At that time, most of the instruments in the FMMA program were filter based and thus, the intercorrection factors were, in large part, dependent on the fixed characteristics of the individual instruments (e.g., dependent on filter wavelength and bandwidth). What was under operator’s control, however, was linearity and primary slope, and it was not known at the time what effect controlling these key determinants would have on the variability of intercorrection factors within and between instruments.
As previously mentioned, intercorrection factors are routinely evaluated every 6 mo by determining them experimentally and comparing them to the previous values. In the absence of criteria for evaluation, the decision was made to change them by default each time they were checked. What we found was that intercorrections factors remained very constant within an instrument (when linearity and primary slope were maintained within tolerances) and there was little effect of most instrument repairs on intercorrection factors. Furthermore, intercorrection factors tended to be very similar among instruments with the same wavelength characteristics.
Table 6 presents the most recent intercorrection factors for the instruments that are or have participated in the precalibration evaluation program, grouped by filter/wavelength. The standard deviation (SD) is indicative of the variation among instruments. For the categories in which information is available for a reasonable number of instruments (i.e., the 12 filter- and 6 FTIR-based instruments using 4 primary and 4 reference filters/wavelengths), intercorrection factors are similar among instruments with the same characteristics (SD ranging from 0.008 to 0.029 and 0.002 to 0.022 for the filter- and FTIR-based instruments, respectively) and, with the exception of the effects of lactose and protein on fat B, the differences between the corresponding intercorrection factors for the 12 filter- and 6 FTIR-based instruments are also very small (average difference ranging from 0.002 to 0.034).
|
The effect of reference filter wavelength on the intercorrection factors is also illustrated in Table 6. Although the intercorrections for both the filter and FTIR instruments using the same 4 reference wavelengths were similar, the instruments using just 1 or 2 reference filters were not. When a single reference wavelength of 6.7 µm (protein reference wavelength) was used, significant changes in the magnitude and sign were observed for most of the intercorrections except the effects of fat B and lactose on protein (Table 6). When 2 reference wavelengths were used (5.6 µm for fat A, protein, and lactose; 3.6 µm for fat B), the effects were even more variable. The main message here is that, assuming common primary wavelengths, any differences between instrument type (filter or FTIR) or between models appear small when compared with the effect of the selection of reference wavelengths (and bandwidth). A clear definition of both primary and reference wavelengths is needed in official methods to ensure consistency. The most desirable situation for analytical testing is to have the intercorrections as small as possible.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Received for publication November 13, 2005. Accepted for publication January 13, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONPRINCIPLE OF INFRARED ANALYSISPRECALIBRATION, CALIBRATION, AND...PRECALIBRATION PROCEDURESUSDA FMMA PRECALIBRATION...CONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. E. Kaylegian, J. M. Lynch, J. R. Fleming, and D. M. Barbano Influence of fatty acid chain length and unsaturation on mid-infrared milk analysis J Dairy Sci, June 1, 2009; 92(6): 2485 - 2501. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. E. Kaylegian, D. A. Dwyer, J. M. Lynch, D. E. Bauman, J. R. Fleming, and D. M. Barbano Impact of fatty acid composition on the accuracy of mid-infrared fat analysis of farm milks J Dairy Sci, June 1, 2009; 92(6): 2502 - 2513. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. De Marchi, C. C. Fagan, C. P. O'Donnell, A. Cecchinato, R. Dal Zotto, M. Cassandro, M. Penasa, and G. Bittante Prediction of coagulation properties, titratable acidity, and pH of bovine milk using mid-infrared spectroscopy J Dairy Sci, January 1, 2009; 92(1): 423 - 432. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Dal Zotto, M. De Marchi, A. Cecchinato, M. Penasa, M. Cassandro, P. Carnier, L. Gallo, and G. Bittante Reproducibility and Repeatability of Measures of Milk Coagulation Properties and Predictive Ability of Mid-Infrared Reflectance Spectroscopy J Dairy Sci, October 1, 2008; 91(10): 4103 - 4112. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. E. Kaylegian, J. M. Lynch, J. R. Fleming, and D. M. Barbano Lipolysis and Proteolysis of Modified and Producer Milks Used for Calibration of Mid-Infrared Milk Analyzers J Dairy Sci, February 1, 2007; 90(2): 602 - 615. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. E. Kaylegian, J. M. Lynch, G. E. Houghton, J. R. Fleming, and D. M. Barbano Modified versus producer milk calibration: mid-infrared analyzer performance validation. J Dairy Sci, August 1, 2006; 89(8): 2833 - 2845. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
