Editor�s Note: The following is an excerpt from a much larger paper prepared by Dr. Billo, et al. The full text, which contains considerably more information, tables, explanations and mathematical models, is available in the Members Only section of the Parcel Shipping & Distribution Web site at www.psdmag.com. We are very pleased to be able to offer this ground-breaking research exclusively to our readers and Web site members.
    In a high-speed automated sortation environment, with millions of packages being processed each day, a 1% rejection rate can lead to thousands of packages requiring extra handling by the sortation system.
    A contributing factor to rejection is a low symbol read rate, usually attributed to a variety of factors associated with either the symbol�s print quality, its placement on the package or issues related to the conveyor. According to the ANSI X3.182 Barcode Print Quality Guideline specification, factors such as symbol contrast, the X dimension (the nominal width of a barcode symbol�s narrowest element) and bar growth and reduction can all impact the read rate of a barcode tracking system. In addition, factors such as aspect ratio (height of symbol compared to width), orientation, quiet zones and conveyor speed can further impact read rates.
    In response to this problem, many distribution companies have arbitrarily set high barcode print quality limits without any systematic experimentation of either the independent or interactive effects of the above factors. These requirements place undue costs on both the package supplier and shipper. For the supplier, costs associated with different quality label stock and printing technologies must be incurred. For the shipper, equipment and labor costs associated with ensuring quality control of the symbols on incoming packages must be incurred.
    Presumably, as the quality of the barcode symbol improves, the probability the symbol will scan faster and decode properly increases. In an environment where millions of packages are sorted and distributed around the world, the quality of a barcode symbol should be at its highest possible level. However, the question arises as to the impact on the read rate when the barcode symbol�s quality is reduced.
    Quality at Varied Levels
    The approach used in this research was to generate a series of barcoded labels with systematically varied quality levels and then present the symbols to a fixed mount scanner, commonly used for high-speed sortation. An appropriate experimental design was used to isolate the effects of symbol quality variables. The tests were conducted using the Code 128 barcode symbology. The symbols were printed with an Eltron 2242 TLP direct thermal printer with a resolution of 203 dpi. An AccuSort Quad X omnidirectional laser scanner was used to read the symbols. For testing, symbols were mounted to a drum roller that was operated at speeds from 0 to 550 feet per minute (fpm). Symbols were presented to the scanner at a controlled rate and a fixed distance.
    The read rate served as the major dependent measure. Read rate is defined as the ratio of the number of successful reads to the total number of read attempts. Seven independent variables were tested, including symbol contrast, X dimension, bar growth/reduction, symbol orientation, aspect ratio, symbol quiet zone and line speed.
    A Box-Behnken experimental design for the seven variables with three levels for each variable was used. A total of 62 combinations of factor levels were used to test the significance of the independent variables and 2,500 samples were collected for each combination of factor levels.
    Each independent variable had three test levels. In other words, symbol quality for each variable was systematically varied upon printing. The levels included the standard or �suggested best� level for each factor; then symbols were printed one level below and one level above the suggested level. The suggested best levels were determined by criteria currently used by transportation companies. The low and high levels were determined based on our own knowledge of barcode print quality as well as equipment limitations.
    Defining Independent Variables
    Symbol contrast is defined as the difference between the maximum reflectance value and the minimum reflectance value of a symbol scan. Scanners require a high-symbol contrast to be able to distinguish bars from spaces. The levels for symbol contrast were achieved through an initial study of different label stocks from several manufacturers. Symbols were printed on 16 different label stocks and the symbol contrast was measured using a PSC barcode verifier. Three clusters of symbol contrast levels included a low symbol contrast level of 54, a high of 62 and a suggested standard level of 58. For the study, three different label stocks for each level were tested.
    X dimension is the intended width of the narrowest element in the symbol. At the time of the study, the scanner manufacturer recommended an X dimension of 15 mils for the sortation process. The printer resolution of 203 dpi prints an X dimension at three printer dots per narrow bar, which is equivalent to 14.8 mils. The low level for X dimension was set to be two printer dots, equivalent to an X dimension of 9.9 mils, while the high level was set at four printer dots, equivalent to an X dimension of 19.7 mils.
    Bar growth and reduction results from the bar width unintentionally becoming narrower or wider during the printing process, thereby either enlarging or reducing the width of the spaces between the bars. Bar growth can result from several conditions, including the printing method (e.g. ink spread from an inkjet printer), the speed of the label feed, the medium on which the symbol is printed or the printer�s overall condition (e.g. errors caused by misalignment or heat settings). For this study, bar growth was controlled using a barcode generating software package that allowed for adjustment (growth or reduction) of the bars by a single printer dot. The standard level for bar growth was a setting of zero. The low level reduced the bar width by one printer dot, while the high level increased the width by a single printer dot. The associated percentage change for the bars was calculated relative to the X dimension. For example, adding a single printer dot to a bar with an initial X dimension of four printer dots would increase it 25%, while increasing a bar with an initial X Dimension of 3 printer dots would be a proportional 33% increase.
    Symbol orientation is the angle at which the barcode symbol passes through the scan path of the scanner. The orientation levels were as follows: 0�, -22.5� and +45� with respect to the conveyor travel.
    Aspect ratio is the ratio of the height of the barcode symbol to the width of the symbol. For the shipping company for which this study was conducted, the barcode symbols are required to be a minimum 1.2 inches tall. The nominal X dimension is 15 mils and each barcode symbol was encoded with 22 characters, which would yield a barcode that is 2.5 inches wide with an X dimension of 14.8 mils. This would give a standard aspect ratio of 0.48. The standard level for aspect ratio was set at 0.50 while the low level was set at 0.375 and the high level was 0.625.
    Quiet zone is the white space to the left and right of the barcode symbol. The quiet zone aids the scanner in locating the symbol and adjusting the amount of light being reflected back from the label. If the quiet zone is too small, the scanner will not be able to adjust to the light reflected back from the label, and thus, will not be able to read the symbol. According to the ANSI X3.182 Barcode Print Quality Guideline, the minimum quiet zone that should be used for a barcode symbol is 10 times the X dimension of the symbol. The fixed mount scanner manufacturer has recommended that the minimum quiet zone be approximately three times the widest bar for the symbol. The widest bar for a Code 128 barcode symbol is four modules wide. A module is the narrowest nominal width unit of measure in a barcode symbol3. The standard level for the quiet zone was set at 12X, which is equivalent to three times the intended X dimension. The low level was set to 8X, while the high level was set to 16X.
    Line speed represents the speed of the conveyor. In most distribution environments, the line speed in each scan tunnel varies. Therefore, for this study it was difficult to set a standard level. It was arbitrarily decided that 500 fpm was a reasonable estimate of the standard line speed. All of the independent variable levels were chosen to be symmetric about the standard level. This symmetry, along with equipment limitations on upper level speed settings, led to the choice of the other line speed settings. The low level was set at 450 fpm, while the high level was set at 550 fpm, which is the maximum attainable speed for the drum roller used in this study. Line speed determines the amount of time a symbol is visible to a scanner pattern and, therefore, the number of scan attempts the reader can make.
    Eight labels were affixed to the drum roller in such a manner that only one label at a time would be within the field of view of the scanner. The distance between the label and the scanner was fixed at 34 inches throughout the test. The drum roller was then started at the predetermined speed defined by the experimental design and allowed to reach a steady state speed before scanning would begin (approximately 30 seconds). A total of 2500 scanner read attempts (i.e., the number of times a symbol was presented to the scanner) were electronically collected for each test condition.
    Dissecting the Data
    Using a logistical regression procedure, data was analyzed to determine the significant factors that impact read rate. Results of the logistical regression showed that read rate in a high-speed sortation environment is mostly affected by three factors: symbol contrast, the X dimension, and bar growth/reduction. Symbol orientation had an interactive effect with the X dimension. Factors that had little or no impact were aspect ratio, line speed and quiet zone.
    The data indicate that the read rate for a symbol contrast value less than 58 cannot reliably be read, while symbols at the higher levels produce higher read rates. In general, we found that a symbol with a symbol contrast higher than 58 produces higher read rates (typically greater than or equal to 98% if X dimension and bar growth are controlled). At first glance, this finding may appear to be in stark contrast to the ANSI print quality criteria for symbol contrast. According to the specification, symbol contrast receives a failing grade with a value less than 20. However, the reader must take into consideration that this ANSI guideline must address all applications in general. The results of the current investigation suggest that for the high-speed automated sortation process, a much higher value is required to achieve reliable read rates.
    As expected, the results also show increasing the X dimension increases the read rate of the symbol. In the current study, the 9.9 mil symbols could not be read with the available scanner. If other factors are tightly controlled, symbols with a 14.8 X dimension can be reliably scanned. However, as can be seen from Figure 2, if other factors are left uncontrolled, even a 14.8 mil symbol will be unreliable. The data also suggests that through increasing the X dimension to 19.7 mils, problems caused by other poor print quality factors can be overcome. If a label has space to increase the X dimension to 19.7, we recommend this action be taken.
    Reducing the bars by a single printer dot negatively impacts the read rate of the symbol. In contrast, growing the bars by a single dot resulted in higher read rates. This statement can only be made for symbols with 14.8 X dimensions. A single dot subtraction or addition to a 14.8 mil symbol results in a 33% change in the size of the bar. When we tested the impact of bar growth/reduction on 19.7 size symbols, we found there was no significant change in read rate when a single dot was added to or subtracted from the bar.
    Orientation impacted the results by making it more difficult for the scanner to obtain a complete scan pass over the entire symbol. This condition was especially apparent as the X dimension of the symbol decreased. As expected, the greater the deviation of the symbol from 0 degrees, the lower the read rate. In general, test results showed the best read rates when the symbol was oriented at 0 degrees and had a 14.8 or 19.7 mil X dimension. Good read rates can also be achieved when the symbol is oriented at a maximum of 22.5 degrees if the X dimension is 19.7 mils.
    As stated above, the results indicate that neither aspect ratio, quiet zone nor line speed significantly impacted the read rate of the symbol. Early in the study, it was hypothesized that the scanner would be able to scan symbols with more success at slower speeds than at faster speeds. In reality, this assumption appeared to be false with the scanner performing equally well for all line speeds.
    Label Recommendations
    One goal of this study was to be able to make recommendations that would increase the probability that a package would be successfully sorted in the high-speed sortation environment. For this to be possible, the read rate should be as close to 100% as possible. The recommendations for printing labels were based on models that were produced in this study as well as some real-world limitations.
    Richard E. Billo, PhD is department head of Oregon State University�s Industrial and Manufacturing Engineering Department. Dr. Billo also serves as director of the University�s Automatic Identification and Data Capture (AIDC) Research Center. Stephen J. Brown is an MS student at the University of Pittsburgh and an engineer for FedEx Ground. He completed this investigative study on the impact of barcode print quality factors as part of the MS degree requirements at the University of Pittsburgh. Mainak Mazumdar, PhD is professor of Industrial Engineering at the University of Pittsburgh and a renowned leader in reliability, statistics and energy distribution systems. To see the article in its entirety, refer to our Web site at www.psdmag.com in the �Members Only� section.