Id Systems
Automatic Feeding Bar Code Readers
At some point in your bar code assisted application, perhaps when you start reading thousands, or tens of thousands, bar codes a month, you realize you have too many people using bar code wands, or even laser scanners, to get your data into your application. Yes, it's faster, and definitely more accurate, than traditional keyboard data entry. But there has to be a better way.
Or perhaps you are avoiding implementing bar code assisted data entry because the volume already appears to be too great for manual scanning methods.
Who are you going to call? Not the Ghostbusters. But you could enlist the aid of some relatively new devices that automate much of the bar code scanning process.
Enter Automatic Feeding Bar Code Readers
In the past few years a group of automatic feeding bar code scanners have come onto the market at surprising low cost. With prices in the $7,000 to $15,000 range, they can offer a return on investment from a couple of months to less than a year in labor savings alone.
They all work by taking a stack of small slips of paper (also called documents or tickets) and feeding them one at a time past a wand type or laser bar code reader.
And they do it fast, reading from 100 to 600 documents per minute. Running 24 hours a day, 7 days a week, that's over 300 million a year. Of course no machine of this type can operate all the time, if for no other reason than the operator needs to feed document stacks.
There appear to be almost as many uses for these units are there are people using them. Aside from the obvious bill payment processing, these units are being used by transit authorities, parking structures, amusement parks, casinos, a lottery system, manufacturing and distribution applications.
Not the Only Choices
Lest you think there are only three manufacturers who make units capable of automatically scanning bar coded sheets (and this magazine receive a number of angry calls from other manufacturers), let me point out that several other manufacturers have units that can be used in this application. However, in each case they were designed for another purpose such a optical mark reading (OMR), or optical character recognition (OCR), applications and offer bar code reading as an option to that other primary application.
So if you have a need for OMR, OCR, or other application that could involves the technology of one of these other units, then you should definitely look into bar code options for them. The incremental cost of adding bar code reading capabilities is low compared to buying a dedicated bar code reader in addition to one of these other units.
But Not All My Documents Can Be Automated!
One of the common reasons given for not automating the collection of data via these systems is that many applications have exceptional conditions that can't be handled with a pre-printed bar code receipt.
For example, consider the case of a utility recording customer payments. Not everyone pays their utility bill in full every month. Some people can't afford to pay all the bill, so they make a partial payment. Other people mistakenly write their check for a few cents or dollars more than what is due.
So do those exceptions make automated bar coded data entry unsuitable for processing utility payments? Yes, if you take the view that every customer's payment must be processed by the automated system.
But if you look at the fact that the great majority of the customers pay their bill in full each and every month, then utilizing an automated system such as these for that majority and separating the exceptions for manual processing makes a lot of sense.
That's exactly what the Loyal Order of the Moose, a fraternal organization, did with their membership renewals. Since 90% of their members renew each year for one year, their payment stubs contain a bar code with their membership number on it. Those stubs are simply scanned into an application using a Duplo reader. The balance are processed semi-manually--the membership number is still scanned with a wand reader, and the renewal information keyed by hand. They estimate they have eliminated the need for a second shift in processing the approximately 800,000 renewals each year that have been converted to this system.
So if you feel that your automated system has to be able to handle all possible exceptions, stick with your old way of collecting data. But if you can see how a vast majority of your data can be handled automatically, let one of these units handle the routine work and process exceptions manually.
Equivalent Bar Code Reading Capabilities
Interestingly, the units' ability to read bar codes is really a sine qua non, or an assumed capability. They all use an OEM bar code reading module to do the actual scanning and decoding, focusing their efforts on the paper handling aspects. In our evaluation, we were unable to generate bar codes that could be read by one unit and not another (we did generate some bar codes that couldn't be read by any unit, or any other bar code scanner).
There are some slight differences in the symbologies read by each unit. All the units read Code 39, Interleaved 2 of 5 and UPC/EAN/JAN. Both the Brandt and Duplo units also read Codabar. The Brandt alone reads Code 128. The Duplo adds regular and matrix 2 of 5 to the Interleaved 2 of 5. And the Kowa specification sheet indicates it reads KW-7 symbology, a symbology I have never heard of.
Finally, the Brandt 3120, all the Duplo and Kowa units utilize autodescriminating decoders that will accept variable length data, while the Brandt 3140/3141 decoder requires the user to select the bar code symbology and data length during setup. This selection does give the Brandt 3140/3141 an additional error check capability at the expense of some flexibility in the kind of documents that can be scanned. However, both the Duplo and Kowa units can be configured via DIP switches to disable symbologies not being used and increase their error detection capability.
While the need to select symbology and data length may at first appear to be a limitation, remember that in most of the applications using these devices, the user is also the bar code supplier and thus has total control over what information is printed in the bar code and how.
Pickets and Ladders
Where there is a significant difference in the various units' bar code reading ability is the underlying scanning technology they picked for the reader. The units evaluated use one of two technologies: a single beam CCD based wand or a moving beam laser scanner.
The wand scanners read the bar code off the document in what is called ladder reading orientation. It's called ladder reading because the bars of bar code look like the rungs of a ladder to the scanner. The bar code is perpendicular to the axis of paper motion and the movement of the document itself is responsible for the scanning action. An illustration of ladder reading is shown in figure 1.
The laser scanners, on the other hand, read the bar code in what is called picket reading. When a scanning beam traverses the bar code, it the bars look much like a picket fence. In this type of reading, the movement of the document is parallel to the bar code and the scanning action is supplied by the laser. Figure 2 shows picket reading by a laser scanner.
Each method has its advantages and disadvantages and care should be exercised in matching the scanning method to the application.
In ladder reading, the orientation of the document in the scanner is often important unless by chance it happens to be along the center of the document. That's because the scanning beam travels across a fixed location of the document and if the document is rotated 180 degrees, the beam won't scan the bar coded part of the document. So while the scanners are capable of reading the bar code in either orientation, the fact that the position of the bar code usually isn't along the centerline makes scanning impossible.
In ladder reading, the bar code height can often be much shorter than in picket reading, allowing bar codes with relatively low aspect ratios. Because the document is being pulled past a fixed beam, once the beam is carefully adjusted to the centerline of the bar code, it will travel across the bar code every time. The only limitation on the height of the bar code is the readers ability to consistently hold the document in registration as it passes by the scanner. If the unit allows the document to skew or shift from side to side, then its possible the scanning beam will miss all or part of the bar code resulting in a bad read. Registration errors can be overcome somewhat with higher bar code heights.
And in ladder reading the bar code length can be much longer because there is no theoretical limit on the scanner whereas with a laser scanner the width of the scanning beam limits the length of the bar code that can be scanned.
Finally, in theory a ladder reader is limited only by the speed of the scanner decoder. While scanner decoder speeds aren't documented, one would expect them to be extremely fast and not a limit. However, it is interesting to note that the ladder reading units were among the slowest although I suspect that's because they also were the least expensive and may be limited more by their paper handling than the scanning speeds.
Why would a user want a laser picket scanner over a fixed beam ladder scanner? First and foremost, the laser scanners were all able to read documents whether they were rotated or not. While it is possible that a bar code could be printed far enough to one side that when turned upside down it was outside the laser scanning beam, the lasers are less sensitive to position than ladder readers.
That becomes an important concern because it determines the amount, and level, of operator interaction needed to read the bar codes. Since the goal of these units is to minimize operator involvement, any step that can be removed, such as orienting the documents, will save on-going operating costs.
One real limit on the laser picket readers is the height of the bar code relative to the speed of the document handling. Figure 2 shows that at slower speeds, the laser beam makes several passes over the bar code as it moves underneath. But as the document moves faster and faster, as shown in figure 3, the laser can make fewer complete passes over the bar code until finally a complete scan can't be made.
Most of the laser scanners by default depend on several scans of the bar code to generate a reliable read by self checking the results. While that feature can be turned off, it reduces the level of error checking the unit can perform.
If the height of the bar code can be increased, then the document can be handled faster since there is more chance of the bar code being scanned multiple times.
We noticed that with some deliberately under-height bar codes (only 1/6" high), that the laser based scanners rejected more documents than the ladder readers. Generally many of those documents would read on a second try, indicating the timing of the document passing through the beam was affecting the read rate.
Finally, with long, narrow bar codes, any skew introduced by the document feeding mechanism may affect the unit's ability to read a particular document as shown in figure 4.
Paper Handling is Key
Having concluded that the units were all equally capable in their ability to read any given bar code, we felt the real differences in selecting one of these automated bar code readers would be their ability to handle paper. And indeed each unit uses a different technique to separate the stacks of paper into individual sheets for reading.
Interestingly however, we were unable to generate any test cases that could cause any of the units to consistently fail despite the expertise provided by a large printing company in paper and paper handling. The units were able to process coated and uncoated stocks in a variety of paper thicknesses with generally equal capability. Yes, from time to time we would have to be more critical adjusting the units, but they all could be made to reliably read everything we tried.
This observation was borne out talking to users of these units who reported no troubles with paper mis-feeding during use.
One recurring theme talking to vendors and users alike is to involve your printing company in selecting the paper for the documents to be read. Often imperceptible changes in the paper can make differences in the paper feeding characteristics when read by automated bar code readers.
Another document design that should be emphasized is the desirability of having as tall of bar codes as possible within the constraints of the document format. The taller the bar codes, the less susceptible the document will be to feeding induced reading errors and the higher the first read rates will be.
So what is different about each unit's paper feeding system? The differences lie in two areas: how the unit separates individual documents from the stack, and what happens with mis-feeds or mis-reads.
All the units except the Brandt 3120 and 3140 readers have two output trays, one for documents read correctly, the other for documents that have mis-fed or been mis-read. The Brandt 3120 and 3140 units stop when they detect either a mis-feed or mis-read. This is useful when the order of the documents being read contains important information. One example is a distribution application where a picker document is followed by a set of item documents and it is imperative the sets stay together with the picker document to be properly coded. The Duplo unit can also be configured to stop on mis-feed or mis-read, although it is not documented in the manual.
The Kowa unit uses two sets of rollers just like many fax machines you may be familiar with. Each set of rollers is soft rubber, giving them a tacky ability to pull individual documents off the stack. One set of rollers is used to feed the documents, the other to position the stack. Unlike most fax machines, the Kowa rollers are adjustable to allow a broader range of document thickness.
The Duplo unit is slightly different, using a belt to feed and position the stack and a set of feed rollers to pick individual documents off the stack one at a time. There are several adjustments to give the operator considerable control over the feed roller tension and stack positioning.
Both the Kowa and Duplo units sense the length of the document and its opacity. This lets them detect when two documents are fed simultaneously or when two documents are overlapping each other. In both cases, the documents are rejected. The Brandt units all use just an opacity, or density, check for double document feeding.
While the Kowa and Duplo units pull individual documents off the "top" of the stack, the Brandt units pull individual documents off the bottom of the stack. This is accomplished by a cam driven set of "pickers" that position documents to be "grabbed" by two small stripper/feed wheel assemblies that pull the document into the unit. My paper feeding experts felt this method was less prone to mis-feeds than the other techniques, although we were unable to prove or disprove the feeling in our testing. In addition, feeding from the bottom makes it easier for the operator to keep adding documents while the unit is operating.
Expect Some Difficulty During Installation
All the units evaluated interface to the host computer via RS-232 serial communications. Unfortunately, while the signals for RS-232 are standardized, especially for terminal/modem connections, when used with many computer applications, the choice of connector size (DB-25 or DB-9), sex (male or female), and whether a a unit is DCE or DTE aren't standardized and left open to the vendor. In addition, there isn't a really good vocabulary available to make documenting the connections straightforward.
As a result, connecting any RS-232 device to a computer is somewhat a black art. It is usually done with a level of educated guessing, coupled with an inexpensive break-out box that lets the installer watch the control signals, and a handful of adapters. After a few tries, the experienced installer can generally get a unit communicating and then select or build the proper cable.
In addition, the Duplo and Brandt units used RS-232 control signals to control the feed motor so the unit couldn't be used until the computer is ready to receive the data and to keep the unit from sending data faster than the computer can process it. But neither one documented in its manual what signal was being used, how, or that they were being used at all. This probably is a reflection of the relative immaturity of the market that will go away as more units are sold.
The manuals in general were minimalistic. Some were as short as 10 xeroxed pages. In general, they would cover the various controls, but they offered little about interfacing or adjusting the units' operation. Given the price range of the equipment, it is surprising how little printed information the vendors supplied. This may lead to a cult of experienced installing dealers. The exception to this observation was the manual for the Brandt 3140/3141 units which included comprehensive operator and routine maintenance instructions.
Plan on needing help from the vendor while you are setting up your system. The physical aspects were easy enough, but every unit evaluated required a call to the vendor to clarify some point in the interfacing to our PC based computer system. Fortunately, all the people I talked to were knowledgeable enough to answer most of the questions I had.
But don't let any difficulty getting the unit interfaced to your computer scare you away from these useful devices. Remember, you only have to connect them once and they will continue to operate with little or no additional effort on your part. So like the purchase, the effort spent interfacing the unit is an investment in the productivity you will gain from the reader.
Finally to get the best possible paper feeding from these units plan on carefully adjusting them during set-up. While most are relatively tolerant of not being adjusted exactly, they do tend to mis-feed more often than when they are tuned up optimally.
Maintenance Issues
All these units are mechanical devices with many moving parts. Any time you have moving parts you are going to require service at some point in the operation of the unit. While none of the users I talked to required any significant service in the average year of use, most had one service call during the first year. Brandt recommends an on-site service contract.
When considering which unit to purchase, inquire about where the closest service agent is. While all the manufacturers have a nationwide dealer service network, the coverage of particular areas differs. In some cases, the manufacturer uses part of their other products service network. In those cases, the person providing your service may never have seen your particular unit before. So while they may be qualified to work on the mechanical portion, they may be of little help during installation.
In addition, all of the units require some form of routine maintenance such as cleaning the feed wheels and strippers that the user can perform. But this maintenance is important to ensure the units will continue to reliably feed the documents with fewest mis-feeds.
Fast Readers for Volume Applications
If you are reading tens of thousands, hundreds of thousands or even millions of bar coded documents a year, these automatic feed bar code readers can save you considerable labor costs. And if you are still relying on manual data entry with such volumes, you should re-evaluate whether the way you are operating is as cost effective as using one of these units for the routine transactions and handling exceptions manually. No matter which unit you pick, you will find your data entry operations significantly more productive and cost efficient.
-----------
I would like to thank EagleDirect, a large marketing and printing firm in the Denver area, for its expertise in paper handling and for supplying some sample applications for use during the evaluation of these units.
Specifications
Mfg. Model Cost Reader # Barcodes Minimum Maximum Paper Slips
per pass document document Thickness per min
Brandt 3120 $6,995 CCD/Ladder 1 2" x 4" 4.25" x 7.25" .003" - .007" 500
Brandt 3140 $13,225 Laser/Picket 1* 2" x 4" 4.25" x 9.5" .003" - .010" 600
Brandt 3141 $15,495 Laser/Picket 1* 2" x 4" 4.25" x 9.5" .003" - .010" 600
Duplo NB-80E $6,995 CCD/Ladder 1 2" x 2.5" 4.75" x 8" .004" - .08" 100
Duplo NB-100 $9,995 CCD/Ladder 1 2" x 5.5" 5" x 9" .002" - .08" 100
Duplo NB-180E $12,995 CCD/Ladder 2 2" x 2.5" 4.75" x 8" .004" - .08" 100
Duplo NB-200S $11,995 CCD/Ladder 1 2.25" x 5.5" 5" x 9" .002" - .08" 100
Duplo NB-200 $12,995 CCD/Ladder 2 2.25" x 5.5" 5" x 9" .002" - .08" 100
Kowa LB130 $12,900 Laser/Picket 3 4" x 2.75" 11.8" x 8.25" .002" - .008" 400
* two bar codes possible with firmware change
Vendors
Brandt
1750 Woodhaven Dr.
Bensalem, PA 19020
(215) 638-3600
FAX (215) 638-1188
Duplo USA Corporation
17481 Mr. Cliffwood Circle
Fountain Valley, CA 92708
(714) 540-8222
FAX (714) 540-9139
KOWA Optimed Inc.
20001 S. Vermont Ave.
Torrance, CA 90502
(213) 327-1913
FAX (310) 327-4177
Related Vendors
National Computer Systems
4401 West 76th Street
Minneapolis, MN 55440
(612) 830-7600
Optowand Incorporated
P. O. Box 166558
Irving, TX 75016
(214) 257-2100
FAX (214) 257-2129
Scantron
1361 Valencia Ave
Tustin, CA 92680
(714) 259-8887
FAX (714) 259-1177