We have a client that we have been working with for quite a long time that has been seeing an alarming high rate of barcode misreads on their scanning stations.  They currently scan thousands of documents a day and each and every one of them is separated by what we refer to as a "Batch Header Sheet". You will find a few samples attached to this post as "fig1 Redacted.pdf".  During the scan process into Quick Fields it is checked against 6 document classifications, once classified, it runs a process on the Batch Header sheet to pull information from up to three more barcodes, uses that information to perform a Realtime look-up against a SQL database to fill more indexing information, which is then used to sort and name the document.  

The issue occurs during the barcode reads, many times the barcode is not picked up which of course causes more issues down stream.  The issue started occurring when we moved the client from Laserfiche 7 to 8.  During that upgrade they also placed new scanning stations and Kodak scanners.  Their current scanning stations are Dell Pcs with dual core 3.2Ghz i5 CPUs, 4GB memory, 64bit Windows 7, and the scanners are Kodak i1420s and i4600s.

So here is the process that we took to test this issue and the results which seem to show two patterns.

Process 1.  I used their machine named CoCd10765 and confirmed all scanner settings, firmware versions, and LF versions. This PC has a Kodak i4600 attached directly to it and is linked into Quick Fields via Scan Connect and ISIS driver.  The staff gave me a sample set of documents which included 95 pages and 20 documents all divided by batch header sheets.  When I scanned the documents twice stopping between each set I ended up with the following:

Process 1 results:
40 documents with the possibility of 4 barcode reads per document.
Total possible barcode reads: 4x40 = 160
Total actual misreads: 12
Percentage of misreads: 7.5%

Ran this test a few more times and each time it might misread a different barcode here and there but many misreads stayed the same from test to test.

Process 2. I used their machine named CoCd10765 and confirmed LF versions.  This PC has a Kodak i4600 attached directly to it.  I used Kodak's Scan Validation Tool (SVT) to scan the same 20 documents, at the same DPI, ithresholding, and aggressive cropping settings, in the same order they were scanned earlier.  I then used Universal Capture as my scan source for Quick Fields.  When I scanned the documents twice stopping between each set I ended up with the following:

Process 2 results:
40 documents with the possibility of 4 barcode reads per document.
Total possible barcode reads: 4x40 = 160
Total actual misreads: 6
Percentage of misreads: 3.75%

Ran this test a few more times and each time the misreads stayed the same from test to test.

Process 3. This time I scanned them into a Kodak i1420 scanner attached to another machine.  I used Kodak's Scan Validation Tool (SVT) to scan the same 20 documents, at the same DPI, ithresholding, and aggressive cropping settings, in the same order they were scanned earlier.  I then moved those .tif files to their machine named CoCd10765.  I then used Universal Capture as my scan source for Quick Fields.  When I scanned the documents twice stopping between each set. I ended up with the following:

Process 3 results:
40 documents with the possibility of 4 barcode reads per document.
Total possible barcode reads: 4x40 = 160
Total actual misreads: 0
Percentage of misreads: 0%

I than ran this test over and over again until I had scanned 300 documents and my misread rate with still 0%.

This made me believe that it was an issue with the Kodak i4600 scanners so I began talking to the project manager from their IT group.  They expressed that the i1420 was misreading also but not as much as the i4600.  They attributed this pattern to the fact that the i4600s were used more often than the i1420.  Taking this extra piece of information into account we started looking very deeply at the Batch Headers that I was using verses the misread logs they captured for each scanning station.  We noticed another pattern, the longer/wider the barcode is the more it misread across any of the scanners.  So we did the following process to continue testing.

Process 4.  Took 100 new documents with much wider barcodes and used Kodak's Scan Validation Tool (SVT) to scan them into the Kodak i1420 scanner attached to another machine. Again at the same DPI, ithresholding, and aggressive cropping settings as they were used earlier.  Moved those .tif files to their machine named CoCd10765.  Used Universal Capture as my scan source for Quick Fields.  When I scanned the documents I ended up with the following:

Process 4 results:
100 documents with the possibility of 4 barcode reads per document.
Total possible barcode reads: 4x100 = 400
Total actual misreads: 4
Percentage of misreads: 1%

Additional notes:  
-During testing the documents were always scanned in the same order.
-During earlier testing we tried different manufactures of paper for the Batch Header sheets – No noticeable difference.
-During earlier testing we tried different printers – No noticeable difference.

Updates:
Since running these tests misreads have gotten worse.  I tried adding image enhancements, smooth and despeckle to the barcode process in the Quick Fields session.  This actually made the misreads worse.

The two patterns I am seeing:
-The misreads are more frequent on the Kodak i4600s.
-Barcodes that are longer/wider have a better chance of misreading.

At this point we are reaching out to Kodak and of course Laserfiche to get any help possible to correct this issue.

One thing that we thought of was the type of barcode that is used on the batch header sheets.  It is barcode 39 and is a font that is used to create the batch header sheets.  A Kodak rep explained that there are much better barcodes to use, and also each printer could space the barcode slightly differently because it is a font and not a rasterized image during the print.

So what is the best barcode to use?  I don't believe that is really the issue here, there is something else going on, but a test that I would like to try next is to use a different/better barcode to see if the misreads drop.

Other specs:
Quickfields: 9.0.0.460
Barcode: 9.1.0.0
Scan Connect 8.3

Kodak i4600: 1.1.11212.10001