This is a copy of the presentation given by Ellen Fleurbaay and Marc Holtman of the Amsterdam City Archives at the the MARAC Plenary Session in Jersey City on Friday October 30, 2009.
20. You ask Scanning on customer’s request, economic principles, technical issues and work process We Scan MARAC Conference October 30 2009
21. You ask We Scan MARAC Conference October 30 2009 We Store We Do Scanning on customer’s request, economic principles Image quality and workflow principles Compression and filesize Workflow, tools and practical issues
22. Q. How long does it take to scan it all? MARAC Conference October 30 2009 1 feet = 2.000 scans Production = 10.000 scans a week A. 406 years Will this be our ultimate solution? Q. How many scans can be made from 20 miles of archives? A. 739.200.001 scans
23. The user doesn’t commit to anything by placing a request, but neither does the archive You ask We let our users set priorities in digitization In principle all requests are honored, unless We speak of a request for digitization and not of an order MARAC Conference October 30 2009 1. Scanning at customer’s request It can not be digitized for material reasons Copyright material Disclosure restrictions apply All archive files can be requested for digitization via the online the finding aids
24. Costs for purchasing scans are equal for all users (the more you buy, the cheaper it gets) Scans available are integrated in the online finding aids All scans made are available for all users The requester is not obliged to buy all scans MARAC Conference October 30 2009 You ask 1. Scanning at customer’s request
25. Customers think a low price is important This means that costs for producing and storing scans have to be as low as possible Archival research easily runs into the use of dozens to hundreds of documents You ask The price of an ordinary copy in our reading room should be the benchmark MARAC Conference October 30 2009 2. Low costs 100 scans should not cost $ 100 The costs when purchasing scans online should be competitive with travel costs when visiting our reading room
26. This asks for a streamlined, efficiently organized work process You ask Digitization takes time, but research should not have to be planned weeks ahead Delivery time in a scanning on request service should be as short as possible MARAC Conference October 30 2009 3. Fast delivery Aim is a delivery time of 2 – 3 weeks
27. An efficiently organized work process Low incidental and structural costs You ask MARAC Conference October 30 2009 Conclusion If we can make sure that All finding aids can be selected for digitization by users The scans are delivered in short time For low costs it can be stated that we have no backlog in digitizing and the objective that the customer is able to consult digitized item has been achieved We need:
28. Besides scanning on request projects are based on: In this presentation the focus is on large scale digitization at customer’s request We scan However, scanning on request is only a part of all digitization that takes place in the archives MARAC Conference October 30 2009 Digitization at the Amsterdam City Archives in general Grant money (often on specific topics, like WWII) Selections of photographs, drawings etc for the Imagebank (Beeldbank) Cooperation with Amsterdam district councils and services
29. Goals of digitization projects vary from access to substitution of the originals In every project quality standard and method are set, depending on purpose and type of material For all projects we have one workflow We always work on project basis We scan MARAC Conference October 30 2009 Digitization at the Amsterdam City Archives in general
30. Experience shows that a constant production of 10.000 scans (at cutomer’s request) each week is achievable This way tasks can be planned best and deployment of staff is most efficient We scan 1. At large scale the more scans being made, the lower the price per scan Large scale production is a prerequisite in order to keep production costs as low as possible MARAC Conference October 30 2009 2. With a constant production Large scale production can only be organized effectively when constant production is assumed
31. Documents that are being digitized in this reproduction process can have the following forms We scan MARAC Conference October 30 2009 Small and large size Bound and loose-leafed entities Card indexes Old and modern material Low and high contrast documents Text alone, text and image together Hybrid forms 3. A broad spectrum of document types
32. Costs for producing and storing scans are determined to a high extent by the quality standard set for the scans Purpose of the scans: archival research using the web, straight from screen or print We scan 4. For archival research from screen or print The higher the standard of quality, the higher the costs will be In order to keep costs low it is prudent to allow the standard of quality follow from the requirement the end user places on the scan Textual information legible in de originals must be legible in the scans MARAC Conference October 30 2009
33. But has no added value for the customer at all A quality higher than that inevitably will push up both incidental and structural costs We scan 4. For archival research from screen or print Specified (basic) quality standard: MARAC Conference October 30 2009 Reproduction of all significant information Reproduction of details which are not part of the textual information is not required
34. We scan MARAC Conference October 30 2009 Scan quality and legibility High quality scan Modified scan (contrast) Optimal tonal range Example: very “light” original Excellent flexibility Poor tonal range Little flexibility Experience in practice learns that what is experienced as being “good legibility” is very personal. We decided to solve this problem with a smart filter in the document viewer. Poor legibility Excellent legibility Which one would you buy?
35. Skimming on the quality of scans (it can be better) is purely an economic decision, not one taken on principle We scan MARAC Conference October 30 2009 4. For archival research from screen or print It does make sense to let the standard of quality follow from the purpose the end-uses places on of the scans Price rates scanning, external partner 0,05 $ Legibility, auto-feed 0,30 – 0,75 $ Legibility 3 – 10 $ High-end Price comparison scanning costs
36. This way damage or loss of the originals is ruled out After digitization the originals can not be requested in the reading room anymore We scan 5. For conservation and security The scans in the scanning on request service are made for the purpose of access / archival research Not as a substitute for the originals Nevertheless, digitization does have a real conservation function MARAC Conference October 30 2009 Conservation of the originals remains the major concern
37. A file can contain one – hundreds of documents We scan By definition the entire file is scanned Never just a selection of pages There are a few reasons for this: MARAC Conference October 30 2009 6. Always complete files The costs for scanning are not so much a factor of quantity, but rather of the manual processing involving in it In the originals or the metadata it has to be indicated which documents are being digitized When shown in the Archiefbank, the user expects completeness When non-scanned pages have to be digitized later, the entire preparation process has to be gone through once again
38. Contracting out of scanning was a logical choice We scan The in-house scan facilities are not designed for large-scale digitizing The complexity of the workflow and material to be scanned calls for Investing only makes sense by very high production, organized on a large scale MARAC Conference October 30 2009 7. Contracting out the scanning to external partners Specialized hard- and software Specialized set-ups Knowledge Very complex technical infrastructure
39. This calls for intensive collaboration Also, the workflows of archive and digitizer have to dovetail We scan There are many scanning companies Most do have experience in bulk processing But not in this degree of complexity and diversity MARAC Conference October 30 2009 7. Contracting out scanning is more than awarding a contract to a supplier Contracting out the scanning to external partners
40. We use a combination of 1 and 3 We store Storage costs still are considerably high when producing large quantities of scans In order to bring structural costs down file size of the scans has to be as low as possible This can be achieved in three ways MARAC Conference October 30 2009 Scans with a file size as small as possible 1. Skimming on resolution 3. Using (lossless or lossy) compression on the files 2. Skimming on bit depth / amount of colors (only possible in formats like TIFF and PNG)
41. We store Resolution, compression and legibility: an example MARAC Conference October 30 2009 300 dpi, high quility JPEG 200 dpi, low quility JPEG Scans with a file size as small as possible
42. We store Storage of compressed files as master images was “not done” The main arguments where Research after these arguments learned: MARAC Conference October 30 2009 When using lossy compression you’ll loose information Compressed files are more vulnerable (preservation) Even when using strong lossy compression legibility is still guaranteed Compressed files are not more vulnerable to loss then uncompressed files But no compression means: large files high storage costs Storage of uncompressed files is not necessary Scans with a file size as small as possible
43. MARAC Conference October 30 2009 Comparison between file format, compression, resolution and file size Scans with a file size as small as possible We store 55% 6 Tb 12 MB 24 bits 300 dpi Lossless Part 1 JPEG2000 0,5% 59 Gb 120 Kb 24 bits 300 dpi Lossy Part 6 34% 3,7 Tb 7,5 Mb 24 bits 300 dpi Lossy Qua (ps) 12 JPEG 10% 1,1 Tb 2,1 Mb 24 bits 300 dpi Lossy Qua (ps) 10 1,1% 124 Gb 255 Kb 24 bits 200 dpi Lossy Qua (ps) 4 Filesize 3,3 Mb 22,1 Mb Avg Lossy --- Type 15% 100% % 1,6 Tb 11 Tb 500.000 400 dpi 300 dpi Resolution Qua (ps) 10 No Compression 24 bits 24 bits TIFF Color Format
44. TIFF uncompressed MARAC Conference October 30 2009 Comparison between file format, compression, resolution and file size Scans with a file size as small as possible We store 55% 6 Tb 12 MB 24 bits 300 dpi Lossless Part 1 JPEG2000 0,5% 59 Gb 120 Kb 24 bits 300 dpi Lossy Part 6 34% 3,7 Tb 7,5 Mb 24 bits 300 dpi Lossy Qua (ps) 12 JPEG 10% 1,1 Tb 2,1 Mb 24 bits 300 dpi Lossy Qua (ps) 10 1,1% 124 Gb 255 Kb 24 bits 200 dpi Lossy Qua (ps) 4 Filesize 3,3 Mb 22,1 Mb Avg Lossy --- Type 15% 100% % 1,6 Tb 11 Tb 500.000 400 dpi 300 dpi Resolution Qua (ps) 10 No Compression 24 bits 24 bits TIFF Color Format
45. JPEG (psd) 10 MARAC Conference October 30 2009 Comparison between file format, compression, resolution and file size Scans with a file size as small as possible We store 55% 6 Tb 12 MB 24 bits 300 dpi Lossless Part 1 JPEG2000 0,5% 59 Gb 120 Kb 24 bits 300 dpi Lossy Part 6 34% 3,7 Tb 7,5 Mb 24 bits 300 dpi Lossy Qua (ps) 12 JPEG 10% 1,1 Tb 2,1 Mb 24 bits 300 dpi Lossy Qua (ps) 10 1,1% 124 Gb 255 Kb 24 bits 200 dpi Lossy Qua (ps) 4 Filesize 3,3 Mb 22,1 Mb Avg Lossy --- Type 15% 100% % 1,6 Tb 11 Tb 500.000 400 dpi 300 dpi Resolution Qua (ps) 10 No Compression 24 bits 24 bits TIFF Color Format
46. JPEG (psd) 4 MARAC Conference October 30 2009 Comparison between file format, compression, resolution and file size Scans with a file size as small as possible We store 55% 6 Tb 12 MB 24 bits 300 dpi Lossless Part 1 JPEG2000 0,5% 59 Gb 120 Kb 24 bits 300 dpi Lossy Part 6 34% 3,7 Tb 7,5 Mb 24 bits 300 dpi Lossy Qua (ps) 12 JPEG 10% 1,1 Tb 2,1 Mb 24 bits 300 dpi Lossy Qua (ps) 10 1,1% 124 Gb 255 Kb 24 bits 200 dpi Lossy Qua (ps) 4 Filesize 3,3 Mb 22,1 Mb Avg Lossy --- Type 15% 100% % 1,6 Tb 11 Tb 500.000 400 dpi 300 dpi Resolution Qua (ps) 10 No Compression 24 bits 24 bits TIFF Color Format
47. JPEG2000 lossless MARAC Conference October 30 2009 Comparison between file format, compression, resolution and file size Scans with a file size as small as possible We store 55% 6 Tb 12 MB 24 bits 300 dpi Lossless Part 1 JPEG2000 0,5% 59 Gb 120 Kb 24 bits 300 dpi Lossy Part 6 34% 3,7 Tb 7,5 Mb 24 bits 300 dpi Lossy Qua (ps) 12 JPEG 10% 1,1 Tb 2,1 Mb 24 bits 300 dpi Lossy Qua (ps) 10 1,1% 124 Gb 255 Kb 24 bits 200 dpi Lossy Qua (ps) 4 Filesize 3,3 Mb 22,1 Mb Avg Lossy --- Type 15% 100% % 1,6 Tb 11 Tb 500.000 400 dpi 300 dpi Resolution Qua (ps) 10 No Compression 24 bits 24 bits TIFF Color Format
48. We store Comparison storage costs MARAC Conference October 30 2009 Storage of 500.000 images Avg size per scan uncompressed = 22,1 MB Price rate : 1 TB, storage in a controlled e-repository environment on two separate locations, including IT costs $ 7.000 (NLD, nov 2009) Scans with a file size as small as possible (File)size still does matter! $ 420.000 $ 8.680 $ 77.000 $ 770.000 Costs 10 years $ 42.000 $ 868 $ 7.700 $ 77.000 Costs 1 year 6 TB 124 GB 1,1 TB 11 TB Storage JPEG 2000 (part 1, ll) JPEG 4 (200 dpi) JPEG 10 Tiff uncompressed Fileformat
49. Projects with different goals, document types and partners take place at the same time A streamlined, standardized process is indispensable when digitizing on a large scale Guidelines and best practices often take no account of these complex factors and the amount of scans to be produced We developed a process in which large scale and flexibility are starting points All digitization projects follow this process Developing the reproduction process MARAC Conference October 30 2009 We Do
50. We developed a simple, but effective workflow application in-house This asks for workflow management with a user-friendly application For all projects, at any moment, it has to be clear: We Do MARAC Conference October 30 2009 What the current status is of each to digitize unit Where each unit can be located What current and succeeding tasks are to be performed on each unit Developing the reproduction process
51. In the following slides we focus on the weekly production of 10.000 scans in the digitizing on request service We developed a simple, but effective workflow application in-house This asks for workflow management with a user-friendly application For all projects, at any moment, it has to be clear: We Do MARAC Conference October 30 2009 What the current status is of each to be digitized unit Where each unit can be located What current and succeeding tasks are to be performed on each unit Developing the reproduction process
52. All public files can be requested for digitization via the findings aids in the Archiefbank Just by clicking on the “digitize” button Production of 10.000 scans on weekly basis 1. Requesting for digitization MARAC Conference October 30 2009 We Do
53. A unit to be digitized must be able to be identified at each step of the handling process The units therefore get a unique meaningless order number An order number is provided by the metadata management system and is the basis for In practice: all units to be digitized get an order ticket 2. Providing ordernumbers MARAC Conference October 30 2009 Communication with the digitizer Scanning Assigning filenames Registration of filenames Billing by digitizer We Do
54. A unit to be digitized must be able to be identified at each step of the handling process The units therefore get a unique meaningless order number An order number is provided by the metadata management system and is the basis for In practice: all units to be digitized get an order ticket 2. Providing ordernumbers MARAC Conference October 30 2009 Communication with the digitizer Scanning Assigning filenames Registration of filenames Billing by digitizer We Do
55. The workflow system generates a list of all originals to asses from the repositories The list is sorted on repository / shelf to make retrieval efficient We Do 3. Assessing the originals MARAC Conference October 30 2009
56. MARAC Conference October 30 2009 All assessed originals are stored in a special room In this room all checks are executed We Do 4. Checking the originals
57. MARAC Conference October 30 2009 Information about the originals in our management systems is not always complete If an item falls into one of these categories the request is rejected B. Condition of the material A rough check of the originals takes place A. Content We Do 4. Checking the originals Copyrights Publicity Privacy Items that are in such a condition that digitizing or transport could cause damage, or are packaged in a way that scanning in conventional set-ups is not possible do not qualify for standard way of digitization
58. MARAC Conference October 30 2009 Information about the originals in our management systems is not always complete If an item falls into one of these categories the request is rejected B. Condition of the material A rough check of the originals takes place A. Content We Do 4. Checking the originals Copyrights Publicity Privacy Items that are in such a condition that digitizing or transport could cause damage, or are packaged in a way that scanning in conventional set-ups is not possible do not qualify for standard way of digitization
59. Material preparation is limited to the most minimal We Do 4. Checking the originals MARAC Conference October 30 2009 Staples are being removed as a rule Small reparations are executed by our restoration employees The sequence of the originals as found in the repository is not checked or altered We Do We don’t The originals are not numbered
60. But this is only true when the numbering tallies exact, because: Numbering the originals has one advantage: We Do Not number the originals MARAC Conference October 30 2009 The completeness of the scans (compared to the originals) can be guaranteed Numbers that are assigned double lead to illogical end numbers (100 scans: scan 100 has been numbered as 99) Experiments with numbering in practice learned that faultless numbering can not be realized A missing number in a sequence of scans leads to the conclusion that there is one original that has not been scanned
61. Securing completeness can be realized by other means: We Do MARAC Conference October 30 2009 Comparing scans to originals 1:1 after digitization Scanning the originals twice # scans = 365 # scans = 365 Low quality High quality master files Not number the originals
62. For secure transport, special flight cases are used We Do 5. Transport MARAC Conference October 30 2009
63. It has to be perfectly clear which filenames this should be After scanning the scan operator or data manager has to assign filenames to the scans Because, when the meaning changes, filenames should change too As a rule filenames contain no meaningful information We Do 6. / 7. Scanning and assigning filenames MARAC Conference October 30 2009 Filenames are the key between scans metadata
64. Assigning filenames at City Archives Amsterdam MARAC Conference October 30 2009 Customer request Management systems First 6#: ordernr Last 6#: serial nr Order ticket Filename Scanning the order A20758000001 A20758000002 A20758000003 Range A20758000001 – A20758999999 Archive 195 File 836 Order: A20758 A20758000004 A20758000005 Scan report A20758000001 A20758000002 A20758000003 A20758000004 A20758000005 12 digits Registration filenames Import
65. An application from which all checks can be executed is in development Scans and metadata are checked efficiently Where possible checks are automated 10. 11. Checking scans and metadata Basic checks We Do MARAC Conference October 30 2009 Depends on project Completeness Script Filenames Visual check production scans Visual check reference scans Quality scans Jhove File format validity MD-5 checksum comparison Data integrity Virus checker Viruses Method Check
66. After import the “order for digitization” of each unit is completed After approving of all checks, scans and metadata are imported into the management systems The imports are executed automatically, on basis of scripts and standard protocols for file transfer 13. 14. Import metadata and scans into management systems We Do MARAC Conference October 30 2009
67. After import the metadata are optimized for the search system For exchange of finding aids we use EAD From any workstation at the archive, directly via the CMS of the website The website is hosted from an external location Metadata are uploaded to the webserver by simple HTTP transfer 18. Import metadata into the website We Do MARAC Conference October 30 2009
68. Until then scans are transported by use of portable USB harddisks Bandwith of the internet connections at the archive is still too small for direct sFTP (or suchlike) upload of large quantities of scans to the webserver It seems likely that in the near future this will change 17. Import scans into the website Transport medium We Do MARAC Conference October 30 2009
69. Derivates for use of thumbnails and zoom / contrast functionality are made After connecting the harddisk to the server the import process starts Some basic checks are executed on the scans Import 17. Import scans into the website We Do MARAC Conference October 30 2009
70. MARAC Conference October 30 2009 The requester can decide whether to buy scans or not When both scans and metadata have been imported, automatically an e-mail is sent to the requester for digitization This email contains a link to the finding aid and thumbnails on the website Request complete! The happy customer: We Do
71. MARAC Conference October 30 2009 The requester can decide whether to buy scans or not When both scans and metadata have been imported, automatically an email is send to the requester for digitization This email contains a link to the finding aid and thumbnails on the website Request completed We Do The happy customer:
I will take you a step deeper into the workprocess of creating large amounts of scans. I’ll tell you about starting points and choises we have made and I’ll show you the result of some research we have done, particularyu towards image quality and filesize. Also, I’ll sohw you some back- and frontoffice tools from our webstie.
I will take you a step deeper into the workprocess of creating large amounts of scans. I’ll tell you about starting points and choises we have made and I’ll show you the result of some research we have done, particularyu towards image quality and filesize. Also, I’ll sohw you some back- and frontoffice tools from our webstie.