Saturday, April 25, 2009

To err is human


As Albert Einstein noted, "Anyone who has never made a mistake has never tried anything new.”

However, from a print production business perspective it's always helpful to learn the cause when mistakes happen since it can reveal what needs to be improved in the production process.

The three most common ways people make errors are:

Perception-based. These occur when there is incomplete or ambiguous information. For example: “We need a quote on a four-page folder” could mean many different things. Perception-based errors can be avoided by providing clear and distinctive instructions, standardizing instructions, and avoiding assumptions intended to fill in missing information.

Decision-based. These occur because of stress, pre-existing biases, assumptions, and over-confidence. This type of error can be avoided by using checklists, decision trees, and go-no-go flow charts.

Knowledge-based. These occur due to a lack of knowledge, information, and/or poor communication. These can be avoided by standardizing terms and operational conventions as well as through formal training.

Determining and documenting the source of mistakes helps clarify whether issues are random, intermittent, systemic, or trending in some way. This clarification informs your decisions. The goal being, not to make the same mistake twice.

There is no security on this earth – there is only opportunity.
- General Douglas MacArthur

Thursday, April 23, 2009

Rosettes – everything you didn't realize you needed to know


Rosette basics

Printing depends on halftoning to simulate shades of gray, color, and image detail. In four color process printing, four halftones – one for each of the cyan, magenta, yellow, and black inks are overlaid to produce the image. Unfortunately, overlapping two or more halftone grids can create an objectionable pattern called a "moiré" which, interestingly is the basis of the rosette.
Here, the overlaid halftone grids are 5 degrees and 10 degrees apart:

Here, the overlaid halftone grids are 15 degrees and 20 degrees apart:
As you can see, the greater the difference in angle between overlapping grids, the smaller the resulting moiré and the less apparent it is.
Here, the overlaid halftone grids are 30 degrees and 45 degrees apart:

Once the second grid has been rotated to 45 degrees, the moiré pattern is at its smallest and at a sufficient viewing distance seems to disappear.

Because a halftone screen is a quadratic grid (e.g. 90 degrees appears the same as 0 degrees, 135 degrees is the same as 45 degrees) the largest angle difference possible between two screens is 45 degrees, while the largest angle offset between three screens is 30 degrees (90/3=30). As a result, the defacto standard in four color printing has the three most visible process colors 30 degrees apart (C at 105 degrees, M at 75, and K at 45). Since Yellow is the least visible color it is angled at zero degrees – just 15 degrees from cyan. To further reduce moiré, the yellow screen is usually run at a higher frequency – typically about 108% of the other process colors.

The two kinds of rosettes

When screens of cyan, magenta, and black are overlaid at their respective angles (105, 75, 45) they form a moiré pattern called a "rosette."
To make the structure easier to see, here is the same graphic but with C, M, and K all black. Note that the yellow screen is not included since, because of its higher frequency, it does not form part of the rosette.

This type of rosette is called a "dot-centered" or "closed-centered" rosette because each of the patterns has a dot in its center.

Here is a gradient using the dot-centered rosette:
The second type of rosette is called a "clear-centered" or "open -centered" rosette. It is created by shifting one of the process colors one half row of dots from the other two colors.
Here it is in color:
And in black only for clarity:
And as a gradient:
In general, dot-centered rosettes:
• show a less visible pattern than clear centered ones
• have individual dots that land on top of one another - reducing chroma/gamut slightly
• produce color slightly differently than clear-centered rosettes
• tend to lose shadow detail
• with slight misregistration cause significant color shift
• are more popular with low screen frequencies - 100 lpi and lower

In general, clear-centered rosettes:
• show a more visible pattern than dot centered ones
• look slightly lighter due to more paper showing between dots
• produce color slightly differently than dot-centered rosettes
• tend to preserve shadow detail better
• resist color shifts better when slight misregistration occurs
• are more popular with high screen frequencies - 150 lpi and higher

Halftone dots are built inside halftone cells. Those cells have to fit together seamlessly. In order to rotate the screen, you have to rotate the cell – and there are only certain frequency/angle combinations at a given resolution where this seamless tiling is possible. The result is that at screen angles other than zero and 45 degrees, like cyan and magenta, the angles are not exactly as requested. As a result, the rosette can drift from being clear-centered to being dot-centered.

In this image the cyan is off by just two degrees and you can see the rosette going from dot-centered in the upper left to clear-centered in the middle and back to dot centered in the lower right:

In black only for clarity:
And reduced in size for clarity:
As it can appear in an image:
A well designed halftone screen will usually be able to maintain a clear-centered rosette across the largest diagonal plate that will be used. A less well designed screen may see "rosette drift" occurring over a distance of a few inches.

Rosette drift can also be caused by slight press misregistration caused by issues such as back sheet flare, web growth, or "waggle" (lateral sheet movement in the press). In this case rosette drift is not localized but occurs in the entire press sheet area.

In register - clear-centered rosettes:

Out of register by one half row of dots - now dot-centered rosettes with a subsequent tone and color shift:

With either cause of rosette drift, the problem can appear in presswork as:
• a moiré. Since a rosette is itself a high frequency moiré it is very sensitive to angular shifts.
• as "noise" or a grainy appearance in flat screen tint areas. This is because as the rosette drifts it has the effect of lowering the frequency of the halftone.
• as a shift in tone as the clear-centered rosettes are filled with a dot and then cleared again.
• as a color shift as the overprinting colors change their relationships with the shift from clear-centered rosettes to dot centered rosettes.

Tuesday, April 21, 2009

The Wayback View – My 25th Anniversary with the Apple Macintosh

Twenty-five years ago this week I bought my first Macintosh.
(click on image to enlarge)

Normal computer stores wouldn't carry it. The salesman at the Xerox store discouraged me and tried to sway me into buying a "conventional" computer – but I bought the MAC anyway. $3195 was a lot of money in 1984 when the average annual wage was around $17,000. I was a freelance commercial artist at the time and mostly used it for wordsmithing and accounting (there were no page layout programs back then). In the fall of 1984 I bought a "Thunderscan" which converted my Imagewriter dot matrix printer into a one bit scanner.
BOOM! Now MacPaint became really useful in my work. I began using its scans in my design projects. Here's a brochure I did for Quebecor/Ronalds Printing using Thunderscan images for the illustrations (click image to enlarge):
A close-up of MacPaint with one of the illustration scans used in the Quebecor brochure:
By 1985, with the introduction of the Apple Laserwriter, I could leverage the extra output resolution and quality to combine Thunderscan images with in-camera film-based special effects to produce images like this one which was used in a MacDonald Dettwiler corporate brochure:
When Microsoft Word for the MAC came out in 1985, a company called "Set and Send" in Vancouver introduced some software that allowed Word documents to drive a Compugraphic imager. This allowed me to set my own galley type – at $12 per foot of type, one text column wide and ready for paste-up. Of course I had to learn the Compugraphic formatting codes - however this innovation allowed me to eliminate outside typesetting services which was where the real profits in production occurred.

For me, this was the beginning of desktop publishing since I could create the entire document – on my desktop – with final press printed results that couldn't be distinguished from one created conventionally.

Later, in July 1985, Aldus Pagemaker was released and the rest, as they say, is history.

Friday, April 17, 2009

Image resolution for printing - LPI vs DPI a.k.a. LPI vs PPI a.k.a. LPI vs SPI

Background - pixels make the original image

A digital "raster" image acquired from a scanner, a digital camera, or created directly in a "paint" application like Adobe Photoshop is made up of a mosaic of "pixels" (picture elements)."
Here is an original image at actual size:

Here is a close up view showing the actual pixels that form the image:

The physical size of the image is described by two numbers which can be expressed two ways:

1) The number of pixels per inch/centimeter.
and
2) The number of pixels in both horizontal and vertical dimensions.

Or:

1) The number of pixels per inch/centimeter.
and
2) The horizontal and vertical dimensions expressed in inches/centimeters.

Those are just two ways of saying the same thing.
Here is the original image with a dialog box showing its dimensions:

Note that the dimensions have a "lock" icon beside them. This is because the relationship of pixels per inch (ppi) and vertical/horizontal size are "locked" together. Changing one changes the other as you can see in the below dialog boxes (click on image to enlarge):

Note that as the resolution is changed (from 600 to 300 and 300 to 150 pixels per inch) only the density of the pixels changes, not the number of total pixels in the image, in this case 1412 pixels x 2028 pixels, therefore the file size remains the same. Put another way, each time the resolution in ppi is increased, or lowered, the physical image size changes but the total number of pixels forming the image (and hence the detail) remains the same.

Note that I use the term "pixels per inch" - ppi. Very often the term that is used is "dots per inch" or dpi. Technically the terms are not interchangeable - however, in daily usage, when speaking about digital images the terms are considered as meaning the same thing. You may sometimes hear the term "spi" - samples per inch. This refers to a scanner's resolution - i.e. it ability to acquire an image at so many samples per inch (e.g. 300 spi). Again, in practical usage, when speaking about digital images - ppi, dpi, and spi can be understood as meaning the same thing.

Interestingly, digital cameras typically do not have a resolution assigned to them.

Instead a digital camera captures data based on the "megapixel" ability of its CCD sensor. For example, a 14.2 megapixel camera might capture an image that's 4592 pixels by 3056 pixels, which equals 14,033,152 total pixels. When you open the file into an image-editing program a resolution must be assigned to the file. Most programs, including Photoshop, use 72 ppi as the default resolution.

Background - halftone dots make the image reproduction

Because printing presses can only lay down 100% ink or 0% ink, digital images acquired from scanners, digital cameras, or created directly in "paint" applications need to be converted into a binary (on/off) format. This is done through a process called halftone screening. The result is that the image will be converted to dots of either 100% or 0% ink with the original tones being simulated, in this case, by the size of the dots. Bigger dots represent darker tones - smaller dots represent lighter tones:

The fineness of the screen, and hence the level of detail in the original that can be preserved, is determined by how densely packed the dots are and is indirectly described by how many rows - or lines of dots are used per inch (or centimeter) to create the image. These virtual lines are highlighted in red below:
In this example the image is made up of 85 lines of dots per inch – expressed more commonly as an 85 lines per inch halftone - or more simply stated: an 85 lpi halftone image.

The key thing to remember is that although the halftone image is made up of dots - the level of detail that it can reproduce is described in terms of lpi NOT dpi.
So, original image pixel density/detail = ppi, spi, or dpi. Halftone reproduction dot density/detail = lpi.


Of course, in order to pack more lines of dots into an inch - the smaller the dots become and hence the greater amount of image detail that is preserved.

40 lpi halftone:

100 lpi halftone:

200 lpi halftone:


It is the relationship of how densely packed the original pixels are (see part 1) compared to the frequency of lines per inch of the halftone screen dots that determines what image resolution is appropriate for its reproduction in print.

The relationship between dpi/ppi and lpi for
grayscale
images


The guiding principle for understanding what original image resolution (ppi/dpi) is needed compared to the halftone screen (lpi) that will be used is that the image pixels should always be more densely packed (ppi/dpi) than the detail resolving ability (lpi) of the halftone screen that is used.

To illustrate this principle I'll take a section of the same image at different resolutions (ppi/dpi) and reproduce it using the same 150 lpi halftone screen:

Original 75 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is one half of the halftone screen resolution (lpi). As a result the halftone reproduces the individual pixels of the original. This visible artifact is termed "staircasing," the "jaggies," or "pixelation."

Original 100 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is two thirds of the halftone screen resolution (lpi). As a result the halftone still reproduces the individual pixels of the original - but they are less visible.

Original 150 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is equal to the halftone screen resolution (lpi). As a result the halftone still reproduces the individual pixels of the original - but they are much less visible.

Original 225 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is 1.5 times greater than the halftone screen resolution (lpi). Although some original image pixels may still be visible, in general, the halftone no longer resolves the individual pixels of the original - just the tones they represent.
This minimum required original resolution can be represented by the formula: 1.5 X lpi = ppi @ 100% reproduction.

Original 300 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is twice the halftone screen resolution (lpi). As a result the halftone no longer resolves the individual pixels of the original - just the tones they represent.
This ideal required original resolution can be represented by the formula: 2 X lpi = ppi @ 100% reproduction.

Original 600 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is four times the halftone screen resolution (lpi). The image file size is about 7 times larger than the 225 ppi/dpi image but provides effectively no difference in the final reproduction.

The relationship between dpi/ppi and lpi for CMYK
images


As with grayscale images, the guiding principle for understanding what original image resolution (ppi/dpi) is needed compared to the halftone screen (lpi) that will be used is that the halftone screen should not reproduce the image pixels themselves but instead the tones the pixels represent. It is worth comparing these images to their grayscale equivalents in part 3.

To illustrate this principle, I'll take a section of an image rendered at different resolutions (ppi/dpi) that has been converted from grayscale to CMYK and reproduce it using the same 150 lpi halftone screen:

Original 75 ppi/dpi - halftone screen 150 lpi:Here the image ppi/dpi is one half of the halftone screen resolution (lpi). As a result the halftone reproduces the individual pixels of the original. This visible artifact is termed "staircasing," the "jaggies," or "pixelation." That being said, the jaggies are less severe than we saw in the grayscale image at the same ppi/dpi. Also the numbers on the sail appear clearer. This suggests that it might be possible to use a lower image resolution for reproducing a CMYK image than can be used for a grayscale image.

Original 100 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is two thirds of the halftone screen resolution (lpi). As a result the halftone still reproduces the individual pixels of the original - but they are less visible.

Original 150 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is equal to the halftone screen resolution (lpi). Because the CMYK image is a composite of four individual halftone images it tends to lessen the visibility of the individual pixels of the original.
This minimum required original resolution for a CMYK image can be represented by the formula: lpi = ppi @ 100% reproduction.

Original 225 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is 1.5 times greater than the halftone screen resolution (lpi). The halftone no longer resolves the individual pixels of the original - just the tones they represent.
This ideal original resolution can be represented by the formula: 1.5 X lpi = ppi @ 100% reproduction.

Original 300 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is twice the halftone screen resolution (lpi). As a result the halftone no longer resolves the individual pixels of the original - just the tones they represent.
This maximum required original resolution can be represented by the formula: 2 X lpi = ppi @ 100% reproduction.

Original 600 ppi/dpi - halftone screen 150 lpi:
Here the image ppi/dpi is four times the halftone screen resolution (lpi). The image file size is about 7 times larger than the 225 ppi/dpi image but provides effectively no difference in the final reproduction.

The below table provides image resolution requirements for a variety of typical print applications:Note that this table refers to conventional "AM" halftone screening where the lpi signifies the dot density and hence the resolution of the halftone screen. However, there is another type of halftone screen in use which does not have a traditional lpi. Instead, this type of screening organizes the halftone dots in random appearing patterns. Below are three different vendor's offerings (click on images to enlarge):
This type of halftone is called "FM" or "Stochastic" screening (covered in other posts in this blog). Rather than indicating resolution according to "lpi" - the average actual dot size, specified in microns, of the screen pattern is used instead. Typical dot sizes are: 10 - 20 micron for commercial work, 20 - 25 micron for magazine work, and 35 micron for newspaper work. Because this type of screening has a higher average resolution than conventional AM screening - it's a good idea to use images at a higher resolution to take advantage of this screening's detail rendering capability. Typically 400 ppi/dpi for 10-20 micron FM, 300 ppi for 25 micron, and 200 ppi/dpi for 35 micron.

Image resolution "gotchas" – where things can go wrong

Whether you are targeting your images for AM or FM screening, there are at least three places where the resolution of the images may be accidently altered:

1) If the image is resized/scaled in the page layout application – it may no longer have an appropriate resolution:
2) If the image is resized/scaled when the file is converted to the PDF format – it may no longer have an appropriate resolution:
3) If the printshop's workflow is setup to resample incoming documents – they may no longer have an appropriate resolution. Most prepress RIPs are set, by default, to downsample incoming files to 300 ppi/dpi.