Print and prepress glossary: terms printers actually use

Most jobs that go wrong go wrong because two people used the same word for different things. This page collects the terms that come up between a printer and a customer, with what each one means and where it gets muddled. Definitions are drawn from the ISO and ICC specifications and the vendor documentation where those exist, and marked as trade convention where they don’t. Use the headings to link straight to a term when you need to settle something over email.

Files, page boxes and what gets supplied

Vector

Artwork described as geometry: coordinates, straight lines, Bézier curves, fills and strokes, held in a device-independent coordinate space. In PostScript and PDF the default user space unit is 1/72 inch. ISO 15930 treats vector data as the encoding for line art and text. The description is mathematical, so it carries no inherent resolution and pixel dimensions get decided at the RIP.

File extension proves nothing. An .ai, .eps or .pdf is a container and can hold a single placed 300 ppi TIFF and nothing else. Vector also gets confused with editable: converted type and flattened transparency are still vector, they just can’t be edited any more.

Raster

An image stored as a rectangular grid of samples, each pixel carrying fixed colour values. The grid’s dimensions are set when the image is made or scanned, and they can’t be increased without inventing data. PostScript defines a raster output device as one whose output is a rectangular array of individually addressable pixels. Enlarging a raster enlarges the pixels, and the detail ceiling is the sample count.

Strictly, bitmap means one bit per pixel, though everyday use makes raster and bitmap synonyms. Raster also gets confused with halftone dots (a 300 ppi image is not 300 dots of ink per inch) and with Illustrator’s destructive Rasterise command.

DPI

Dots per inch: the addressable marking dots an output device can place along an inch. Platesetters and imagesetters commonly work at 2400 dpi, office lasers at 600 to 1200 dpi, and inkjets quote 1200 dpi and up for drop placement. It’s a property of the device, and it sets how finely a halftone dot can be built.

The trade and most software dialogs use dpi and ppi interchangeably, so “send 300 dpi artwork” means 300 ppi at final size. There’s more on that in what dpi you need for print. Don’t confuse it with lpi either. 2400 dpi builds a 150 lpi screen with a full 257 levels and starts running short of levels above that, and inkjet dpi marketing figures don’t convert into halftone detail.

PPI

Pixels per inch: the sampling density of a raster image at a stated physical size. A 3000 x 2400 pixel image placed 10 inches wide is 300 ppi. The customary target for halftone offset is 1.5 to 2 times the screen ruling, so 225 to 300 ppi at 150 lpi, which is where the 300 ppi convention comes from. Coated work at 175 lpi is usually specified at 300 ppi.

Scanners deliver spi, samples per inch, a third measure again. The ppi figure in a file header is only metadata, so changing it without resampling changes the print size and leaves the detail alone. Photoshop’s dialog says pixels/inch and everyone says dpi.

Effective resolution

The ppi an image actually delivers after layout scaling. Effective ppi equals actual ppi divided by the scale factor, so a 300 ppi image placed at 200% is 150 ppi effective, and the same image at 50% is 600 ppi. InDesign’s Links panel reports Actual PPI and Effective PPI separately, and preflight tools measure the effective figure because that’s what reaches the RIP.

Actual against effective is the standard mix-up on both sides of a job, with the designer quoting the Photoshop number and the printer measuring the placed one. Non-uniform scaling or rotation gives two different effective values, horizontal and vertical, and some tools report only the lower.

Interpolation and resampling

Changing an image’s pixel count by computing new values from neighbours. Nearest-neighbour subsampling takes one pixel per area; bilinear and bicubic weight several. Downsampling discards data, upsampling invents it and adds no detail. Distiller and PDF export downsample only when the input exceeds the target by a threshold ratio, default 1.5, settable from 1.0 to 10.0. PDF’s Interpolate flag, false by default, asks the reader to smooth low-resolution images at render time.

Resampling gets confused with resizing. Unchecking Resample in Photoshop only rewrites the ppi tag and the print size. Downsample and subsample are used interchangeably, though subsampling is the crude one, and a scanner’s “interpolated resolution” is not optical detail.

Compression (JPEG and lossless)

JPEG (ISO/IEC 10918-1, DCTDecode in PDF) is lossy. The image is split into 8 x 8 sample blocks, transformed by DCT, and the coefficients quantised. That quantisation is where data goes, which is why artefacts appear at hard edges and accumulate with each re-save. The lossless filters, FlateDecode, LZWDecode, RunLengthDecode and CCITTFaxDecode, reproduce the original data exactly. JPEG 2000 (JPXDecode) can be either. Lossy compression only ever applies to sampled image data.

“ZIP is lossless” holds for the filter and not for the pipeline, since Flate applied to an already JPEG-damaged image preserves the damage faithfully. Quality sliders aren’t comparable between applications, and maximum quality JPEG inside a PDF is still lossy.

PostScript

Adobe’s page description language, introduced in 1985 and reaching LanguageLevel 3 in 1997. It’s a dynamic programming language with graphics primitives, so a page is a program executed by an interpreter in the output device, describing text, graphics and sampled images in device- and resolution-independent terms. The same file images correctly at 600 dpi and at 2400 dpi.

PostScript and PDF get conflated. PDF borrows the imaging model but is static data with no operand stack and no control flow. “PostScript font” refers to Type 1 or CFF outlines, a separate thing from the language, and EPS is PostScript constrained for placement inside another page.

PDF

Portable Document Format: a structured binary file format whose imaging model derives from PostScript, but which is static data, so pages can be addressed randomly. PDF 1.7 became ISO 32000-1:2008 after Adobe donated the specification to ISO in January 2007. PDF 2.0 is ISO 32000-2, first published July 2017, second edition 2020, available at no cost from the PDF Association.

PDF version numbers (1.3, 1.4, 1.6, 2.0) get confused with Acrobat version numbers and with PDF/X levels. Valid also gets confused with printable: a file can conform perfectly to ISO 32000 and still carry RGB images, unembedded fonts and no bleed.

PDF/X

ISO 15930: restricted subsets of PDF for complete or blind exchange of print-ready data, requiring no prior agreement between sender and receiver. The common rules are all fonts embedded, an output intent identifying the characterised printing condition, no encryption, no OPI, and either a TrimBox or an ArtBox but never both, with a BleedBox where bleed is used. The levels in general use are X-1a (CMYK and spot), X-3 (colour-managed), X-4 (PDF 1.6, live transparency and layers), X-5 (partial exchange, external references) and X-6 (PDF 2.0).

It gets confused with PDF/A, which is for archiving under ISO 19005, and with PDF/VT for variable data, which has to conform to PDF/X as well. The more damaging confusion is treating conformance as a quality check. PDF/X says nothing about image resolution, total ink coverage, trapping or bleed, which is preflight’s job.

Preflight

Automated checking of an incoming file against a rule set before production: fonts embedded, colour spaces and spot-colour names, effective image resolution, total ink coverage, page-box geometry and bleed, overprint settings, transparency, hairline widths. The term comes from pilots’ checklists and was first used publicly by consultant Chuck Weger at the 1990 Color Connections conference. Ghent Workgroup profiles build on PDF/X-4 and are published per market segment.

Preflight gets confused with PDF/X validation, which only tests conformance to ISO 15930. A valid PDF/X-4 file routinely fails a printer’s preflight on resolution or ink limit. Fixups, the automatic correction step, are a separate and riskier operation.

RIP (raster image processor)

The software or hardware that interprets a page description (PostScript, PDF, PDF/X, PDF/VT) and renders it to the pixel array a marking engine needs. The usual order is interpret, render to continuous tone, colour-manage against the output intent or device profile, then screen to halftone or dither at device resolution. 2400 dpi is typical for platesetters.

The RIP gets confused with the workflow system wrapped around it, since Prinergy, Apogee and Automation Engine each contain one, and “ripped” gets used loosely for any prepress processing. RIP resolution in dpi is a different measure from screen ruling in lpi. Two conformant RIPs can legitimately render the same PDF differently.

Rasterisation

Turning a geometric description into a pixel array by deciding, for each device pixel, whether and with what value it’s marked. The classic non-anti-aliased rule turns a pixel on if its centre falls inside the outline, with interiors determined by a non-zero winding count, plus dropout-control rules so thin features don’t vanish. Anti-aliasing and halftone screening are separate later steps.

It gets confused with screening, which converts rasterised contone values into printable dots, and with flattening, which is a transparency operation. Rasterise is also a destructive menu command in Illustrator and Photoshop, and that isn’t what a RIP does to a page.

Transparency flattening

Converting live transparency (blend modes, opacity, soft masks, all introduced with PDF 1.4) into opaque objects that older imaging models can reproduce. The flattener divides artwork into vector-based areas and rasterised areas, controlled by a raster/vector balance and two resolutions. Adobe’s [High Resolution] preset uses 1200 ppi for line art and text, 300 ppi for gradients and meshes. Required for PostScript, EPS, PDF 1.3 and PDF/X-1a, and unnecessary for PDF/X-4 and PDF 2.0.

Flattening transparency isn’t flattening layers in Photoshop, and it isn’t Acrobat’s Flatten command, which merges annotations and form fields. Stitching lines and colour shifts across flattened regions get blamed on the RIP when they come from the flattener’s atomic regions and its spot-to-process conversion.

Overprint preview

A display mode that simulates on screen what overprinting will do on press, replacing the normal opaque composite with overlapping inks mixing. Acrobat’s Page Display preference “Use Overprint Preview” defaults to Only For PDF/X Files. InDesign and Illustrator offer View > Overprint Preview, and Acrobat Pro’s Output Preview adds separation viewing and output-intent simulation.

The preview gets confused with the attribute. Switching it on changes nothing in the file. With the overprint parameter false, the default, painting erases underlying colourants; true leaves them undisturbed. Overprint mode OPM 1, nonzero overprint mode, additionally makes a 0.0 tint in DeviceCMYK leave that colourant untouched.

MediaBox

The only required page box: the boundaries of the physical medium on which the page is to be displayed or printed, in default user space units, and inheritable through the page tree. It’s the largest box, since crop, bleed, trim and art boxes are effectively reduced to their intersection with it, and content falling outside it may be discarded without affecting the file’s meaning.

People read the MediaBox as the press sheet. It’s the PDF page’s medium and has nothing to do with the paper the job runs on. It also gets confused with the page size Acrobat displays, which comes from the CropBox, and it’s wrongly assumed to include bleed, which it does only if whoever made the file deliberately made it larger than trim.

CropBox

The rectangle defining the visible region of the page, to which contents are clipped when displayed or printed. Alone among the boxes it has no defined meaning in terms of physical page geometry or intended use. It only imposes clipping, and it defaults to the MediaBox. With no imposition instructions it also determines how content sits on the medium. The Ghent Workgroup advises against relying on it in prepress.

Most viewers report the CropBox as the page size, so it gets mistaken for trim. Cropping in Acrobat only rewrites boxes; the content stays put, ready to reappear downstream. Because bleed, trim and art boxes all default to the CropBox, one wrong CropBox propagates into all of them.

TrimBox

The rectangle defining the intended dimensions of the finished page after trimming, expressed in default user space units of 1/72 inch. Optional in ordinary PDF, where it defaults to the CropBox. It may be smaller than the MediaBox to leave room for marks and colour bars. Imposition software positions pages on the press sheet by the TrimBox and can generate cut and fold marks from it.

The usual mix-up is with the CropBox, since Acrobat’s Document Properties reports Page Size from the CropBox, so a file can look correct and still carry a wrong or missing TrimBox. Trim marks drawn in the artwork are decoration; only the box drives imposition. PDF/X allows a TrimBox or an ArtBox, never both.

BleedBox

The rectangle to which page contents shall be clipped when output in a production environment, allowing for the physical limitations of cutting, folding and trimming equipment. Optional, defaulting to the CropBox. It typically sits 3 to 5 mm outside the TrimBox, 3 mm being the European convention and 1/8 inch (3.175 mm) the North American one. Printer’s marks may legitimately fall outside it.

Setting a BleedBox does not create bleed. If the image stops at the trim edge, the box just clips empty space. The box also gets confused with the general term bleed, and with the slug area, which carries marks and instructions outside the bleed and isn’t a PDF page box at all.

ArtBox

The rectangle defining the extent of the page’s meaningful content, including any intended white space, as the page’s creator intended it. Optional, defaulting to the CropBox. Its designed use is placing a page inside another document, so an advertisement dropped into a magazine layout is positioned by its ArtBox and clipped to the art or bleed box depending which edges bleed off.

In conversation people use ArtBox and TrimBox interchangeably, though PDF/X permits only one of the two and some advertisement specifications demand a particular one. The ArtBox is also the box most often left at an application default, so an ArtBox equal to the CropBox carries no real information.

Trim

The finished cut line, and by extension the finished size. In a print-ready PDF that’s the TrimBox, and imposition software positions pages on the press sheet by the TrimBox. Flat work is cut to trim on a guillotine. Bound work is trimmed on head, foot and fore-edge after binding, so trim allowance has to exist on all three edges.

Trim gets confused with the CropBox, which drives what a viewer displays and shouldn’t be relied on to carry finished size. Some tools honour it in production and some don’t, which is the reason imposition works off the TrimBox. Trim size is also used loosely for flat size on folded work, where the finished folded dimensions and the flat unfolded dimensions are different numbers.

Bleed

Artwork extended past the trim line so that cutting variance can’t leave a white edge. 3 mm is the European convention and 0.125 in the North American one, with the BleedBox typically sitting 3 to 5 mm outside the TrimBox. Bound and large-format work often calls for 5 mm. Bleed has to be genuine extended image; scaling the same image up doesn’t count.

Bleed gets confused with the slug, which sits outside it carrying instructions, and with the safe area, which sits inside the trim. Bleed is used as a verb too, for an image running off the page. The commonest supply fault is crop marks present with no extended artwork behind them.

Mirrored bleed

Also called synthetic or faked bleed: a bleed band rebuilt from artwork that stopped at the trim, by reflecting the edge pixels outward. It’s what RIPs and prepress tools do when a file arrives without bleed. It holds up on flat colour, texture, grain and soft photo edges. It gives itself away on hard lines crossing the trim at an angle, on regular repeating patterns, on gradients running perpendicular to the trim, and on type or a logo near the edge, where it produces a reversed ghost of the lettering.

It can’t rescue artwork with a thin border, which the mirror doubles into a stripe, and it can’t help artwork that stops short of the trim, since mirroring a white margin only produces more white. The band is only ever seen if the cut misses, which is the case it exists to cover. The numbers are in how much bleed a print job needs.

Safe area and margin

The zone inside the trim where live matter (text, logos, folios, barcodes) is kept so cutting tolerance can’t clip it. 3 mm inside trim is the usual stated minimum and 5 mm the safer working figure. Perfect-bound work needs considerably more on the binding edge. No ISO standard fixes the value, so each printer publishes its own.

Safe area gets confused with the design margin, which is a typographic decision, whereas this is a production tolerance. Some workflows carry it in the ArtBox, which isn’t what the ArtBox is for. It’s also distinct from the quiet zone required around a barcode.

Crop marks and trim marks

Pairs of perpendicular hairlines set outside each corner of the trim, showing the cutter where the finished edge falls. The mark offset has to be at least the bleed distance or the marks sit inside the bleed, so with 3 mm bleed, offset 3 mm or more. Crop marks and trim marks are the same thing, and Adobe’s dialog calls them crop marks.

They get confused with registration marks, which are bullseye targets carried in every separation to check colour-to-colour fit and are useless as a cut guide. They also get mistaken for the TrimBox, and imposition reads the TrimBox while ignoring drawn marks. Marks drawn inside a page-size PDF with no bleed is a standing prepress complaint.

Slug

The area outside both the page and the bleed, carrying job name, date, file name, operator notes, colour bars and sign-off boxes. It images on the press sheet and is trimmed away. InDesign defines it separately from bleed in Document Setup, and objects placed there output only when the slug area is explicitly included at print or export.

The word collides with the hot-metal sense, where a Linotype slug is a whole line of text cast as one piece of lead-antimony-tin alloy, and with editorial use, where a slug is a story’s short working title. It also gets confused with bleed, and the slug lies outside the bleed.

Type and fonts

Outline fonts and the “infinite zoom” property

Fonts whose glyphs are stored as filled contours on an em grid. Coordinates are relative to the em square, 1000 units per em for Type 1 and CFF, commonly 2048 for TrueType, so one master serves every size. At output, pixel coordinate = em coordinate x ppem / upem, where ppem = point size x device dpi / 72. Rescaling is arithmetic, so edges never degrade.

“Infinite zoom” is overstated in the trade. A design is fixed at one weight and detail level, so a text-optimised master blown up to 200 pt gives clean edges with wrong optical proportions. The property also gets confused with converting to outlines, which is an operation performed on a document.

PostScript Type 1

Adobe’s outline font format, introduced in 1984 and published as the Adobe Type 1 Font Format, second edition 1990. Glyphs are cubic Bézier charstrings in a 1000-unit character space, FontMatrix [0.001 0 0 0.001 0 0], with declarative hints in an eexec-encrypted Private dictionary. It shipped as separate outline and metrics files. Adobe ended support for authoring with Type 1 fonts in January 2023.

Type 1 gets confused with Type 3, which uses arbitrary PostScript procedures per glyph and has no hinting mechanism, and with CFF, sometimes called Type 1C, the compact re-encoding of the same outlines used inside OpenType and PDF. In casual use “PostScript font” can mean Type 1 or any CFF-flavoured OpenType.

TrueType

Apple’s outline format from the early 1990s, licensed to Microsoft, now the glyf outline flavour inside OpenType. Contours are second-order quadratic Bézier splines built from on-curve and off-curve points. Point coordinates are FUnits, each within -16384 to +16383, units per em is chosen by the maker in the range 16 to 16384, and a power of two such as 2048 scales fastest. Hinting is bytecode executed by a stack-based interpreter.

TrueType and OpenType get treated as alternatives, though nearly every .ttf shipped today is an OpenType font containing TrueType outlines. Curve order is the real difference against Type 1, quadratic here and cubic there, which is why conversion between the two is an approximation.

OpenType

A font format built on the sfnt table container, carrying outlines either as TrueType glyf data or as PostScript CFF or CFF2 data, in a single cross-platform file. It supports the whole of Unicode including supplementary planes, up to 65,535 glyphs per font, GSUB and GPOS layout features, font collections and variable fonts. OpenType 1.4 was contributed to ISO and underlies ISO/IEC 14496-22, Open Font Format.

The .otf and .ttf extensions mark only the outline flavour, though RIPs and flatteners can behave differently on CFF and TrueType outlines. OpenType features such as ligatures and small caps get confused with the format itself, and variable fonts get confused with multiple master.

Font embedding

Storing the font program itself inside the file. In PDF it’s a stream referenced from the font descriptor: FontFile for Type 1, FontFile2 for TrueType, FontFile3 with subtype Type1C, CIDFontType0C or OpenType. ISO 32000 notes that a licence may forbid embedding, and that an embedded font shall be used only to view and print the document. Every PDF/X level requires all fonts used to be embedded.

Embedding gets confused with subsetting, which embeds only some glyphs, and with outlining. “Fonts embedded” in a preflight report doesn’t mean the glyph set is complete or the embedding legal.

Font subsetting

Embedding only the glyphs a document actually uses. ISO 32000 requires a subset font’s PostScript name to begin with a tag of exactly six uppercase letters followed by a plus sign, as in EOODIA+Poetica, with different subsets in the same file carrying different tags, and a CharSet or CIDSet entry may list what’s present. It cuts file size sharply on large CJK and OpenType fonts.

Subsetting gets confused with full embedding. A subset PDF can’t be text-edited beyond the glyphs present, and merging documents can leave several differently tagged subsets of one typeface. The six-letter tag is arbitrary, so matching tags don’t prove matching subsets.

Converting type to outlines

Replacing live text with filled vector paths, via Type > Create Outlines in Illustrator or InDesign. The glyphs stay vector and image identically at platemaking resolution. The text stops being editable, searchable or accessible, hinting is discarded, and the anchor-point count can inflate the file. Very small type and hairline serifs may render marginally heavier once unhinted.

Outlining gets treated as equivalent to embedding. Take it as a last resort for fonts that can’t be embedded legally or technically. It fixes nothing about colour, overprint or missing glyphs, and it’s irreversible, so once the file is saved the typeface can’t be identified or corrected from it.

Hinting

Data or code that distorts a glyph outline onto the pixel grid, known as grid-fitting, so stems keep even weight and thin features don’t drop out at low pixels-per-em. TrueType hinting is imperative bytecode executed by an interpreter using the fpgm, prep and cvt tables. Type 1 and CFF use declarative hints in the Private dictionary that the rasteriser interprets. Its influence fades above roughly 35 ppem.

Hinting gets confused with anti-aliasing, which is a rendering technique, and with kerning. Its irrelevance in print is overstated: 8 pt type at 2400 dpi is about 267 ppem, so hinting is moot on plate, while the same PDF at 100% on a 96 dpi screen is about 11 ppem, where hinting governs the rendering the client sees.

Colour

RGB

Additive colour model in which red, green and blue light are summed, where 0,0,0 is black and full scale is white. RGB numbers are device-dependent and mean nothing until an encoding fixes the primaries, white point and transfer function. Cameras, scanners and displays work in RGB, presses don’t. RGB is a model; sRGB, Adobe RGB (1998) and Display P3 are spaces. An untagged RGB file has no defined colour at all, and most software silently assumes sRGB. The other standing error is believing RGB gamut always contains CMYK gamut, when sRGB can’t hold a printed process cyan.

sRGB

Standard RGB encoding, IEC 61966-2-1:1999. Primaries R 0.6400/0.3300, G 0.3000/0.6000, B 0.1500/0.0600, white point D65 at 0.3127, 0.3290 and 80 cd/m². The transfer function is piecewise, linear at x12.92 below 0.0031308, then 1.055 x C^(1/2.4) minus 0.055. Reference viewing environment is 64 lux ambient with a 4.1 cd/m² surround. It’s the default assumption for untagged files, the web and consumer devices. People call it gamma 2.2 routinely. The actual curve uses exponent 2.4 with a linear toe and an effective gamma of roughly 2.2, so the two aren’t interchangeable in calculation. It also gets confused with a monitor’s sRGB mode, which is a display preset. sRGB is the smallest of the common working spaces and clips saturated printable cyans and yellows.

Adobe RGB (1998)

Adobe’s wide-gamut RGB working space. Primaries R 0.6400/0.3300, G 0.2100/0.7100, B 0.1500/0.0600, white point D65 at 0.3127, 0.3290, single-exponent transfer function of 563/256, which is 2.19921875. Reference display white 160 cd/m², 32 lux ambient, 32 cd/m² proximal field. It was designed to enclose most CMYK-printable colours, particularly the cyan-green region sRGB clips, which is why prepress took it up. People talk about “Adobe 1998” as though it were a separate space, and mix it up with ProPhoto RGB, which is far larger and needs 16-bit. Adobe RGB sent untagged into an sRGB-assuming workflow prints flat and desaturated, so the wider gamut produces a worse result. Its green primary is widely reported to derive from a mis-transcribed SMPTE-240M value. Treat that as anecdote.

Display P3

Apple’s display encoding, now also a CSS Color 4 predefined space. It takes the DCI-P3 primaries, R 0.680/0.320, G 0.265/0.690, B 0.150/0.060, and pairs them with a D65 white point and the sRGB piecewise transfer function. Roughly 25% wider gamut than sRGB, mainly in reds and greens. Standard on Apple hardware and increasingly on Windows and Android panels. Display P3 and DCI-P3 aren’t the same. DCI-P3 is the cinema space, with 2.6 gamma and a green-shifted DCI white, which is why Display P3 is sometimes written P3-D65. It’s a poor proxy for a print space, since much of its extra red and green falls outside any CMYK gamut while its blue-cyan advantage over sRGB is small.

CMYK

Subtractive four-ink model for process printing: cyan, magenta, yellow and K, the key or black printer. Transparent inks are overprinted as halftone screens, each on its own separation. CMYK values are device-dependent, so the same numbers give a different colour on a different press, ink set or substrate. They only describe a colour when tied to a named printing condition and profile such as GRACoL 2013 (CRPC6) or PSO Coated v3, which makes “convert to CMYK” meaningless until you say which CMYK. CMYK is a colorant model that gets treated as a colour space. K stands for key, the key plate the others register to. Extended-gamut work adds orange, green and violet as further process colours, so CMYK no longer means four inks everywhere. The route from one model to the other is covered in RGB, CMYK and what happens to colour before press.

Process colour

The inks used to build most colours by overprinting halftone screens, conventionally cyan, magenta, yellow and black. Real pigments are impure, so equal CMY doesn’t give neutral grey and full-strength CMY gives a muddy brown, which is why black is added for shadow density, detail and neutral control. Extended-gamut printing treats orange, green and violet as further process inks. Process is set against spot, though the boundary is soft: an OGV ink is a fixed process ink even though it comes from a named pot. The old ink names mislead, since process blue is essentially cyan and process red is close to magenta, and neither is PMS Blue 072 or Red 032. “Four colour” to a printer means process, while “four colours” may mean four spot units.

Colour separation

Splitting artwork into the single-ink component images that will actually be printed, normally one per process ink plus one per spot ink. Traditionally shot photographically through filters onto film, today generated in the RIP at platesetting. Each separation carries its own halftone screen ruling and angle. Designers now supply composite PDF/X rather than pre-separated files, with ISO 15930 defining the delivery format. Separation and plate get used interchangeably. The separation is the image data, the plate is the physical carrier made from it, and separations exist for digital presses with no plates at all. Separation also gets confused with a Photoshop channel, which has no screening, and in flexo with colour breaks, the artwork-side decision about which element goes on which ink.

Colour plate

The physical printing forme carrying one separation, mounted on one press unit and inked with one ink. A four-colour job needs four, and each spot ink adds another. Modern plates are imaged directly by laser via computer-to-plate with no intermediate film. Plate count drives cost, press selection and pass count, so it’s usually the first commercial question on a quote. Plates and separations get confused. “How many plates” is a costing question, “how many separations” is a data question, and they diverge on multi-pass or ganged work. ISO calls it a printing forme, which almost nobody says out loud, and the trade still says “plate change” on digital presses that have none, meaning a job or media change.

Gamut

The range of colours a medium, device or process can actually reproduce under stated viewing conditions. It’s a three-dimensional volume in colour space, and gamut volume is commonly quoted in cubic CIELAB units when spaces are compared. For a press it depends on ink set, substrate, screening, ink limit and measurement condition, so a gamut statement with no printing condition attached is incomplete. Gamut gets confused with colour space: the space is the coordinate system, the gamut is the reachable volume inside it. “Bigger gamut” says nothing about accuracy or smoothness. The classic error is comparing gamuts on a flat CIE xy chromaticity diagram, which throws away lightness and makes CMYK look hopeless when at high lightness it can beat sRGB.

Out of gamut

A colour present in the source that the destination can’t reproduce, and which therefore has to be mapped, clipped or compressed on conversion. Whether a colour is out of gamut depends on the source, destination and intent in play, so the file alone can’t tell you. ICC colorimetric intents preserve in-gamut colorimetry and sacrifice out-of-gamut colours, and the ICC spec deliberately leaves that mapping unspecified, so vendors differ. It gets confused with clipped or over-range channel values in the file itself, which is a different problem. Photoshop’s gamut warning depends entirely on the chosen proof profile, so the same image shows different warnings against different presses. The condition runs both ways, since printed process cyan and some yellows sit outside sRGB.

Colour profile

Loose trade term for any data set describing the colour behaviour of a device, medium or space. In practice it nearly always means an ICC profile, though printers also use it for a characterisation data set in CGATS or CxF, a monitor calibration curve, a device-link, or a RIP’s internal media setting. ISO 12647-2 keeps “characterisation data set” and “ICC output profile” as separate deliverables for exactly this reason. Every ICC profile is a colour profile and the reverse doesn’t hold. It also gets confused with calibration. Calibration puts a device into a known repeatable state and the profile describes the state it’s in, so a profile made from an uncalibrated device stops matching as soon as the device drifts.

ICC profile

A file in the ICC.1 format that maps a device’s values to and from the Profile Connection Space. The PCS is CIEXYZ or CIELAB with a D50 adopted white, nCIEXYZ X 0.9642, Y 1.0000, Z 0.8249. Profile classes include input, display, output, device link, colour space, abstract and named colour. The current specification is ICC.1:2022, and version 2 and version 4 profiles both remain in daily circulation. A profile alone alters no pixel; the CMM does the work. Within ICC work the live confusions are v2 against v4, whose perceptual reference medium differs so the same image renders differently, and embed against assign against convert. Assigning a profile reinterprets existing numbers and changes appearance; converting changes the numbers to preserve appearance.

Rendering intent

The style of mapping colour values from one description to another when gamuts don’t match. ICC.1:2022 defines exactly four: ICC-absolute colorimetric, media-relative colorimetric, perceptual and saturation. The colorimetric pair preserve in-gamut colorimetry at the expense of out-of-gamut colours. Perceptual and saturation modify colorimetric values to account for differences in device, medium and viewing conditions, and their exact rendering is vendor specific. Intent is a property of the transform, chosen at conversion or print time, and it gets thought of as a property of the file. An intent only exists if the destination profile carries tables for it, and asking for a missing intent makes the CMM substitute another silently. “Perceptual for photos, relative for logos” is a rule of thumb.

Rendering intent: perceptual

In perceptual transforms the PCS values represent a hypothetical reproduction on the ICC reference reflective medium. The intent compresses the whole source gamut into the destination and preserves tonal and colour relationships at the expense of exact values. It’s intended for pictorial images moving between substantially different media, and the ICC specification states plainly that the exact colour rendering of the perceptual intent is vendor specific. Perceptual gets read as “more accurate for photographs”, when what it does is move colours that were already reproducible in order to make room for those that weren’t. Two profiles for the same press from different vendors will render perceptually differently, and v2 and v4 perceptual results diverge. Some CMYK profiles carry poor perceptual tables that visibly lighten shadows.

Rendering intent: relative colorimetric

Called media-relative colorimetric in the ICC specification. It rescales in-gamut chromatically adapted tristimulus values so the actual medium’s white point maps to the PCS white point, leaves in-gamut colours otherwise alone and clips out-of-gamut colours toward the gamut surface. It’s the usual default for RGB-to-CMYK conversion in Adobe applications, and it’s normally paired with black point compensation. The name differs between spec and software: media-relative colorimetric in ICC.1, Relative Colorimetric in Photoshop. It gets confused with absolute colorimetric, where for most jobs the only practical difference is paper-white simulation. Used without black point compensation it blocks shadows badly, and ICC White Paper 40 says the intent “has little value without BPC”.

Rendering intent: saturation

A vendor-specific rendering that, in the ICC specification’s words, involves compromises such as trading off preservation of hue in order to preserve the vividness of pure colours. Historically it was intended for business graphics, so charts, diagrams, maps and flat solid fills, where a strong clean colour matters more than a faithful one. It isn’t intended for photographic content. Saturation gets mistaken for a saturation boost, as if it were a slider applied to the image, when it’s a separate lookup table in the profile. In practice many output profiles simply copy their perceptual tables into the saturation tag, so choosing it changes nothing, which is why some operators believe the intent doesn’t work.

Rendering intent: absolute colorimetric

ICC-absolute colorimetric transforms leave the chromatically adapted nCIEXYZ values of in-gamut colours unchanged, and unlike media-relative they don’t rescale to the destination medium’s white. On a proofer that means the proof stock gets tinted to simulate the target paper white. Where no separate transform exists it’s derived by scaling media-relative values by the ratio of source to destination mediaWhitePointTag. “Absolute” gets read as “most accurate”, and operators are then surprised by a grey-yellow cast across the proof, which is the whole point of the intent. The ICC spec itself notes this definition is what CIE terminology would call relative colorimetry, since data is normalised to a perfect diffuser. Black point compensation is generally disallowed with it.

Black point compensation

A conversion option that lives in the CMM. It linearly scales the source black point onto the destination black point in XYZ so the destination’s full dynamic range gets used and shadow separation survives. It originated in Adobe Photoshop in the late 1990s and is documented in ICC White Paper 40, which recommends it be available for perceptual and saturation as well, and enabled by default in non-interactive systems. Since the profile doesn’t carry it, the same profile gives different output with BPC on or off, a frequent cause of proof-to-print mismatch between two shops using the same ICC set. It isn’t black generation (GCR or UCR), and it isn’t setting a black point in Levels or Curves. Generally disallowed for ICC-absolute colorimetric.

White point

The chromaticity treated as achromatic white by a space or device. sRGB, Adobe RGB (1998) and Display P3 all use D65, while the ICC Profile Connection Space uses D50 at nCIEXYZ 0.9642, 1.0000, 0.8249. In print, white point usually means the substrate white recorded in the profile’s mediaWhitePointTag, the value media-relative colorimetric rescales to and absolute colorimetric simulates on the proof. Three different things share the name: the encoding white point of a colour space, the calibrated white point of a display, and the media white point of a substrate. It also gets confused with luminance or brightness, which is separate. Whether to calibrate a monitor to D50 or D65 is a live disagreement in the trade, with D65 plus visual adaptation to the booth being common practice and D50 being common advice.

Illuminant D50 and D65

CIE daylight illuminants. D50, correlated colour temperature about 5003 K at x 0.3457, y 0.3585, is the graphic arts reference, used for ISO 3664 viewing booths, ISO 13655 measurement and the ICC PCS adopted white. D65, about 6504 K at x 0.3127, y 0.3290, is the display reference used by sRGB, Adobe RGB and Display P3. ISO 3664:2009 sets 2000 lx for critical comparison (P1) and 500 lx for practical appraisal (P2). A 5000 K lamp isn’t automatically a D50 simulator. ISO 3664:2009 tightened the ultraviolet tolerances and the metamerism index precisely because cheap lamps passed on CCT and failed on spectrum. “D50” in an instrument specification refers to the computation illuminant under ISO 13655, which has nothing to do with the light in the room. The whole workflow expects a print judged at D50 to match a screen built at D65, which holds only because the eye adapts.

Spot colour

ICC.1:2022 defines it as a single colorant, identified by name, whose printing tone values are specified independently from values in a colour coordinate system. In practice it’s a pre-mixed ink on its own unit and plate. Used for colours outside CMYK gamut, for brand consistency, to run a job economically on a one- or two-colour press, and for metallics, fluorescents, opaque whites and varnishes. Spot gets equated with Pantone, though Pantone is one library of spot inks among several including HKS, Toyo, DIC and house mixes. A spot swatch set to convert on output has stopped being a spot colour while still looking like one in the file. In a PDF, a spot is any named separation, so a die line, a varnish or a technical layer may carry no visible ink at all.

Pantone / PMS

The Pantone Matching System: a proprietary library of named spot inks, each defined by a mixing formula from a set of base inks and shown in printed fan guides on coated and uncoated stock. The current Formula Guide set carries 2,390 colours, though the count is edition-specific and changes between releases. Suffixes identify the reference stock: C for coated, U for uncoated, CP and UP for process simulations. Color Bridge shows each solid beside its nearest CMYK build, an approximation that must not be quoted as the spot. PMS is the system, so “a PMS” is trade shorthand. The same number on C and U stock is a visibly different colour, so specifying PMS 286 without the stock is incomplete. Guides fade and want replacing regularly. Recent licensing changes removed Pantone libraries from Adobe apps, which breaks older files.

Duotone

In Photoshop, an image mode where a single 8-bit greyscale channel is printed with up to four inks, each carrying its own transfer curve. The family covers monotone with one ink, duotone with two, tritone with three and quadtone with four. It’s used to extend the tonal range of a monochrome image beyond what one black can hold, or to tint it. Only 8-bit greyscale images can be converted. Duotone gets confused with the fashionable two-colour gradient-map look, which is an RGB effect, and with “a two-colour job”, which may use two inks and contain no duotone image. The mode keeps one channel and multiple inks, which surprises operators expecting two channels. Ink order matters on press, and the file has to be saved as PSD, EPS or PDF to retain the ink and curve data.

Rich black

A solid black reinforced with percentages of cyan, magenta and/or yellow beneath it, so large black areas read dense instead of washing out to grey. Typical builds run around 60/40/40/100 or 65/50/50/100, chosen to sit safely under the printing condition’s total ink limit. Cyan-heavy builds read cool and magenta-heavy builds read warm. Suitable for large areas and display type; keep it off fine text. The build numbers are conventions and not standards, and the ink limit they have to sit under comes from the printing condition, with ISO 12647-2:2013 giving figures for offset. Rich black gets confused with registration black at 100/100/100/100, which belongs on marks only, won’t dry and sets off in the pile, and with plain 100K. Rich black under small text shows coloured fringes as soon as register drifts. The recipe is press-, ink- and substrate-specific, so ask the printer before reusing a house number.

K-only black

Black built from the black separation alone, 0/0/0/100, also called flat, plain or single black. It’s the standard for body text, rules and line work, because a single plate can’t misregister with itself, and it’s normally set to overprint. Over large areas it reads thin and greyish next to a rich black, since one ink film can’t reach the density four can. K-only gets confused with rich black in specification, and with what happens on conversion: a greyscale image or a K-only element placed in a CMYK document can separate into all four inks unless it’s converted through a black-preserving or K-only device-link. On screen it looks identical to rich black, so the problem gets caught at preflight.

Total ink coverage (TAC / TIL)

The sum of C+M+Y+K percentages at any single point. ISO 12647-2:2013 calls it tone value sum and specifies that on coated substrates it should be under 330% and shall not exceed 350% sheet-fed, and shall not exceed 300% heat-set web; for other substrates the figures are 300% sheet-fed and 270% heat-set web. The GRACoL 2013 (CRPC6) ICC profile is built to 320%. TAC, TIL, total ink limit, total area coverage and tone value sum are all the same quantity, which makes specs look inconsistent when they aren’t. It’s distinct from maximum black, the K limit, which is a separate control in profile building, and from solid ink density, which is a measurement. Exceeding TAC causes set-off, poor trapping, picking and drying failures.

GCR (grey component replacement)

A separation strategy that replaces the neutral component of any CMY combination with black across the whole tone scale, not only in the shadows. It reduces total ink and improves grey stability and run-to-run repeatability, because neutrals then depend chiefly on one ink instead of three held in balance. Strength is selectable from light to maximum, and the GRACoL 2013 profile uses medium-plus black generation with maximum black at 100%. GCR gets confused with UCR, which acts only in dark neutrals, so the difference is one of scope. Heavy GCR shrinks the gamut in dark saturated colours and makes later re-separation or spot-colour matching harder. GCR is fixed when the profile or separation is built and can’t be adjusted downstream. Photoshop’s Black Generation in Custom CMYK is the same control under another name.

UCR (under colour removal)

Reduces cyan, magenta and yellow in dark neutral areas and substitutes black, holding total ink down in the shadows so the sheet dries and traps properly. It acts only in the dark neutrals, and the rest of the tone scale keeps its full CMY build. Adobe’s documentation puts UCR on newsprint and uncoated stock, because the separation lays down less ink overall. UCR gets confused with GCR, UCR being the shadow-only case, and with UCA, under colour addition, which puts CMY back into deep shadows after heavy GCR to restore density. In casual use UCR gets applied to any ink-limiting operation including a RIP’s ink limit, which is a different mechanism with different consequences for hue.

Measuring colour

Delta E (ΔE)

The numerical colour difference between two measurements in CIELAB. ΔEab (CIE76) is Euclidean distance in Lab. ΔE00 (CIEDE2000) adds lightness, chroma and hue weighting functions plus a hue-rotation term for blues, and tracks what the eye sees far better. ISO 12647-7 moved proof tolerances from ΔE*ab to ΔE00 in its 2016 revision. A ΔE00 near 1 is about the threshold of perceptibility.

Qualify any ΔE in a specification. ΔE76 3.0 and ΔE00 3.0 aren’t the same tolerance, and CMC(l:c) is a third formula common in textiles. Numbers only compare when the measurement condition matches, since M0 and M1 readings of one patch differ on brightened stock. ΔE gets confused with ΔH, hue difference alone.

Metamerism

Two samples with different spectral reflectance curves that match under one illuminant and observer but diverge under another. Illuminant metamerism is the common case, where a match under D50 fails under store lighting or illuminant A. Observer metamerism is the match that holds for one person and not another. It turns up whenever a spot ink is simulated with a CMYK build, or a proof and a print use different colourants.

Strictly, metamerism is the phenomenon that lets different spectra match at all. The trade uses the word for the failure, properly called metameric mismatch. It also gets confused with fluorescence from optical brightening agents in paper, a UV problem addressed by M1 measurement, and with plain colour inconsistency across a run.

Densitometer

Instrument that measures optical density, the base-10 log of the reciprocal of reflectance, through wide-band filters. ISO 5-3 defines the spectral status responses, chiefly Status T for North American graphic arts and Status E in Europe, and ISO 5-4 fixes the 45:0 or 0:45 reflection geometry. On press it reports solid ink density, tone value, dot gain, ink trap and print contrast.

Density isn’t colour. Two inks can share a density reading and look plainly different, so a densitometer controls consistency and says nothing about accuracy. Status T and Status E give different numbers for the same patch, with the gap largest on yellow, so a density spec is useless without the status. It’s also distinct from a colorimeter, which is tristimulus with no spectral data, and from ink film thickness, which isn’t measured directly.

Spectrophotometer

Measures spectral reflectance across the visible band, typically in 10 nm steps, and computes CIELAB, ΔE, density and tone value increase from it. ISO 13655 defines the measurement conditions: M0 unfiltered incandescent-like source, M1 D50 including the UV component for brightened stock, M2 UV-cut, M3 UV-cut plus polarising filter. Graphic arts colorimetry uses D50, the CIE 1931 2° observer and 45:0 or 0:45 geometry.

It gets confused with a spectrodensitometer, which is a spectrophotometer that also reports status densities, and with a colorimeter, which uses filters and can’t detect metamerism. When two shops disagree, check the measurement condition before blaming the instrument, since M0 against M1 on optically brightened paper is a real difference. Backing, white against black, also shifts readings and has to be stated.

Colour bar / control strip

A strip of defined patches printed outside the trim, usually across the sheet at the tail or gripper margin, so the crew can measure the sheet. It typically carries solids, tints such as 25/50/75%, two- and three-colour overprints, grey balance patches and slur and register targets, aligned to the press ink zones. Its proofing equivalent is the Ugra/Fogra Media Wedge CMYK V3, 72 patches, laid out to ISO 12647-7.

A press bar and a proof media wedge do different jobs. The bar controls a live run zone by zone, and the wedge verifies a proof against a printing condition. The bar also gets confused with registration marks and other printer’s marks, which carry no measurable data. A bar that doesn’t line up with the ink zones can’t be used for zone control, and it has nothing to do with SMPTE broadcast colour bars.

Colour cast

An unwanted overall tint across an image or press sheet, seen first in areas that should be neutral. On press it usually traces to ink and water balance, density drift or grey balance being off. In a file it traces to capture white balance or a wrong source profile. Diagnose it by measuring supposed neutrals, where a cast shows as a* and b* departing from zero.

A cast gets confused with a deliberate warm or cool grade, so ask before correcting. A cast that appears only under some lighting is metamerism or an OBA fluorescence effect. Blue-white brightened stock reads as a cast and is the substrate. Judging a sheet under anything other than D50 produces an apparent cast that isn’t in the ink.

Screening, dots and tone

Halftone

Reprographic method that renders continuous-tone imagery as discrete dots varying in size or spacing, read by the eye as smooth tone. ISO 12647 speaks of half-tone colour separations, and its tone value is calculated from a reflectance measurement. Halftone dots are built from many smaller device spots: at 2400 dpi a 150 lpi cell is 16 x 16 spots, giving (2400/150)^2 + 1 = 257 grey levels.

Halftone dot, device spot and pixel get run together. A 2400 dpi platesetter doesn’t produce 2400 halftone dots per inch, it produces 2400 addressable laser spots clustered into halftone dots. Designers also use “halftone” loosely for the deliberately coarse dot effect.

Screen ruling / LPI

Frequency of the halftone grid, in lines (rows of dots) per inch, or per centimetre in European and ISO usage, where you multiply cm-1 by 2.54 to get lpi. ISO 12647-2:2013 clause 4.2.3 puts four-colour offset at 48 to 80 cm-1 (120 to 200 lpi) on coated stock and 48 to 70 cm-1 (120 to 175 lpi) on uncoated. Newsprint typically runs 85 to 100 lpi.

Lpi, dpi and ppi are three different measures: lpi is the halftone screen, dpi the output device’s addressable spots, ppi the image’s own pixel density. Convention puts image resolution at 1.5 to 2 times lpi, so 300 ppi for 150 lpi. “Line screen” means the same as screen ruling, and also names a line-shaped halftone dot.

Screen angle

Orientation of an AM halftone grid, in degrees. ISO 12647-2:2013 clause 4.2.5 says that for dots with no principal axis, circular or square, the nominal difference between cyan, magenta and black should be 30 degrees, with yellow separated 15 degrees from another colour and the dominant colour at 45 degrees. For dots with a principal axis such as elliptical, the separation is 60 degrees and the dominant colour sits at 45 or 135 degrees. A common set is K 45, M 75, Y 0, C 15.

C 15 and C 105 name the same screen for a symmetrical dot, which repeats every 90 degrees. An elliptical dot repeats every 180 degrees, which is why the trade quotes two apparently different angle sets. Vendors also disagree on writing yellow as 0 or 90. Nothing to do with mesh angle in screen printing.

AM screening

Amplitude-modulated screening, also called conventional, periodic or clustered-dot. Dot centres sit on a fixed grid defined by the screen ruling and angle, and dot area grows with tone value. ISO 12647-2 calls these periodic screens and permits circular, square or elliptical dot shapes. For dots with a principal axis, first link-up should occur no lower than 40 per cent tone value and second link-up no higher than 60 per cent.

AM, conventional, periodic and clustered-dot all name the same thing, and ISO’s own word is periodic. The link-up rule is why elliptical or chain dots exist: a circular dot joins its neighbours on all four sides at once around 50 per cent and produces a visible tone jump.

FM / stochastic screening

Frequency-modulated or non-periodic screening: dots of essentially fixed size, pseudo-randomly dispersed, with the number per unit area carrying the tone. ISO 12647-2:2013 clause 4.2.4 gives spot sizes of 20 to 40 micrometres for four-colour work, 20 to 30 on coated and 30 to 40 on uncoated. First-order FM keeps every dot the same size, second-order FM lets dots grow once the area is filled. Having no angle, it produces no rosette and no screen moiré.

FM gets confused with hybrid or cross-modulated screening such as Agfa Sublima, which is AM in the mid-tones and FM in highlights and shadows, and which sales material often calls FM. “Stochastic” overstates the randomness, since placement is a deterministic dispersed pattern. FM raises tone value increase sharply, so ISO 12647-2 assigns non-periodic screens their own aim curve.

Dot gain / tone value increase (TVI)

The arithmetic difference, in percentage points, between the tone value requested and the apparent tone value measured on the print, so a 50 per cent patch reading 65 per cent is 15 per cent TVI rather than 30 per cent. Apparent tone value is derived from reflectance by the Murray-Davies equation, measured per ISO 13655 with D50 illuminant, 2 degree observer and 45:0 or 0:45 geometry. Mechanical gain is physical dot spread, optical gain is light scatter within the substrate.

ISO deprecated “dot gain” in favour of tone value increase because non-dot processes also gain. The measured TVI and the compensation curve applied to correct it are opposites, and both get called the dot gain curve. Aim-curve letters are edition-dependent: the 2004 curves A to F were 13, 16, 19, 22, 25 and 28 per cent at 40 per cent, and the 2013 edition reletters them, so a curve letter with no edition is ambiguous.

Moiré

Low-frequency interference or beat pattern produced when two or more periodic structures overlap at close frequencies or angles. In four-colour AM work, screens less than 30 degrees apart are at risk. Yellow lies only 15 degrees from cyan, so most screening systems run yellow at roughly 108 per cent of the other frequency, and ISO 12647-2 notes frequency differences of up to 6 per cent between C, M and Y for the same reason.

Screen moiré, a conflict of angle or frequency between separations, gets confused with subject moiré, where image content such as fabric, mesh or an already-printed original being rescanned beats against the screen. FM screening cures the first and does nothing for the second.

Rosette

The regular flower-shaped pattern formed where cyan, magenta and black AM screens overprint at 30 degree intervals, a deliberate moiré whose frequency is high enough to be invisible at normal reading distance. Two forms exist, clear-centred (open, the usual choice) and dot-centred (closed). Yellow, only 15 degrees away and often at a higher frequency, doesn’t join the pattern. Drifting between the two forms across a sheet reads as a colour shift.

A rosette is moiré of the wanted kind, which is where the confusion with unwanted moiré comes from. A rosette visible at normal viewing distance means the screen ruling is too coarse for that distance. Rosettes are AM-only, and FM screening has none, which is why FM work looks different in flat tints.

Continuous tone

An image whose tone varies smoothly with no discrete structure, as in photographic emulsion or dye sublimation. In prepress the term also covers the unscreened digital image handed to the RIP, where 8 bits per channel gives 256 levels and 16 bits gives 65,536. Screening converts that data into the binary on/off information a plate or press can carry.

Contone as a physical ideal, meaning film, gets confused with contone as a file state, where a 300 ppi 8-bit scan is quantised and pixelated. Contone-capable proofers and inkjets vary drop size or density per addressable location and so need no halftone screen, and the label says nothing about resolution.

Line art / bitmap

1-bit raster data in which every pixel is either fully inked or blank, with no intermediate values, so it needs no screening and images directly to the plate. Photoshop calls this colour mode Bitmap. Because edges can’t be softened, resolution has to be high, with 1200 ppi the usual print minimum and sometimes matched to platesetter resolution. Page layout applications can colourise a placed 1-bit image.

Bitmap as the 1-bit colour mode, bitmap meaning any raster image at all, and the .BMP file format are three different things carrying one word. Line art in this sense is raster, so a scan of it has fixed resolution and pixelates on enlargement. Traced vector artwork gets called line art too.

Anti-aliasing

Smoothing of stair-stepped raster edges by giving boundary pixels intermediate values in proportion to coverage. Useful on screen and within continuous-tone images, and a liability in prepress, where anti-aliased type or logos placed as raster carry grey fringes that screen into halftone dots, softening what should be a hard edge and undermining trapping. It can’t exist in 1-bit data. RIPs also anti-alias internally when rendering to lower device resolutions.

Anti-aliasing is softness already in the file, and dot gain is softness added on press. It’s also distinct from Photoshop’s diffusion-dither bitmap conversion, and from Acrobat’s on-screen smoothing, which is a display setting with no presence in the PDF.

Banding

Visible discrete steps where a gradient should read as smooth. Causes include too few levels in the file, since 8-bit gives 256 per channel, a blend stretched over too long a distance, correction curves compressing tones, and too few grey levels available in the screen itself. That last is (device resolution / screen ruling)^2 + 1, so 2400 dpi at 150 lpi yields 257 levels and only 145 at 200 lpi. Adding noise or working in 16-bit breaks the steps up.

Tonal banding in the data gets confused with mechanical banding on inkjet and digital presses, which is regular streaking caused by nozzle, drum or feed defects and is a hardware fault no file change will fix. Customers report both as banding. It’s also called posterisation, though posterisation normally describes deliberate level reduction or bit-depth loss.

Overprint, trapping and register

Overprint

Instruction that an object’s ink prints on top of the inks beneath, so colourants add and 100 per cent cyan over 100 per cent yellow gives green. In PDF it’s carried by the OP (stroke), op (fill) and OPM entries of the graphics state, per ISO 32000. Setting OP alone sets both parameters unless an op entry is present in the same dictionary, in which case OP applies only to strokes. PostScript and PDF knock out by default, so overprint has to be set explicitly. Used for black text, trapping strokes and deliberate ink mixes.

Overprint gets confused with the multiply blend mode. Multiply is a transparency operation resolved at flattening; overprint is a separation-level ink instruction. They look alike on screen and behave differently on plate. Overprint applied to white makes the object disappear, and overprint preview has to be switched on or the display lies about the result.

Knockout

The default behaviour, where an object removes the underlying inks within its own area so it prints on bare substrate. Colours stay clean and predictable, and any misregistration exposes a paper-white sliver at the join, which is why knockouts are trapped. PostScript and PDF knock out unless overprint is explicitly set, a reversal of pre-PostScript practice where overprinting was the norm.

Knockout as ink behaviour gets confused with the Knockout entry of a PDF transparency group, a different mechanism entirely. In an all-CMYK file with overprint mode 0, overprinting is visually identical to knocking out. The difference shows with spot colours, or with zero-valued components once overprint mode 1 is in play.

Trapping

Deliberate small overlap built between abutting colours so a slight misregistration shows no paper gap. Typical widths run 0.05 to 0.2 mm. InDesign’s built-in trapping defaults to a 0.25 pt trap width, about 0.09 mm, and 0.5 pt against solid black, with black defined by a neutral density of 1.6 or above. Convention has the lighter colour moving into the darker, so the darker colour keeps its edge shape.

The word carries two unrelated meanings. Prepress trapping is horizontal overlap built into the file. Ink trapping, or wet trapping, is a press phenomenon covering how well a wet ink film accepts the next ink laid over it, quantified by Preucil’s 1958 densitometric equation. ISO 12647-2 uses the phrase in that second sense when it warns about poor ink trapping at high tone value sum.

Choke

Trap in which the lighter background is expanded inward into a darker object lying within it, so the knockout hole is cut fractionally smaller than the object and the object looks very slightly squeezed. That’s Adobe’s definition in the InDesign and Illustrator documentation. Older trade names for the same move are shrink, thin or minus.

Direction follows which colour is lighter. Stacking order doesn’t decide it, which is where choke and spread get swapped. Trade usage is genuinely inconsistent: some shops name the trap after the element that was modified, others after the apparent effect on the foreground object, so the same job can be called a choke by one operator and a spread by another. Confirm which is meant before setting a value.

Spread

Trap in which a lighter foreground object is enlarged slightly so it laps into the darker background around it. The object looks fractionally expanded and the darker background keeps its visual edge. That’s Adobe’s definition in the Illustrator and InDesign documentation. Older trade names are fatty, plus or grow.

Spread and choke are one decision viewed from opposite ends, both chosen so the darker colour retains the edge the eye reads. Spread also collides with the layout sense of a two-page spread, and with mid-tone spread in ISO 12647-2, the difference in tone value increase between the process colours, which is toleranced separately.

Registration

The accurate positioning of each printed colour relative to the others so they combine as intended. ISO 12647-2:2013 clause 4.3.5, tolerancing for image position, requires that the largest deviation between any two printed colours not exceed 0.10 mm. Registration also names the swatch made of 100 per cent of every separation, used for marks that must image on all plates.

Three different things get called registration: colour-to-colour register, image position or fit on the sheet, and front-to-back backup. The registration swatch is neither rich black nor 100K, and artwork coloured with it lays down 400 per cent ink and won’t dry. Register and registration are used interchangeably, with ISO and British trade usage preferring register as the noun.

Misregistration

Failure of separations to align, showing as white slivers between abutting colours, coloured fringes on edges, blurred or doubled type, and hue shifts where overlapping tints move relative to each other. Causes include press mechanical variation, plate mounting, blanket and cylinder condition, and paper stretch or fan-out from moisture and web tension. Trapping conceals small amounts of it without correcting the underlying error.

It gets confused with doubling and slur, which are dot defects on a single printing unit. Doubling is a faint second impression, slur a directional smear, and both mimic register error on the sheet. What gets blamed on failed trapping is usually a knockout that was never trapped at all. Flexo practice ties acceptable register to line screen, so there’s no fixed millimetre figure.

Registration marks

Targets imaged on every separation so colours can be aligned, typically a circle of around 5 mm carrying crosshairs, coloured in registration at 100 per cent of every ink, placed outside the trim area and cut away in finishing. The 5 mm circle is common practice and isn’t a standardised dimension. Registration marks are distinct from trim or crop marks, bleed marks, fold marks and the colour control bar, each of which serves a different purpose.

Almost any printer’s mark gets called a registration mark, trim marks most often. They also get confused with the micro-register targets built into the colour control bar for on-press measurement, and with the fiducial marks read by die-cutters, laminators and converting cameras, which are placed for machine vision.

On press

Substrate

Whatever is being printed on: paper, board, self-adhesive laminate, film, foil, metal, glass or textile. The word is deliberately neutral about material, which starts to matter once a shop runs anything besides paper. Stock is the paper-specific synonym. Substrate choice drives ink set, cure method, adhesion, drying and the achievable colour target.

Substrate, stock and media get used interchangeably. Only substrate reliably covers non-paper materials, and a quote specifying media may mean the vendor’s own consumable list. In lamination, substrate can mean the base layer specifically, with the applied film called the overlaminate.

Coated and uncoated stock

Coated stock carries a mineral layer, typically kaolin clay or calcium carbonate, that seals the surface so ink sits on top as a defined dot. You get better ink holdout, higher gloss, lower tone value increase and a wider gamut. Uncoated absorbs and spreads, and gains more. ISO 12647-2:2013 codifies eight paper types, PS1 to PS4 coated and PS5 to PS8 uncoated, with TVI curves A, B and C.

The coating goes on at the mill. It isn’t the varnish, aqueous or lamination added after printing, and matte coated is still coated. Pantone C and U numbers are the same ink on two stocks, and the U book shows how much duller it prints.

GSM / paper weight

Grammage: mass per unit area in grams per square metre, determined to ISO 536. Because A0 is exactly one square metre, an A0 sheet of 80 gsm stock weighs 80 g. North America quotes basis weight instead, the weight in pounds of a 500-sheet ream at that grade’s basic size. Basic size varies by grade, so 70 lb text and 70 lb cover aren’t equal.

Gsm measures weight, never thickness. Caliper (microns, or points meaning thousandths of an inch) and bulk are separate properties, so a bulky uncoated 100 gsm can be thicker than a coated 130 gsm. Basis weight can’t be converted to gsm without knowing the grade’s basic size.

Grain direction

The alignment fibres take as the wet web travels through the paper machine, running in the machine direction. A sheet is long grain when the grain is parallel to its longer dimension, short grain when it runs across it. Paper folds cleanly and is stiffer along the grain, and cracks or tears when folded across it.

Mills say MD and CD, printers say long and short grain, and they mean the same axis. Long and short are named for the sheet dimension, so the same reel yields either depending on how it’s cut. Grain has nothing to do with laid or wove surface texture.

Offset lithography

Planographic printing: image and non-image areas lie in the same plane on the plate and are separated chemically, with no relief. Fountain solution wets the desensitised non-image areas so they reject oil-based ink. The inked image transfers from plate to a rubber blanket and from the blanket to the substrate, which is the offset step, protecting the plate and letting it print rougher stock.

Offset names that plate-to-blanket-to-sheet transfer. Set-off is the smearing of wet ink onto the next sheet, yet the trade calls set-off “offsetting”. Litho and offset get used as synonyms, with litho the imaging principle and offset the transfer. Waterless offset uses plates with an ink-repellent silicone layer and no dampening, relying on temperature-controlled inking instead.

Digital press

A press that images every impression from the raster held in memory, with no fixed image carrier, so each sheet may differ. That’s the basis of variable data printing. Two families dominate: electrophotography, either dry toner or liquid such as HP Indigo’s ElectroInk, and production inkjet. No plates means no plate cost and minimal make-ready, which shifts the crossover point against offset.

A digital press isn’t the same thing as “digital printing” meaning a digital prepress workflow, since CTP is fully digital and still feeds an offset press. HP Indigo transfers via a blanket and gets mistaken for offset regularly. Against an office copier, a production digital press differs in registration, substrate range and colour control.

Sheetfed

A press fed pre-cut sheets one at a time, lifted from the pile by suction heads and registered against front lays and a side lay before being gripped through the units. Sizes commonly run from B3 up to B1 and beyond. Sheetfed offset suits short to medium runs, heavy stocks, and work needing tight register or special finishes.

Sheetfed describes the feed, so a sheetfed press may be offset, digital, letterpress or screen. Sheet size isn’t trim size: an SRA1 sheet carries several trimmed pages plus gripper edge, bleed and marks. The gripper edge is unprintable and has to be allowed for at imposition.

Web

A press fed from a reel, printing a continuous ribbon of paper, the web, and usually cutting and folding it in line to deliver folded signatures. Heatset web drives solvent off in a hot-air dryer followed by chill rolls, so it can run coated stock. Coldset, or non-heatset, dries by absorption and is limited to uncoated grades such as newsprint.

Web here means the paper ribbon, and specifications sometimes read ambiguously against the internet sense. Web-fed isn’t synonymous with web offset, since flexo, gravure and inkjet also run webs. Heatset and UV are different curing routes. Web make-ready waste is far higher than sheetfed, which is why the crossover run length gets quoted in thousands of impressions as a trade rule of thumb.

Printing plate

The image carrier that transfers ink to the blanket or substrate. A lithographic plate is grained and anodised aluminium coated with a light- or heat-sensitive layer which, once imaged and processed, presents a hydrophobic ink-receptive image area and a hydrophilic non-image area. One plate is needed per separation per side, so a four-colour sheetwise job takes eight.

A plate isn’t a separation: the separation is the single-colour image data, the plate is the metal carrying it. It isn’t the blanket either, which is the rubber intermediate. It has nothing to do with the die in die cutting. In gravure the equivalent carrier is an engraved cylinder.

CTP (computer to plate)

Imaging offset plates directly from digital data on a platesetter, with no film and no contact exposure. Standard practice since the late 1990s. Two laser families dominate: thermal at 830 nm, daylight-safe and long-running, and violet at 405 nm, faster to image with shorter typical run lengths. Processless and chemistry-free plates skip the developer stage entirely.

CTP gets confused with digital printing, when what it feeds is a conventional offset press. It’s distinct from CTF, computer to film, where an imagesetter outputs film that’s then contacted to a plate, and from direct imaging presses that image the plate on-press. Platesetter and CTP device name the same machine.

Make-ready

Everything done to bring a press from idle to a signed sheet: hanging plates and blankets, setting ink keys and dampening solution, adjusting ink and dampening rollers and cylinder packing, setting feeder and delivery, and running sheets until colour and register are right. It ends when the customer or CSR signs the OK sheet. The sheets spoiled along the way are makeready waste.

Make-ready is the operation, and the setup charge on the quote is what gets billed for it. Makeready sheets aren’t overs, which are good sheets run beyond the order quantity to cover finishing loss. In diecutting and letterpress, make-ready also means packing and shimming the die or forme for even impression.

Set-off

Wet ink transferring from the face of a printed sheet onto the back of the sheet landing on top of it in the delivery pile. It’s controlled with anti-set-off spray powder, food-grade starch from corn, potato, wheat or rice in graded particle sizes, which spaces the sheets and admits the oxygen that oxidative-drying inks need. Lower pile heights and IR dryers help too.

Set-off isn’t offset, which is the plate-to-blanket-to-sheet transfer, although the trade routinely calls set-off “offsetting”. It also isn’t show-through or strike-through, which is ink visible or migrating through the sheet itself, and it isn’t ghosting, which comes from ink-train starvation or drying-related image repetition.

Proofing

Proof

Any representation of the job produced for checking or approval before or during the run. Content proofs verify text, images, page order, folding and marks, and include plotter and imposition proofs and the older bluelines. Colour proofs verify appearance against a defined printing condition. A signed proof is normally the contractual reference for what was authorised.

A plotter or imposition proof isn’t colour-accurate and should never be judged for colour. A press check is a separate thing, since the OK sheet comes off the running press. A dummy is an unprinted mock-up made for bulk, feel and spine width, and isn’t a proof.

Soft proof

On-screen simulation of a printing condition using ICC profiles: a source profile, a destination profile and a rendering intent, with black point compensation and paper-simulation options. ISO 12646 sets the display requirements, with white point at D50 chromaticity (u’ 0.2092, v’ 0.4881), white luminance at least 80 cd/m² and preferably 160 cd/m², and limits on uniformity, glare and surround.

Graphic arts works to D50 while sRGB, video and most consumer displays are D65, so a screen at factory settings can’t match a proof. “Colour profile” loosely means the ICC profile; the rendering intent decides how out-of-gamut colour is mapped. A soft proof isn’t a contract proof and carries no approval weight.

Hard proof / contract proof

A physical print simulating a defined printing condition, produced to ISO 12647-7. It has to carry a control wedge, commonly the Ugra/Fogra Media Wedge CMYK, and a job label naming the proofing system, media, ink, the printing condition simulated, the profiles used, and the date and time. The 2016 revision moved to CIEDE2000, with average ΔE00 of 2.5 or less and maximum ΔE00 of 5.0 or less across the wedge.

Any inkjet print off a proofer is a hard proof; only one measured and passing against the wedge is a contract proof. A press proof is pulled on an actual press. Judge a contract proof under ISO 3664:2009 D50 at P1, 2000 lx, because office lighting voids the comparison. “Certified” means the wedge has been measured and the results reported; a calibrated proofer isn’t automatically certified.

Finishing and converting

Imposition

Arranging pages on the press sheet in the positions and rotations that give correct reading order after folding, stitching and trimming, with the marks, bars and fold guides the pressroom and bindery need. Schemes include sheetwise, work-and-turn and work-and-tumble. Imposition software positions pages from the TrimBox and applies creep, gutters and bleed allowance.

Pagination gets confused with it: pagination is reader page order, imposition is sheet position. N-up ganging repeats one item across a sheet and has no folding scheme. An imposed PDF is a press-sheet file and mustn’t be re-imposed or re-trimmed downstream.

Creep

Also called push-out, thrust or binder’s creep. In nested saddle-stitched work the inner sheets project further at the fore-edge than the outer ones, by roughly the accumulated caliper of the sheets wrapped outside them, so trimming costs the inner pages fore-edge margin. The correction is shingling: page content shifted progressively spinewards, inner pages moved most. It’s usually applied above about 24 pages.

Trade formulas disagree. Some compute (sheets - 1) x caliper, others double that to cover both halves of the fold, so check what your imposition software assumes. Creep applies to nested work only; gathered perfect-bound signatures don’t get it. Shingling names both the defect and the correction.

Gutter

Three current senses, and the trade hasn’t settled on one: the inner binding margin of a spread, the space between text columns, which is what InDesign’s Columns and Gutter control sets, and in imposition the space between adjacent pages on the press sheet where the cut or fold falls. Always establish which is meant before quoting a millimetre figure.

The column sense and the binding sense collide inside InDesign itself, which uses the word for both. Gutter is also distinct from the binding allowance added for perfect binding, which compensates for pages curving into the spine and not lying readable.

Signature / section

A printed sheet folded to produce a run of pages, commonly 8, 16 or 32, always a multiple of four. British binderies tend to say section and the pressroom says signature, and the terms are now treated as synonyms. Originally the signature was the letter or figure printed at the foot of the folded unit so the binder could verify gathering order.

Two confusions. The forme is the set of pages printed on one side of the sheet, so a sheetwise 16pp job has two formes and one signature. The sheet itself may carry more than one signature and be cut apart before folding.

Guillotine

A cutting machine comprising a flat metal bed, a long steel or carbide knife, a cutting stick embedded beneath the knife line, a clamp that presses air out of the stack, fixed side guides and a moveable back gauge. It cuts a whole lift in one stroke, one straight cut at a time. Used to cut sheets down and to trim flat work.

The three-knife trimmer is a different machine, trimming head, foot and fore-edge of a bound book in one cycle at the end of a binding line. Cutting generally means reducing to sheet size, trimming means cutting to finished size, and the words get swapped frequently.

Folding

Buckle folders drive the sheet against a stop in a fold plate so it buckles into a pair of rollers, which is fast and suited to lighter stock and parallel folds. Knife folders push a blade down through the sheet between rollers, giving right-angle folds at 90 degrees to the previous fold, and handle heavier stock better. Combination folders carry both unit types.

Fold names aren’t standardised: roll, barrel and letter fold usually mean the same thing, while Z fold and concertina get used loosely. Count panels, since a six-page roll fold has two folds. Folding cross-grain cracks coated stock, so crease first.

Scoring / creasing

Two operations the trade often conflates. A score reduces stiffness along a line, typically by partial penetration of a knife or rule into the substrate. A crease compresses a hinge into the board with a rounded rule pressed into a matrix channel, the matrix width chosen against the caliper in thousandths of an inch. Creasing is the correct choice for coated stock and heavier boards.

The words appear interchangeably on quotes and job tickets, so confirm which is meant. A knife score cuts fibres and can crack or split on the outside of the fold, while a crease displaces them to form a hinge. Creasing against the grain still cracks. Both are distinct from perforation, which cuts through intermittently.

Die line

The vector template describing every cut, crease, perforation, glue tab and window of a converted piece, drawn on its own layer in named spot colours, set to overprint and flagged non-printing. ISO 19593-1:2018 standardises this in PDF as processing-step layers, where a die line belongs to the Structural group under the processing step Cutting.

The die line gets confused with the physical die it specifies, and with the structural engineer’s drawing of the flat blank, which is the same geometry expressed once as manufacturing data and once as artwork. Leave a die line as an ordinary separation and it prints as a coloured outline on the finished job.

Die cut

Cutting a substrate to shape with a shaped die. Flatbed die cutting presses a steel-rule die, hardened steel strip set into a slotted plywood board along the cut and crease paths, into the sheet. Rotary die cutting uses a cylindrical die for high-speed roll work. Kiss cutting cuts only the face layer of a self-adhesive laminate.

Laser and digital knife cutting follow the same die line, need no tooling and cost quite differently. Kiss cut, through cut and perforation are separate operations routinely lumped together as die cut. In a combination die, creasing rule is set lower than cutting rule.

Perfect binding

Adhesive binding of a gathered book block: the spine folds are milled off and the spine roughened to expose fibre, adhesive is applied, a wrap cover is drawn on, and the three open edges are trimmed. EVA hot melt is the cheap standard. PUR reactive polyurethane goes on thinner, stays flexible and holds coated and cross-grain stock far better.

PUR is an adhesive, so “PUR binding” means perfect binding using PUR. Notch binding retains the folds and cuts notches into them, and Smyth or section sewn uses thread. The constraint is a minimum spine thickness of around 1.5 mm (0.06 in); page count doesn’t come into it.

Saddle stitch

Folded sections are nested one inside another, laid over a saddle and stitched through the spine fold with wire staples driven from outside and clinched inside, then trimmed on three sides. Page count is always a multiple of four. Most printers cap it near 64 pages on ordinary text stock, a convention with no standard behind it, and creep compensation is needed well before that limit.

Stitch here means wire, and sewing is a different operation. Side stitch staples through the face near the spine and won’t open flat. A saddle-stitched book has no flat spine to print on, and the outermost sheet carries both the first and the last page.