Choose a DTF halftone by matching screen geometry to the file’s effective density, previewing the actual alpha-screening method at full resolution, and printing a controlled matrix. In NestSheet, LPI maps to a pixel cell only when embedded or NestSheet-assumed DPI is available; angle applies only to Ordered and AM Dot. No LPI or angle is “best” until the complete print path is tested.
First, identify what NestSheet is screening
The Halftone step in Main Image Prep changes artwork alpha. It does not separate CMYK colour plates and it does not set a printer’s native screening. The RGB colour samples remain the colour artwork while the tool changes which pixels are opaque, transparent, or distributed as an opacity pattern.
That scope is different from the white-underbase workflow, where white is a derived production property regenerated for export. It is also different from diagnosing a pale edge: the white-halo guide separates dirty alpha, underbase geometry, and directional registration before any screening decision.
This article assumes, unless a test record says otherwise:
- raster artwork is in NestSheet’s 8-bit RGBA working path;
- final width and height are fixed before comparing screens;
- “DPI” in the LPI mapping means embedded or NestSheet-assumed DPI, while effective image PPI is checked at the intended physical size;
- method, LPI or pixel cell, angle, target, output container, channel representation, and target RIP/version remain fixed except for the one variable under test; and
- physical conclusions name the printer resolution/mode, ink, film, powder, cure, press, and substrate used. A screen preview alone establishes none of those outcomes.
The broader DTF file-requirements guide owns source resolution, physical size, transparency, and container intake. Lock those inputs before comparing halftones.
Keep PPI, DPI, and LPI separate
These units answer different questions. Substituting one for another creates precise-looking settings with no stable physical meaning.
| Unit | What it describes here | Calculation or check | What it does not prove |
|---|---|---|---|
| Image PPI | Source pixels distributed across the intended print size | pixels ÷ placed inches | Printer dot size, ink drop, or surviving film detail |
| Printer DPI | The selected device resolution or addressing mode | Read from the exact printer/RIP queue record | Source detail or NestSheet’s screen cell |
| Halftone LPI | A requested screening frequency in the Prep control | Used with density to derive a pixel cell | An independently measured physical line screen on film |
| Pixel cell | The raster cell passed to the selected method | clamp(round(DPI ÷ LPI), 2, 64) | A universal production-safe dot |
For example, the same requested frequency produces different raster cells when the assumed density changes. These are reproducible mappings, not recommended DTF settings:
| Embedded or assumed DPI | Requested LPI | NestSheet calculation | Effective cell |
|---|---|---|---|
| 300 | 50 | round(300 ÷ 50) | 6 px |
| 300 | 75 | round(300 ÷ 75) | 4 px |
| 600 | 50 | round(600 ÷ 50) | 12 px |
The mapping is clamped to 2–64 px. Ordered adds another important boundary: it uses an 8×8 Bayer matrix and holds each matrix position for the calculated cell size, so its full repeat spans eight cells. “DPI ÷ LPI” is therefore NestSheet’s configured geometry mapping, not proof that a microscope would measure that exact period on film.
Read the current controls literally
Main Image Prep exposes these labels and ranges today:
| UI control | Current choices or range | Applicability |
|---|---|---|
| Method | ORDERED · AM DOT · FS · FM | All four alpha-screening methods |
| Frequency | 20–200 lpi, with a displayed pixel cell | Shown when density is available |
| Dot size | 4–16 px | Fallback when density is unavailable |
| Angle | 0–90° | Shown only for ORDERED and AM DOT |
| Apply to | SEMI-TRANSPARENT · FULL IMAGE | Selects the alpha/luminance source region |
Without usable density, the interface does not pretend that pixels are lines per inch. It shows Dot size and the message “No DPI in file — dot size set in pixels.” Do not convert that fallback to a physical claim unless the actual final density and size are established elsewhere.
A fresh Halftone panel starts at Ordered, 6 px, Semi-transparent, 45°, with persisted LPI off. When density is available, the Frequency row derives a visible position from that cell size; moving the control stores an LPI value. Those are product starting states, not print recommendations. Older saved steps without geometry remain at angle and LPI zero so merely opening them does not rewrite historical output.
What the four methods actually do
The method names are not interchangeable style presets. They produce different binary opacity structures from the same source signal.
| Method in Prep | Current product behavior | Angle? | What the cell control affects | Useful comparison question |
|---|---|---|---|---|
| ORDERED | Compares each processed alpha sample with a repeating 8×8 Bayer threshold pattern | Yes | Scale of the repeating threshold cells | Does the regular pattern preserve the intended tonal transition at final size? |
| AM DOT | Averages processed alpha within each cell and builds a circular dot whose area follows that average | Yes | Cell size and therefore the constructed dot geometry | Do small, middle, and large dot areas remain distinct in the exported raster and print? |
| FS | Serpentine Floyd–Steinberg error diffusion, alternating row direction | No | The LPI-derived or pixel value feeds minimum-island cleanup, not an angleable screen lattice | Does diffusion retain the intended local tone without losing critical small regions? |
| FM | Deterministic seeded thresholding | No | The same value feeds minimum-island cleanup rather than a regular angled grid | Does the reproducible dispersed pattern survive the declared physical path? |
For Semi-transparent, the current tool processes source alpha above 10 and below 240; fully transparent and substantially opaque pixels pass through. For Full image, non-transparent pixels receive a working alpha derived from inverted RGB luminance before screening: darker colour becomes denser working alpha and brighter colour becomes sparser. That is still an alpha operation, not CMYK colour separation.
Ordered and AM Dot skip the minimum-island cleanup because their own geometry defines the pattern. FS and FM apply it after screening. This is why changing Frequency for FS or FM is not equivalent to rotating or tightening a conventional regular screen.
Prep also has a fixed source-edit order: remove background → colour knockout → fringe repair → invisible-pixel repair → halftone. A later Halftone step cannot recover an edge or fade removed earlier. The current product workflow shows where Prep sits before nesting and export; the prepress-software comparison helps when the unresolved question is which application should own screening at all.
Choose by a controlled matrix, not a settings list
Start from one approved target containing the tonal and edge structures that matter to the actual job. Hold the source pixels, physical size, placement, output row, RIP queue, and physical stack constant. Change one screening variable at a time.
- Freeze size and density. Record pixel dimensions, intended millimetres or inches, effective PPI, and whether the mapping uses embedded or NestSheet-assumed DPI.
- Choose the target intentionally. Use Semi-transparent when the question is how existing partial alpha is distributed. Use Full image only when converting the visible artwork’s luminance into screened opacity is actually the intended effect.
- Compare methods before fine tuning. Export one candidate per method that is genuinely under consideration. Do not compensate for a method change by also changing size, LPI, angle, or RIP processing.
- Bracket cell geometry. Compare at least three deliberately separated Frequency values within 20–200 LPI, or Dot size values within 4–16 px when density is unavailable. These are test candidates, not a low/medium/high production recipe.
- Test angle only where it exists. Hold method and frequency constant while comparing declared angles for Ordered or AM Dot. FS and FM have no angle row.
- Inspect the emitted raster. Use the same production container, channels, bit depth, and export row that the real job will use. Reopen the artifact rather than approving a canvas thumbnail.
- Print the same small matrix. Record the exact RIP/version/queue and device stack, then judge the features defined before printing. Promote a setting only from that bounded result.
A compact test record can look like this:
| Field | Hold constant | Change for this comparison | Observation to record |
|---|---|---|---|
| Source | File hash, pixels, alpha, final dimensions | None | Approved edge and tonal intent |
| Prep | Enabled upstream steps and Apply-to target | Method, then cell/LPI, then angle where applicable | Full-resolution pattern and any lost detail |
| Export/RIP | Container, channels, bit depth, profile responsibility, RIP/version/queue | None | Imported size and raster/plane preview |
| Physical output | Printer mode, ink, film, powder, cure, press, substrate | None | Which declared target features survive and how they differ |
The ICC workflow guide is relevant only to keeping colour interpretation controlled; an ICC choice does not validate alpha-screen geometry.
Use the proxy for direction and the loupe for raster truth
NestSheet’s ordinary Halftone preview runs the same tool on a proxy whose longest side is no more than 1,600 pixels. When density exists, the proxy density is scaled so the screen remains proportional to the artwork. That makes slider feedback useful, but a downscaled preview is still not the full-resolution raster.
When the canvas is below 100% zoom, Prep labels the dots approximate and offers the print loupe. The loupe runs the enabled Prep chain at full resolution, cuts a raster tile, and displays one image pixel as one CSS pixel. Use it on hard edges, fine detail, and midtone transitions before export. It is the correct place to verify the digital pattern—not a simulation of ink or film.
For alpha-derived white, changing artwork alpha can also change the generated underbase. Compare the WHITE view and then use the RIP white-preview guide for the downstream inspection boundary. A file-supplied white plane is a separate production source and must be inspected in the exported artifact and target queue rather than inferred from the artwork Halftone panel.
Decide who owns screening downstream
First-party CADlink documentation lists a “Halftone for Spot Colors” job action. That establishes that screening can also exist in a RIP-side workflow; it does not establish its method, range, or result for your version. Record whether Prep or the exact RIP queue owns the operation. If both are enabled, treat that combination as a distinct test configuration rather than assuming the second stage is neutral.
Adobe’s conventional-print guidance distinguishes image PPI, output-device DPI, and halftone LPI, and tells operators to confirm output settings with the print provider. Adobe also warns that an on-screen separation preview does not reproduce halftone screens or output resolution. Those principles support a proof-first method, not DTF settings copied from plate printing.
What remains physically unproven
Current product code proves the controls, ranges, mapping, algorithms, proxy, and full-resolution loupe. External documentation proves the unit distinctions and that downstream screening may exist. Neither proves a best LPI, angle, minimum surviving dot, powder behavior, moiré outcome, feel, cure result, durability, or wash life for a DTF stack.
A clean loupe and RIP preview can reject a bad digital candidate. They cannot certify droplet behavior, film coating, ink gain, powder, curing, pressing, substrate, or washing. Publish a shop setting only with a controlled receipt that names those variables and the observed result.
Sources
- Adobe, Print images to a commercial printing press: https://helpx.adobe.com/photoshop/using/printing-images-commercial-printing-press.html
- Adobe Illustrator, Print color separations: https://helpx.adobe.com/illustrator/using/printing-color-separations.html
- CADlink Digital Factory, Jobs menu — Output: https://help.cadlink.com/website/digital_factory/en/production/menus/jobs_menu_output.htm
External documentation was checked July 2026. NestSheet method names, control ranges, LPI-to-cell mapping, target behavior, proxy, and loupe were checked against the shipped product code on 21 July 2026. No controlled physical-print receipt was available, so this article deliberately gives a test method rather than a universal DTF recipe.
