Halftones for screen printing: lpi, angles, dot shape, film output

First: if you want a halftone look, this is not that

Two completely different things share the word "halftone". If you are after a retro comic-book or newsprint dot texture for a poster, a record sleeve, a web hero or an illustration — Ben-Day dots, big visible circles, a risograph vibe — you want a halftone *effect* filter, and there are a dozen free ones a search for "halftone effect generator" will hand you. This is not one of them, and forcing it to be one would be miserable: the output here is a separation plate, one per ink, at film resolution, screened at an angle chosen so it will not interfere with the plate next to it. It is meant to be printed onto transparency film and burned onto a screen. If nobody is going to burn a screen with it, close this tab and go get the filter — that is a better use of your afternoon than fighting a printer's tool for a design effect.

The four numbers that decide a halftone

Frequency, in lines per inch, is how many rows of dots per inch — garment work typically runs 35 to 65 lpi. Angle is the rotation of the dot grid, and it matters because two grids at similar angles beat against each other and produce moiré. Dot shape — round, elliptical or square — changes how tones join up as they grow; elliptical dots are popular because they avoid the abrupt tonal jump round dots make at 50%, where round dots all touch their neighbours at once. Output DPI is the resolution of the film itself, and 600 or 1200 dpi is what determines how many distinct dot sizes the device can actually resolve at your chosen frequency. Set all four explicitly. Every one of them is a press decision, and a default that suits somebody else's shop is not a decision.

Mesh count sets your ceiling, not your taste

A screen cannot hold a dot finer than its own openings, so mesh is the hard constraint on frequency and the number to start from. The rule of thumb in general circulation puts mesh count somewhere around four to five-and-a-half times your lpi — so 45 lpi sits comfortably on 205 mesh, and 55 lpi on 305 is a real, commonly-run pairing at the upper end of that band — and going finer than your mesh supports does not produce a finer print, it produces dots that break up, fill in, or wash out of the emulsion. Finer mesh is also more fragile and more of your week spent restretching. And higher is not automatically better even when the mesh can take it: a bold one-colour design that reads clean and solid at 35 lpi can come off faded at 55 because the dots holding the light end have gone too delicate to print reliably. Pick the coarsest frequency that holds the detail the artwork actually has.

Angles chosen so they do not fight

For four-color process there is a well-established set of mutually non-interfering angles, with yellow placed where the eye notices its pattern least. For simulated process with six, eight or ten inks there is no single canonical set, so the angle assignment is generated for your actual ink count and ordering to keep the plates as far apart as possible. You can override any of it and enter your own angles — some shops have a set they trust on their equipment, and a tool arguing with a printer about their own press is not a useful tool.

Dot gain, the reason process prints go muddy

Ink spreads. A 50% dot on film becomes something well past 50% on fabric, and the effect is worst in the midtones where a print has the most to lose. Dot gain compensation applies an inverse curve before screening, so the dot laid on the shirt lands where you meant it to. The strength is a slider rather than a fixed profile, because gain depends on your mesh, your emulsion thickness, your ink, your squeegee and your substrate. Print a step wedge once on your own setup, read where it actually lands, and set the value — after that, it is the same number for every job on that press.

What your RIP expects, and when to screen it yourself

A RIP wants to do its own screening. Feed it continuous-tone grayscale plates — one greyscale image per ink, in register, at full print size — and it applies its own frequency, angle set and dot shape with curves tuned for the output device it drives. Screen the plates first and hand it 1-bit film and you get two dot grids fighting: the RIP re-screens your dots and you moiré against your own artwork. So grayscale is the default output here, and it is the right answer if you print through AccuRIP, Wasatch, FilmMaker or anything similar. Pre-halftoned 1-bit film is for the other case — printing to an inkjet or laser through a plain driver with no RIP in the chain — and then the screening has to happen upstream, which is here, with your frequency, shape, angle set and output DPI set explicitly. Either way the plates come out in register at full canvas size, so nothing shifts between films. One practical reason the screening is the paid half: a 14-inch plate at 1200 dpi is a 16,800-pixel image, which is comfortable on a server and genuinely painful in a browser tab on a shop PC.

It comes with the separation, not instead of it

Halftoning is applied per plate as part of the print-ready render, so the same 10 credits that separate the image also screen it: choked underbase, highlight white, choke and spread, trapping, dot gain, halftone, registration marks. Re-renders of that image are free for 24 hours, so trying 45 lpi against 55 lpi, or round against elliptical, costs nothing but the time to look — and a day is long enough to burn one and print it before you decide. If all you need is flat spot plates with no tone, you never need halftoning at all — and that path is free and runs in your browser.

Or skip halftones entirely

Index separation is the escape hatch. Every dot is the same size at 100% ink, tone comes from dot density, and because there are no varying dot sizes there is no angle set to choose and no rotated grid for the next plate to interfere with — moiré stops being the thing you manage. It is coarser than a good halftone and more graphic in character, but it is far more forgiving on press, and on textured garments and older equipment it frequently wins outright. Index separation is free and runs in your browser.