You can draw. That isn't the problem. The problem arrives at the plotter, when sheet 03's walls come out fatter than sheet 02's, the survey layers you thought you'd killed reappear in grey, and the section that looked perfect on screen prints like a wiring diagram. Nobody drew it wrong. The set was assembled wrong.
Here's the thing nobody tells you at the start: a printed line weight is decided in about five separate places, and any one of them can quietly override the other four. The object's own property. The layer it sits on. The plot style table attached to the layout. The page setup that names that table. And the viewport, which can override the layer for that sheet only, invisibly, forever. Get those five agreeing with each other and a set prints identically every time. Leave them to negotiate among themselves and you get what you're currently getting.
A layer standard isn't an organisational scheme. It's a print specification wearing a filing cabinet's clothes. The only question that matters when you're writing the list is: at the smallest scale I'll issue this at, what needs to be the darkest thing on the page, and what needs to nearly disappear?
At 1:50 you can afford nuance — three weights of wall, a distinct weight for joinery, hatching that reads as texture rather than sludge. At 1:100 that same drawing collapses. The 1:50 subtleties merge, the hatch turns into a grey wash, and the furniture starts competing with the structure. This is why sets look inconsistent even when the layer names are perfect: the layer convention was designed at one scale and issued at another.
What we teach is to fix the hierarchy first and name the layers second. Decide on a small number of print weights — the ISO pen sizes give you a ready-made ladder, 0.13, 0.25, 0.35, 0.50, 0.70 — and give each one a job. Cut structure heaviest. Cut non-structure a step down. Everything seen beyond the cut, one step lighter again. Annotation, dimensions, hatch and grid at the bottom. Four working weights is plenty. Five is generous. If you find yourself needing eleven, you are drawing detail into a plan that wants to be a detail drawing.
Only then do you write layer names, and you write one per weight-and-purpose combination, not one per noun in the building. Two door layers with the same pen weight and the same plot behaviour are one layer with a bad filing habit.
If you want the naming to survive contact with a UK project team, BS 1192:2007 set out a field-based layer name — a role code for the originating discipline, a classification for the element, a presentation field for the graphic treatment. ISO 19650, which took over the information-management side of things, doesn't hand you a layer list at all; it's about how information is named, exchanged and approved, not about pen weights. So don't wait for a standard to tell you what to plot at 0.35. It won't.
CTB or STB: Pick One, Office-Wide, Forever
This is the single biggest cause of a set that prints inconsistently, and almost nobody explains it properly.
A CTB is a colour-dependent plot style table. Every one of AutoCAD's 255 index colours maps to a printed pen — a weight, a screen percentage, a print colour. Red might print at 0.50mm black, cyan at 0.13mm. Your layer colours are your pen assignments. It's crude and it works, which is why it's the de facto standard across most of the industry and why almost every consultant file you'll ever receive arrives set up this way.
An STB is a named plot style table. Instead of colour driving the pen, you assign a named style — "Heavy", "Light", "Screened 50%" — to a layer or an object directly. It's the more logical system by miles. You can have three layers all coloured cyan on screen and printing at three different weights, which means screen colour becomes a working aid rather than a plotting instruction.
So why doesn't everyone use STB? Because a drawing file is either colour-dependent or named. Not both. Never both. When you copy geometry from a colour-dependent drawing into a named one, the plot styles don't come with it, and what you get on the other side is whatever the destination decides. Multiply that across a set where three of the sheets were started from an old template, and you have drawings that are genuinely incapable of printing the same.
The practical call: use CTB unless your whole office, including the people who send you files, uses STB. You'll fight the industry otherwise. If you inherit a mixed situation, AutoCAD's conversion commands will take a drawing from one system to the other and a CTB into an STB, but convert the whole set in one sitting and check every layout afterwards — the conversion changes the file, not your assumptions about it.
One rule underneath all of this: everything is ByLayer. Colour, line weight, linetype, plot style. The moment an object carries its own explicit colour, the layer stops controlling it and your plot style table is talking to something that isn't listening. This is the actual reason for most mystery-weight lines in a set, and the fix is SETBYLAYER on a full selection, or QSELECT to hunt down the specific offenders first.
Classes Aren't Layers, and Vectorworks Users Know Something You Don't
Worth understanding even if you never open Vectorworks, because it explains what AutoCAD is missing.
Vectorworks splits the job in two. Design layers answer where — ground floor, first floor, roof, site. Classes answer what — the graphic attributes: line weight, colour, fill, whether it prints at all. A wall on the Ground Floor layer in the Walls-Exterior class. Two independent axes, and you can switch either without touching the other.
AutoCAD has one axis. The layer has to carry both the where and the what, which is precisely why layer names get long and why every office argues about whether the level code goes at the front or the back. There's no right answer, only a consistent one — and when a team can't agree, the tiebreaker is whichever order lets you sort the layer list alphabetically and get useful groupings. That's it. That's the whole argument, and it shouldn't take a meeting.
Revit, for the comparison: categories and subcategories do the graphics, object styles set the defaults, view templates enforce them per view. Worksets are about who's editing what, not about line weight — if someone tells you to control graphics with worksets, they're solving the wrong problem.
The Consultant's File Is Not Your Drawing
A survey or a structural DWG lands in your inbox with 340 layers, half of them frozen, a text style that doesn't exist on your machine, and a proxy object that throws a dialogue box every time you open it. Don't clean it. Clean a copy of it, and reference the copy.
The sequence that works:
Attach it as an Xref, don't insert it. Xrefs keep their geometry at arm's length, arrive on prefixed layers that can't collide with yours, and update when they reissue. Inserting a consultant drawing merges their layer chaos permanently into your file. This is why the External Files lesson in our AutoCAD course spends so long on attachment options rather than treating Xrefs as an afterthought — it's the decision that determines whether the next six months are manageable.
Set VISRETAIN to 1. Without it, every colour, line weight and on/off state you set on the Xref's layers is thrown away the next time that Xref reloads. You'll have carefully screened the survey back on twelve sheets, the surveyor sends a revision, and all twelve go black again. Everyone learns this the hard way once.
Then AUDIT, then PURGE — twice. Purge nested items and zero-length geometry, run it again, and only then look at what's left. Registered applications and unreferenced styles hide behind the -PURGE command line version.
Use the Layer Translator for the layers themselves. LAYTRANS is the tool nobody uses and everybody needs: it maps their layer names onto yours, saves the mapping, and applies it. Do it once per consultant and you've built a reusable translation for every file they'll ever send you. This is the actual answer to reconciling a foreign layer standard, and it beats manually reassigning geometry at eleven at night.
And the thing people get wrong: don't delete layers to tidy up. Freeze or turn off. A deleted layer takes its geometry, and the boundary you binned at RIBA 3 is the one the contractor asks about at Stage 5.
The Same Layer, Printing Differently, On Purpose
Sometimes you genuinely need the demolition layer to print solid on the demolition sheet and screened back on the proposed. The wrong fix is a duplicate layer. The right fix is a viewport override.
Double-click into a layout viewport, open the Layer Properties Manager, and you get extra columns: VP Freeze, VP Colour, VP Linetype, VP Lineweight, VP Plot Style. Change them and the effect is local to that viewport only. The master standard stays intact, the layer name stays the same, and one sheet behaves differently.
Two warnings, because this is powerful and it bites. Overridden layers are highlighted in the Layer Properties Manager and invisible everywhere else — someone opening your file in model space has no way of knowing. And the debugging pain is real: a line printing wrong on one sheet out of thirty, with no fault in the layer standard, is almost always a viewport override somebody forgot. As a rule of thumb, VP Freeze freely, VP Lineweight rarely, and if you override a plot style at viewport level, note it somewhere a human will read.
One Page Setup, Forty Sheets
Here's where sets actually fall apart. Every layout carries its own page setup — plotter, paper size, plot area, scale, and critically, which plot style table it uses. Forty layouts built over six months by three people means forty chances to pick the wrong CTB.
Build one named page setup in your template. Not "Previous plot", not whatever was set last time — a named setup with the device, the sheet size and the plot style table locked in. When a layout needs a different paper size, that's a second named setup, not a manual edit.
Use the Page Setup Manager's import to pull that named setup into any existing layout, including ones you inherited. And at issue, go through Sheet Set Manager or the Publish dialogue and apply a single page setup override to the entire selection — one setting, every sheet, one pass. Publishing the whole set as a single multi-sheet PDF forces everything through the same pipeline, which means if something is wrong it's wrong consistently and you'll spot it. Sheet-by-sheet PDFs are how a rogue plot style survives to the client.
The Pre-Issue Checklist
Run this before every issue. It takes about ten minutes on a mature set and it's caught more than it hasn't.
Select all in model space, run SETBYLAYER. Anything that changes appearance was carrying a hard-coded property.
Open the Layer Properties Manager and sort by name. Layers that don't match the standard stand out instantly in an alphabetical list.
Check every layer's Plot/No Plot flag. Guides, setting-out, working notes — off.
Audit and purge the file. Then check nothing you needed vanished.
Reload all Xrefs and confirm VISRETAIN held your overrides.
Cycle every layout and confirm the page setup name is the one you expect, not "Previous plot".
Scan the layer manager in each viewport for override highlighting. Justify each one or clear it.
Publish the whole set to one PDF, then open it and flick through at 100% on screen. Your eye catches a weight inconsistency across consecutive pages far faster than any command will.
If you're still shaky on the underlying mechanics — object properties, annotation scale, what a layout actually is — start with our AutoCAD basics guide and come back to this.
A drawing set isn't thirty drawings. It's one document that happens to be printed on thirty pieces of paper, and the reader judges it as one document. They won't consciously notice that your wall weights are identical from sheet to sheet. They'll just find it easy to read, and assume you know what you're doing.
