Engineering Drawing Ambiguity: When CAD Generation Must Stop

Ambiguous, missing and conflicting are different problems

They have different causes and different fixes, so name the problem precisely before asking anything.

Ambiguous

The drawing states a value, but more than one reading fits it: a leader that could point to either of two features, a TYP note with no clear scope, a projection angle that is not marked. The fix is a choice between named readings.

Missing

The part needs a value the drawing does not state: a hole with a diameter and no depth, a pocket with no depth, a thickness shown in no view. The fix is a value from the design authority.

Conflicting

The drawing states the same thing twice, and the two disagree: a chain of dimensions that does not sum to the overall, a thread with a tap drill that belongs to a different thread, a material in a note that differs from the title block. The fix is to learn which one is wrong, and usually a drawing change.

Two cases look like problems and are not. A reference dimension in parentheses that duplicates a toleranced chain is information, not a conflict, as long as it agrees within the chain's tolerances. A value fixed by a standard the callout invokes, such as the 1.25 mm coarse pitch of an M8 thread under ISO 261, is stated even though no number is written.

Three illustrations

Illustration 1: an R5 TYP with no clear scope

A 120 × 80 × 10 mm plate has four rounded outer corners and a rectangular pocket, 60 × 30 × 4 mm deep, in its top face. One callout reads R5 TYP, with its leader on an outer corner. Two readings fit. In reading A, the four outer corners are R5 and the pocket corners are not specified. In reading B, all eight corners are R5. The difference matters: an R5 internal corner needs a cutter of Ø10 or smaller, and a mating part designed for one reading may not fit a pocket cut to the other.

Illustration 2: a hole with no depth

An 80 × 50 × 20 mm block carries a callout Ø10 with no THRU and no depth. The only side view is a section that does not pass through the hole, so no view shows its bottom. The hole may be through or blind; if blind, its depth is not stated anywhere. The drawing is missing a value, and no reading of it supplies one.

Illustration 3: a chain that does not add up, and a drill for the wrong thread

A bracket's overall length is dimensioned 120. Below it, a chain of three dimensions reads 40, 45 and 40, which sums to 125. One of the four values is wrong, and the drawing does not say which. Even when a chain like this does add up, dimensioning the overall and every link is double dimensioning, and one of them should be a reference dimension.

On the same bracket, a tapped hole is called out M6, and a note gives a Ø6.8 drill for it. The usual tap drill is about the major diameter minus the pitch: 5.0 mm for M6 × 1, and 6.8 mm for M8 × 1.25 (8 − 1.25 = 6.75, rounded to the standard 6.8 mm drill). Either the thread or the drill note is wrong.

Why guessing is unsafe

A guessed value in a CAD model does not look like a guess. It has the same precision as every other value, and nothing downstream marks it. That is the core of the problem.

• It is checked against itself. When the CAD becomes the reference for CAM and for inspection, the guessed value is machined to and inspected against. The part passes inspection and is still wrong.
• It hides a drawing error. A conflict is often the sign of a change made in one place and not another. Settling it quietly leaves the error in the drawing for the next reader and the next order.
• Fit and function turn on small differences. A bore guessed at a clearance fit where a press fit was meant, or a thread guessed coarse where fine was meant, gives a part that cannot be assembled or that loosens in service.
• Some parts carry safety consequences. Edge distance on a bolted joint, wall thickness on a pressure part, thread engagement on a lifting fixture: a guess in any of them can turn a dimensional error into a failure that hurts someone.
• The guesser takes on design responsibility. The design authority, usually the drawing's owner, answers for what the drawing specifies. A supplier that invents or changes a value without written approval has made a design decision it was not asked to make, and may carry the cost when it is wrong.

How to ask a good clarification question

A good question can be answered in one line by someone with the drawing open, without a meeting.

Recording the answer, and who gave it

An answer that lives in one person's inbox is lost the next time the part is ordered. Keep each answer with its question, so that the record stands on its own.

• The question as sent, and the answer as received, quoted rather than paraphrased.
• Who answered: name, role, and the organization they answer for. An answer from someone without design authority is an opinion.
• When, and through what: the date, and the email, request for information (RFI) or deviation number that carries it.
• Which revision it applies to. An answer about revision C does not hold for revision D by default; check it again when the drawing changes.
• Whether a drawing change will follow. If the answer corrects the drawing, ask for the revised drawing, so the next reader does not need the same answer.
• What it changed in the model, so a reviewer can trace the value to the answer rather than to the drawing.

When to stop entirely

Some drawings cannot be made safe by questions. The right result is to stop before building, with the reason stated, and to ask for a better source.

Raster and scanned drawings

A scanned drawing is a picture. Every number on it has to be recognized from pixels and then attributed to the right feature, and both steps fail in ways that look plausible: a 3 read as an 8, a decimal point lost in the noise, a Ø read as a zero. On a poor scan, a misread cannot be told from a real value. Ask for the native vector PDF or an export from the original CAD. The guide to vector and raster drawings explains how to tell them apart.

Multi-sheet drawings read one sheet at a time

Sections, details, notes and flat patterns often sit on later sheets, referenced from sheet 1 by a view letter or a note number. Reading sheet 1 alone gives a model that agrees with sheet 1 and is wrong. Either the whole set is read together, with every cross-reference resolved, or the work stops.

Parts outside the family a reader was built for

A reader or a template built for one kind of part makes assumptions about it: which features exist, which faces locate the others, what a typical value is. Applied to a different kind of part, those assumptions become silent errors. A tool built for flat plates should stop on a turned shaft, not produce its best plate-shaped approximation of one.

Other reasons to stop

• The drawing is not released: it is marked PRELIMINARY or NOT FOR PRODUCTION, or it has no approval.
• The revision cannot be confirmed against the order.
• So many values have to be asked that the drawing is no longer the source. When most answers come from the customer rather than the sheet, ask for a corrected drawing instead.
• The part is safety-critical and the design authority cannot be reached.

For the most common gaps and how to resolve them, see missing dimensions on engineering drawings. For the misreads that produce false conflicts, see common sources of interpretation error.

Questions engineers ask