An old drawing can contain valuable design information without providing editable geometry. A scan may show dimensions and outlines clearly enough to guide a reconstruction, while a PDF may contain a mixture of vector lines, text and images. The challenge is to recover useful information without giving uncertain source material an appearance of precision it has not earned.
George Omura’s Mastering AutoCAD 2011 and AutoCAD LT 2011 explains using images and PDF underlays as backgrounds for drawing work. The enduring method is to establish the source, calibrate its scale, reconstruct deliberately and check the result independently. For a small business using computer-aided design, or CAD, this is as much an information-quality task as a tracing task.
Identify what the source actually contains
A raster image consists of pixels arranged into a picture. A scanned line is therefore a band of pixels, not a mathematical line with an exact endpoint. Enlarging the image makes those pixels easier to inspect but does not recover missing geometry.
A vector drawing represents shapes through geometric instructions. A PDF may preserve vector information from a drawing application, contain only a scanned page, or combine both. The filename extension alone does not tell you whether selectable geometric information is available.
Inspect a representative area before choosing a workflow. Look at fine linework, dimensions, curves and overlapping details. Establish whether the document contains multiple drawing scales on one sheet, because a single calibration cannot automatically apply to every detail view.
Record the source filename, page, revision and date where available. Also record what is unknown. A legible PDF without a clear issue status may be adequate for an early discussion while remaining unsuitable as the sole basis for a production decision.
Choose reconstruction, attachment or import deliberately
An underlay is an attached reference displayed within the working drawing. It lets you see the source alongside new geometry without treating the source content as ordinary editable drawing objects. This is useful when only a selected part needs to be reconstructed.
Reconstruction means building new geometry from the information in the source. Written dimensions, alignment relationships and identifiable centres can guide that work. Following every visible pixel is not necessary when the design relationship is known more clearly than the scanned outline.
Later AutoCAD releases also support importing eligible PDF content as drawing objects. This differs from the attachment workflow described in the 2011 book. Autodesk’s PDF import guidance distinguishes underlays from imported objects and notes that importing a raster PDF does not perform raster-to-vector conversion.
Select the method according to the next task. Keeping a drawing as a background may be sufficient for comparison. Editable reconstruction is more appropriate when dimensions will change or the geometry will feed another process. Automated conversion can provide a starting point, but its output still needs inspection and organisation.
Preserve the source and define the scope
Keep the supplied document separate from the reconstructed drawing. Work from a copy if image preparation is needed, and record which copy was used. Cropping, rotation and cleaning can improve readability, but they should not obscure the relationship to the original information.
Specify the scope before tracing. A task might require only the external outline and mounting centres, rather than every note, hatch and decorative feature. A narrow scope helps focus verification on the geometry that will actually influence a decision.
Set an explicit intended use for the result. A concept comparison, a layout study and a manufacturing input require different evidence. Converting a source into CAD does not by itself make that source suitable for a more demanding purpose.
For received material, preserve its identifying notes and any stated limitations in the project record. A reconstructed file should remain traceable even when the background is hidden. Otherwise, a later user may reasonably but incorrectly assume that every dimension came from a current measured source.
Calibrate scale using a trustworthy distance
Calibration establishes the relationship between a measured distance in the reference and a known real-world distance. The book demonstrates reference scaling: identify two points in the image, then specify the distance they should represent in the drawing’s chosen units.
Choose a long, clearly identified dimension whose endpoints can be located reliably. A longer reference generally makes a small picking error less influential in the resulting scale. Avoid using a page border or an assumed paper size unless that is genuinely the information you intend to calibrate.
The scale factor is the intended length divided by the current measured length. Apply it to the attached reference with an understood base point. Then remeasure the calibrated distance to check the operation, rather than assuming the entry was interpreted as intended.
Do not stop at the calibration distance. It was used to establish the result, so agreement there is expected. A second dimension elsewhere in the drawing provides a more useful independent check, particularly if it runs in a different direction.
Look for distortion that one scale factor cannot fix
Uniform scaling changes every length by the same ratio. It cannot correct a source that has stretched differently across its width and height. Paper distortion, scanning arrangements or earlier image editing can produce discrepancies that remain after one dimension has been matched.
A photographed drawing introduces another possible problem: perspective. Parallel lines on the original sheet can appear to converge in the image. Treating that photograph as an undistorted plan can produce locally plausible but globally inconsistent geometry.
Check dimensions in different areas and directions. Compare a long horizontal span, a vertical span and a smaller local feature where reliable dimensions exist. Record the differences instead of repeatedly rescaling until one preferred area looks right.
If those checks disagree materially for the intended use, obtain a better source or reconstruct from confirmed dimensions. Do not silently stretch the image in different directions and call the result accurate. Any correction method needs its own stated assumptions and verification.
Rebuild design relationships instead of copying noise
An intended straight edge should normally be reconstructed as a straight line, even when the scan contains a slight waviness. Similarly, a dimensioned circular hole should be built as a circle using confirmed information, rather than a many-sided outline following the pixels.
Use clear reference points and relationships. Establish an origin, main axes and significant centres, then locate secondary features from them. This makes the reconstructed geometry easier to inspect and revise than a collection of independently traced fragments.
When written dimensions disagree with the visible outline, investigate the discrepancy. The annotation may be stale, the image may be distorted, or the drafter may have used a schematic representation. Neither source should be silently preferred when the difference matters to the intended use.
Separate confirmed geometry from approximate interpretation. A working layer or clear note can identify an outline reconstructed from appearance alone. The important point is that the distinction remains understandable after the source image has been switched off.
Make the background readable without confusing it with geometry
The book describes fading images, adjusting contrast, controlling drawing order and clipping a reference to a useful area. These adjustments help distinguish the background from newly created objects. They improve the working view without changing what is known about the original design.
Clipping limits the displayed portion of a reference. It can remove an irrelevant title block or isolate a detail, but it does not necessarily remove the underlying source file from the project’s dependencies. Keep enough identifying information elsewhere to explain which page and area were used.
Check display and plotting behaviour separately. A reference frame may be useful for selection while being unwanted in the output. A faded background that works on screen may become too faint or too strong in a PDF, so inspect the actual output where the reference is intended to remain visible.
Periodically hide the background and review the reconstruction alone. Gaps, duplicate segments and missing features are easier to see when the original image no longer fills them visually. Then reveal the source again to check alignment and completeness.
Treat snapping and conversion as aids to inspection
Object snaps locate recognisable geometric features such as endpoints or centres. With eligible vector PDF content, underlay snapping can be useful. However, the source book itself notes that PDF locations may be approximate representations of the original CAD data.
A successful snap therefore establishes a relationship to the PDF representation, not necessarily to the original design model. Check the resulting geometry against trustworthy dimensions. This is especially important when the PDF was generated with limited precision or later modified.
Imported content can also have a different structure from the drawing that produced it. A dimension may become several objects, or a curve may arrive in a form that is inconvenient for later editing. Inspect object types, continuity and scale before accepting the result as a usable model.
Avoid bulk cleanup based only on appearance. Joining nearby fragments can bridge an intentional gap, while deleting apparent duplicates can remove information serving a different purpose. Make structural changes in small, reviewable groups and retain a recoverable version.
Work through a calibrated reconstruction
This is an illustration. A small Australian business has a scanned drawing of a rectangular display panel. It needs a concept layout showing the panel outline and two mounting centres. The source labels the overall width as 1,200 mm and the height as 800 mm.
After attachment, the distance between the width endpoints measures 240 drawing units. The intended drawing unit is one millimetre, so the initial scale factor is 1,200 divided by 240, which equals 5. The drafter applies that factor and confirms the width now measures 1,200 mm.
An independent check of the visible height measures 795 mm. The discrepancy from the labelled 800 mm is 5 mm, or 0.625% of 800 mm. The team records that result as evidence of uncertainty in the image or the chosen reference points; it does not declare a universal acceptable tolerance.
The mounting centres are dimensioned 100 mm from the left and right edges and 200 mm above the base. With the lower-left corner as the origin, the intended centres are therefore at coordinates (100, 200) and (1,100, 200) mm. Their horizontal separation is 1,000 mm.
| Item | Reconstruction basis | Independent check |
|---|---|---|
| Overall width | Labelled 1,200 mm dimension | Compare calibrated source and model |
| Overall height | Labelled 800 mm dimension | Record the 795 mm image measurement discrepancy |
| Left mounting centre | 100 mm from left, 200 mm above base | Measure both offsets in the model |
| Right mounting centre | 100 mm from right, 200 mm above base | Confirm 1,000 mm centre separation |
The team draws the dimensioned rectangle and locates the centres using these relationships. It does not distort the reconstructed rectangle to follow the shorter scanned height. Instead, it notes that the concept uses the written dimensions and that the original geometry has not been independently verified.
Before the drawing is used for anything beyond the concept discussion, the unresolved discrepancy needs an appropriate source check. That might mean obtaining the original editable file, consulting a later drawing or measuring the actual panel. The conversion task has exposed that need rather than concealed it.
Check the result at several levels
First check geometry: overall extents, key offsets, radii, closure and intended alignment. Use numerical measurements as well as visual overlays. A line can coincide closely with a thick scan while still being displaced enough to matter for a particular downstream task.
Next check information: labels, identifiers, units and source notes. Transcribing a dimension correctly is different from confirming that it belongs to the selected feature. Pay particular attention where crowded annotations cross several nearby edges.
Then check structure: layers, object types, duplicated content and reference dependencies. A reconstruction that looks correct but consists of disorganised fragments can be difficult to maintain. Put objects into an arrangement that reflects their intended role.
Finally check the output or transfer. If another application will use the geometry, verify the representation there. Checking CAD data when changing file formats explains why successful export alone is insufficient evidence that the recipient has the intended information.
Keep the handover honest and useful
Include a concise reconstruction note with the working file. Identify the source, the part rebuilt, the unit convention and the calibration references. List unresolved discrepancies and distinguish dimension-based construction from approximate tracing.
Retain the necessary image or PDF dependencies when the recipient needs to inspect the source alongside the model. Test that arrangement outside the original working folder. Managing CAD references and underlays as shared inputs covers the wider reference-management process.
For a small team, a simple register can be enough: source document, reconstructed file, intended use, checks completed and remaining questions. The register should make it possible to continue the work without repeating the entire investigation or guessing why an edge was placed where it was.
Questions to ask
- Is the source a raster image, vector PDF or mixture of both?
- Which dimensions establish scale, and which independently check it?
- Do different directions or areas reveal distortion?
- Which features are confirmed and which are approximate interpretations?
- Can the next user identify the source, intended use and unresolved discrepancies?
Bringing it together
Rebuilding geometry from a scan or PDF is a controlled interpretation of existing information. Calibrate deliberately, use confirmed relationships and inspect the reconstruction independently. Preserve uncertainty where the source remains uncertain, so a cleaner drawing also becomes a more trustworthy working record.
Source: George Omura, Mastering AutoCAD 2011 and AutoCAD LT 2011 (2010), Chapter 14; Autodesk documentation linked above. Dimensions are hypothetical illustrations. Verify reconstructed geometry against suitable project information before relying on it for manufacture or construction.
