Diagnosing scan defects by their shape

When scans start coming out wrong, the useful question is not “what is broken” but “what shape is the defect”. A mark that appears in the same place on every page is telling you something quite different from one that moves, and a defect that runs along the direction of paper travel points at different hardware than one that runs across it. The shape narrows it to one or two causes almost every time.

The geometry that makes this work

A sheetfed scanner reads one thin line at a time while the paper moves. So there are two axes and they mean different things. Anything that stays put in the sensor’s field of view — a speck on the glass, a dead element in the sensor array — repeats on every line and therefore appears as a stripe running the length of the page, in the same lateral position every time. Anything that happens for a moment in time — a stutter in the transport, a flicker in the lamp — affects a few consecutive lines and appears as a band running across the page.

That single distinction resolves most diagnoses. A flatbed inverts the relationship because the sensor moves and the paper does not, but the principle holds: persistent-in-space means contamination or a sensor fault, persistent-in-time means transport or illumination.

Stripes down the page

A dark or light line running top to bottom, in the same place on every sheet, is almost always dirt or a scratch in front of the sensor. Paper dust, adhesive residue, or a smear of something from a greasy slip sits on the glass or plastic strip, and every line the sensor reads passes it.

Clean the strip properly on both sides of the paper path — most devices have two, one for each face — and rescan the same sheet. If the line goes, that was it. If the line stays after cleaning, the likely cause is a scratch, and a scratch is permanent. Staples are the usual culprit, which is why they matter more than the jam they cause: see why receipts jam.

A line that appears at the exact same position but only on one side of a double-sided scan localises it to that one sensor’s strip. A line that follows the paper rather than the page position — appearing on the left of one sheet and the middle of the next — is not this at all; it is something on the paper itself, or a roller depositing residue.

Bands across the page

A dark band, a pale band, or a stretch of visibly squashed print running left to right means something changed while that part of the sheet was passing the sensor. Two candidates.

Transport speed varied. The paper slipped, or a worn roller lost grip momentarily, or something was hanging off the trailing edge and pulled. The symptom is print that is compressed or stretched vertically in that band, and it is the classic result of feeding a long item with no support behind the machine — covered in scanning very long receipts.

Illumination varied. Less common, and it presents as a brightness change without geometric distortion.

If bands appear on the same sheets repeatably, it is the paper’s stiffness or condition. If they appear at random, it is wear in the drive path, which is the maintenance case in scanner cleaning and consumables.

Haze, ripple and softness around the print

If flat areas of paper are clean but every character has a faint halo or ripple around it, and the smallest print looks mushy while headers look fine, this is not a hardware fault at all. It is lossy compression doing what lossy compression does to fine detail on plain backgrounds, and the fix is a settings change: compression and scan quality covers which one.

The distinguishing test is that a genuine hardware defect has a position and a compression artefact does not. Compression damage is distributed wherever detail is, evenly, across the whole page.

Missing print rather than damaged print

If characters are broken, hollow, or absent — but what remains is crisp — nothing is dirty and nothing is worn. Something threw away tonal information, and there are three candidates: a hard black-and-white threshold, automatic background removal, or a slip that has genuinely faded past capture.

Distinguish them by rescanning the same slip in greyscale with all cleanup disabled. If the print reappears, a setting was eating it, and the reasoning behind the settings is in colour, greyscale, or bitonal and deskew, autocrop and auto-rotate. If it does not reappear, the paper has lost it, and the rescue attempts worth making are in scanning faded thermal paper.

Shadows, dark edges and missing strips

A dark region in one corner or along one edge usually means the paper was not flat against the reference plane at that point — a curled corner lifting away from the glass, or a fold standing proud. On a flatbed, closing the lid properly fixes it. On an overhead capture it is the normal condition and the reason a fixed rig beats a hand-held phone.

A strip of the document simply not present, with a clean edge where it stops, is a cropping decision rather than a defect. That is the silent failure discussed in deskew, autocrop and auto-rotate, and it is worth checking for deliberately because it does not look like damage.

The diagnostic sheet

Keep one reference item and rescan it whenever something looks off. Any consistent piece of paper works — a printed page with fine text and a large flat area, ideally with a ruled border. The value is that you know exactly what it should look like, so any defect in the scan is the machine’s and not the paper’s.

Run it after cleaning to confirm the clean worked, after any settings change to see what the change did, and first whenever you suspect a fault. Most of the time it will tell you within a minute whether you are chasing a hardware problem or a settings problem, which is the distinction that wastes the most time when you get it wrong.

None of this speaks to whether a defective scan still functions as a record, or whether a receipt with a stripe through the total needs recapturing from the paper before the paper goes. Those rules vary by country and by entity type — ask your own tax authority or an accountant where you are, and keep the original until you have.