How a document feeder works

An automatic document feeder has two jobs that sound like one. It has to take exactly one sheet off the top of a stack, and it has to move that sheet past the sensor at a perfectly constant speed. Every irritation you have ever had with a sheetfed scanner is one of those two jobs failing, and understanding which one failed tells you whether to change your paper, your settings, or a part.

Separation is the hard part

Pulling paper is easy; pulling one sheet is not. A stack of receipts is held together by friction, static, and whatever was spilled on it. The feeder’s answer is asymmetry: a driven roller pushes the top sheet forward while a fixed pad or a counter-rotating roller resists everything underneath it. The top sheet wins because it is the only one the driven roller touches.

This works when the sheets are similar in weight and finish. Receipts are not. A thin, slick thermal slip sitting on top of a stiff card invoice behaves differently from the same slip sitting on another thermal slip, and a feeder tuned for office paper is being asked to handle both in the same batch. Two things go wrong. A multi-feed is two sheets travelling as one, which is worse than a jam because it fails silently — you get a file that is missing a page and nothing announces it. A misfeed is nothing being picked up at all, which is loud and harmless.

Better devices watch for multi-feeds with a sensor rather than trusting the mechanism, typically by detecting the difference in what passes through a gap. If your device has such a check, leave it on. If it does not, a page count at the end of each batch is your only defence, and it is worth actually doing.

Transport decides the geometry

Once a sheet is separated, a second set of rollers carries it past the sensor. The sensor reads one thin line at a time and the file is assembled from thousands of those lines in sequence, which is why transport speed matters so much. If the paper speeds up or slows down mid-scan, the image stretches or compresses along its length — the classic symptom is print that looks fine at the top of a receipt and slightly squashed at the bottom.

Skew comes from the same place. If one side of the sheet is gripped harder than the other, the paper rotates as it travels, and the image arrives at an angle. Guides on the input tray exist to stop that starting; they only work if you actually close them onto the stack. A narrow slip in a wide tray with the guides open is a skew waiting to happen, and no amount of software straightening afterwards recovers detail that was smeared on the way through. What the automatic corrections can and cannot fix is covered in deskew, autocrop and auto-rotate.

Everything in the path is a consumable

The parts that touch paper are designed to wear out. Rubber rollers work because they are grippy, and they are grippy because they are soft. Paper dust, toner, thermal coating, and the oil off your fingers all polish that surface over time, and a polished roller slips instead of gripping. A separation pad wears the same way, and when it does the device starts multi-feeding rather than refusing to feed.

This is the single most common reason a scanner that “used to be fine” starts misbehaving. It is not the software, it is a rubber part that has done its life. Most sheetfed devices are built so those parts come out without tools, and treating them as consumables rather than faults changes your relationship with the machine considerably — the practical routine is in scanner cleaning and consumables.

Paper dust also lands on the glass or plastic strip in front of the sensor, which produces a completely different symptom: a line down the length of every page in the batch, always in the same place. Learning to read a defect’s shape saves a lot of guessing, and diagnosing scan defects is mostly that skill.

What receipts do to a feeder that documents do not

Several properties of till receipts are individually mild and collectively hostile.

• They are short. A feeder’s rollers are spaced for a full page. A short slip can be released by one set of rollers before the next set has taken hold, at which point it is travelling on momentum and the geometry stops being controlled.
• They are narrow and inconsistent. Every till uses a different roll width, so a single batch presents the guides with several widths at once.
• They curl. Paper that lived on a roll remembers the roll. A curled leading edge does not present flat to the pick roller and is the usual cause of a receipt that will not feed on the first try.
• They are coated. The thermal face is slicker than plain paper, so the driven roller has less to grip.
• They vary in length wildly. A grocery shop and a coffee produce items with an order of magnitude between them.

None of this means receipts cannot be fed; it means a receipt batch is a harder test of separation than a document batch, and it will find a worn pad long before an ordinary stack of letters does. The failure modes and what to do about each are in why receipts jam.

What to look at before you trust a feeder

If you are evaluating a device, the feed path is where to concentrate, because it is the part you cannot fix with settings. Look at whether the guides adjust down to a narrow slip rather than stopping at envelope width. Look at whether the tray holds short items flat or lets them sag. Look at whether the wear parts are described in the manual as replaceable, and whether there is any way to see how many sheets the device has fed. And feed it your own worst paper before you commit, which is the whole point of testing a scanner with your own paper.

None of this touches the question of whether a scan is an acceptable substitute for the slip that went through the rollers. That depends on rules that differ by country and by entity type, and it is a question for your own tax authority or accountant — not for the machine’s manual.