Before the pipeline, before the LLM skills, before a single court file, the first thing this archive digitized at scale was a stack of my kids’ drawings. The machine that did it was chosen on exactly three requirements, and one of them quietly disqualified half the market.
This is part of the Paper Trail series: turning an entire family’s paperwork, from today’s invoices to court files and century-old records, into a system that reads itself. The map of the whole project is the first entry.
Last time I covered the software bootstrap: paperless-ngx, two ingest doors, coding agents as the clerks. This post is about the hardware door, because “scan everything” is a plan, and the machine you pick decides whether the plan survives contact with an actual drawer of paper.
Four ways to digitize paper
There are four realistic options, and the right question for all of them is the same: volume × fragility × fidelity. How much paper, how delicate, and how faithful does the copy need to be?
A phone. The camera plus a scanning app is genuinely fine for what it’s for: one receipt, one contract page, something you need captured right now. Past a handful of pages it falls apart: you become the feeder, and lighting, skew, and perspective correction become your job. Volume is the killer.
A camera. A proper camera on a copy stand is the highest-fidelity option on this list; it’s how archives digitize fragile and bound material that must not be fed through anything. It is also slow, manual, and a project per page. The right tool for the rare, precious item; the wrong tool for a drawer.
A flatbed. Best per-page quality for anything delicate, bound, odd-sized, or where color actually matters: photos, artwork, old documents that crack when bent. The cost is your time: one page, lift the lid, place, lower, repeat. A flatbed assumes a human is attached to it.
A document scanner. A feeder you load, a duplex pass that reads both sides at once, and a stack of paper that becomes a stack of PDFs while you do something else. Built for exactly one thing (documents in volume) and, as we’ll get to, noticeably worse the further you stray from that one thing.
For an archive measured in drawers, the document scanner is the spine and everything else is a specialist. So the real decision was which one.
Three requirements
One: a network interface. A condition, not a preference. The scanner from the bootstrap post scans straight to a shared folder that paperless consumes. That only works if the scanner is a standalone network citizen: paper in the feeder, button, document in the archive. No computer in the loop. A USB-only scanner means a powered-on machine, a logged-in user, and vendor software running: three failure points in front of a task whose whole point is being frictionless.
Two: Linux. This is a Linux-first household, and that disqualifies a surprising share of the market: not always because the hardware can’t, but because the vendor’s software won’t. Scan-to-network-folder softens this a lot (the scanner talks to the share directly, no drivers involved in the daily flow), but setup, management, and any computer-attached scanning still need to work without a Windows machine in the closet.
Three: duplex and a real ADF. Bills are double-sided. Court files are double-sided. Feeding pages twice, or one at a time, is how a scanning habit dies.
Those three filters left a short list, and the Brother ADS-4300N won it: a wired network port, scan profiles that write directly to an SMB share, an 80-sheet feeder, single-pass duplex rated at 40 pages per minute (80 images, since it reads both sides at once), and Linux support that Brother actually maintains.
The honest part
The review, both halves of it.
For documents, it is excellent and fast out of the box. A stack of invoices goes through the feeder at a rate that makes you stop batching them up; both sides in one pass; straight into the consume folder and on into the archive. Zero babysitting. This is the machine doing precisely what it was built for, and it shows.
Away from documents, it shows too. Photographs and the kids’ drawings come out with unnatural colors: oversaturated in places, flat in others. It can probably be tuned; it is definitely not right out of the box. No real surprise: a document scanner is calibrated to make text crisp, not to reproduce color faithfully. But it’s worth saying plainly, because “scan everything” includes things that aren’t text.
And it is sensitive to bad paper. Older, thinner, more brittle pages: the feeder doesn’t love them. That turns out to be a section of its own, because it is the part that decides whether the interesting half of the archive can be scanned at all.
Not every page goes through the feeder
An ADF is a friction machine: rollers grab the top sheet and bend it around a curve on the way out. A feeder does not scan paper. It scans paper that behaves — uniform weight, intact edges, no tears, no staples. The paper this archive cares about does not behave, and it fails in three ways worth ranking: a jam (annoying), a double-feed that silently swallows a page (dangerous, because you notice when you need that page), and a tear (unrecoverable, because the original is the archive).
The fix for most of it is a carrier sleeve (ordered separately): a transparent pouch the fragile page rides through inside, so the rollers grip plastic instead of a sixty-year-old fold. It costs you the batch, though: one page at a time, fed by hand, with the multi-feed alarm turned off, because a sleeve plus a page looks exactly like the double-feed the scanner is built to catch.
And some paper still doesn’t qualify at all: bound, oversized, or already in fragments. That’s not a sleeve problem, that’s a flatbed. Which matters for this series, because the fifty-year court files and the century-old family papers in later posts are exactly the paper that fails in a feeder.
The flatbed that’s coming
The color experience settles the argument the first section opened: one device does not do everything, and pretending otherwise just produces ugly scans of things you care about. So a flatbed is almost certainly joining the setup. The open question is which kind: one with a transparency unit for negatives and photographs, which would open up the family photo albums and a box of negatives (a whole wing of the archive), or a cheaper, generic flatbed that handles fragile paper and the occasional drawing and nothing more.
The box of 35mm film negatives is the part I keep coming back to. A transparency unit would handle them, but slowly, frame by frame. And it drags an option from the first section back onto the shortlist: the camera. Camera scanning (a macro lens, a light source, a film holder) is how a lot of serious film digitization is done these days, and once the rig is set up it is fast. The negatives might end up being the one job where the “wrong” tool from the comparison wins after all.
That decision isn’t made yet, and I’m not going to pretend it is. The drawings, meanwhile, are already in the archive, odd colors and all. First in, as it should be.
The best scanner is the one that scans while you’re not standing next to it.
The Paper Trail series
- Paper Trail - the map of the whole project
- Paper Trail: Bootstrapping a Family Archive - the software bootstrap
- Paper Trail: Phone, Flatbed, or a Real Document Scanner - this post