a book with nothing in it

In 2025, Trier put on a big exhibition about Marcus Aurelius: emperor, general, philosopher.¹ The half at the Stadtmuseum Simeonstift asked what good rule is, and it was about his writing more than his marble. That's a hard thing to put in a room. The Meditations are notes he wrote to himself, in Greek, with no reader in mind;² it's a dense text, and a museum is a place where people read standing up, for about as long as their feet allow. The research on exhibit labels says visitors read little and briefly,³ and so does everything I've seen on the floor.

The museum had seen a company that makes interactive books, with pictures projected onto the pages from above, and asked me: we've seen this, can we build it too? In this exhibition that kind of book wouldn't have worked well. But the question behind the request was a good one, and it wasn't technical: how close do you stay to a book? This article is about that question, and about the one technical decision that followed from it, how to know where a page is. It also records what half a year of visitors did to the sensing, because that changed what the measurement is for. The book is now in its second exhibition. At the end I say why a device like this gets built when it could be bought.

The book as a symbol

Paper? Pages? Or a screen with one of those page-turn animations from e-readers? Here the book is a symbol before it is anything else. It says that this corner of the exhibition is about a book, and for an exhibition about a text that is the most important thing the object says. The pages carry fragments of the text, mixed differently from page to page. Fragments are short, and short helps when you're standing. The Meditations are written in pieces already.

And the appeal is in the act. You open it, you turn a page, you watch the next one appear. A screen with an e-reader animation would have had the same text and nothing for your hands to do, and the hands hold a standing reader at a page for the length of a fragment.

Two books I didn't build

The most real version would have been a real paper book with the pictures projected onto it live. For a camera to know which page is open and how it lies, every page needs marks the camera can see and visitors can't, in an ink that shows in the infrared. The ink alone would have cost the whole budget, and there was no time to develop the rest. Even then, someone would have had to tell visitors where the book may lie so the camera could find it. An instruction like that puts the technology in front of the visitor, and that was the one thing the design was trying to avoid. This is where the idea got smaller.

The other easy version is flat: one sheet on the table, the pictures projected onto it. A rectangle is easy to track and needs no marks. But a flat sheet reads like a newspaper, not like a book, and the point was the feeling of opening a solid book and leafing through it. So the question wasn't paper or screen. It was which properties of a book carry the symbol. My answer: a spine, pages that turn on it, and the resistance of a page under the hand.

Plates on axles

So the pages became plates that turn on axles, angled up from the spine like a book lying open in a cradle. Four of them turn; the two outer ones stay put. Turning one feels right, and the light gets something more interesting to fall on than a flat sheet. My first version was corrugated cardboard, and the hinge is 3D-printed, the simplest thing that would do.

There's no actual cradle, though. The book sits on the exhibition's table, and the table is a white top with not much underneath it, in an exhibition design that is quiet on purpose.⁴ Whatever I built had to fit in without bulky technology. And with the number of visitors expected, the position of every plate had to be known exactly. Those two constraints, no room and exact knowledge, decided the sensing.

Knowing where a page is

I went through the options. A sliding contact, an end stop or a light barrier would have been hard to fit, too big or too fragile, and all three only tell you that a page has arrived, not where it is on the way. Rotary encoders and potentiometers have a mechanical part: they resist your hand, they make noise, and they wear out after so many turns. In a state exhibition a page is turned many thousands of times by people who aren't careful with it. Optical rotary encoders would have worked, but they're hard to find this small, and building my own would have taken more time than there was. What's left is a magnet.

A small magnet sits in the end of each axle, and right behind it, inside the rear hinge, a chip reads which way the magnet's field points.⁵ Nothing touches, so nothing wears out, and the chip knows the angle all the way round, not just at the ends. At the time, that last property was a nice-to-have. It became the thing that kept the exhibit running.

A board of my own

You can buy the sensor on prototyping boards, which are too big for a hinge, or as leftovers from camera gimbals, on small boards that are miserable to solder. And the plates had to sit close together, for two reasons: it should look tidy, without a big gap in the picture, and nobody should get their fingers caught. Four magnets that close to four sensors might disturb each other, so before drawing anything I ground the prototyping boards down until they fit side by side, and tried. They didn't disturb each other. Negative results like that never get written down, so here it is.

So I drew a board: four sensors in a row, twelve millimetres apart like the axles, and one more chip that lets the small computer tell them apart, because all four sensors answer to the same address.⁶ Along one edge there's a strip with four lights, for testing. Once a board worked, the strip got sawn off. Getting there took cardboard, 3D prints, working models and a breadboard, and only then a board in KiCad. An engineer would tell it in the other order. A designer starts from the idea, asks whether physics permits it, and picks up the tools of the electronic trade last, and uses them backwards.⁷

From magnet to light

The small computer in the book reads the four angles from the board and sends them on, over USB and as one line of text, to the computer by the projector, out of sight. The magnets sit in their axles at whatever angle they happened to end up in, so the first step is calibration: each raw reading becomes degrees relative to the table. From there it's TouchDesigner.⁸ Every page plays a short film, and when the film has played out, it fades into a loop that invites you to turn to the next page. A bigger state machine decides which films are up. The films themselves are by the visual designers Milena Biesel, who made most of them, and Daniela Schurig.

So the projection is the only technology you see, and it doesn't look like technology. There's no button and no screen, the sensors are inside the hinge, and the book itself asks you to turn the page. That reads like a principle. It's a consequence of the table having nothing underneath it.

The table moves

The tracking worked well from the start. Then the exhibition ran for half a year, and over half a year enough people lean on a table, or bump into it, that it wanders, bolted down or not. Every millimetre it moves shifts the plates under a projector that stays where it is. If the picture had depended on the exact angle of each page, someone would have had to re-map it every two weeks, and the technology would have been in the foreground again, as maintenance. Buildings move. Stewart Brand built a book on the observation that the parts of a building change at different rates and the fast layers tear at the slow ones.⁹

So the exact angle now decides only two things: whether a page is lying still, or whether it's on its way. While a page turns, nothing is shown on it. How flat counts as lying still is one number, in degrees, that I can adjust. With a sliding contact, adjusting that would have been fiddly. It's a strange way to use a precise sensor: I measure every page to a fraction of a degree and then mostly ask it one question. The rule I take from it is to measure more precisely than the first use needs, because the building will move, and the question you end up asking a sensor is rarely the one you built it in for.

What the house asks of a device

The book made it through the exhibition, which had a lot of visitors. Now it stands in the museum's exhibition on the 1920s,¹⁰ with new content and new fixings: the same book, a different dense subject.

Why does a thing like this get built at all? Museums aren't short of companies selling them interactive exhibits, and the same house has bought media players that do exactly what they should.¹¹ They're good: plug and play, a warranty, a manual for the staff, a housing, and they survive the power being cut every evening without warning, which is what happens in a museum. For most of what an exhibition needs, playing a film on a wall, you should buy one, and once the hours are counted the self-built version isn't even cheaper. The book still couldn't have been one of them, because nobody sells four axes in a book. There's a Mac mini in the table too, by the way, next to the projector. The project just doesn't live off it.

What the house asks of such a device isn't what a maker would assume. A museum collects objects that have stood somewhere for a few hundred years, and its sense of time is set accordingly. The text in this book has held up for about 1,850 years. The device showing it has to manage five, run by invigilators and the house's IT and not by me, without instructions, without a second attempt by the visitor, and inside somebody else's exhibition design. The lifespan is set by the house. Most of the decisions above, the magnet, the threshold, the board with its test strip sawn off, were made with those five years in view. Whether it reaches them isn't known yet. One exhibition and a half is the evidence so far.