2. Why the Page Stays There When the Power Goes Off

One screen, one bargain

An electronic paper page holds itself without power and is lit by whatever light is already in the room; in exchange it is slow to change. Inside the layer are millions of clear capsules, thirty to a hundred microns across, filled with oil, negatively charged white titanium dioxide and positively charged carbon black — soot and white paint shoved into position by electrodes, with roughly sixteen grey levels made by stopping the pigment part way (how electrophoretic displays are built, and the original E Ink patent).

The pigment stays put when the voltage stops because short-range stickiness beats gravity at that scale, additives damp the random jiggling, and the charge around each particle stays balanced. That bistability — not a bigger battery — is where every long-runtime claim comes from.

The cost: particles must be dragged twenty to forty microns through oil, taking tens to hundreds of milliseconds, and slowing further when cold. A full clearing refresh shakes the particles loose, drives the whole panel to black then white — that is the flash — then places each pixel, in about four hundred to a thousand milliseconds (the shake–erase–drive waveform). Skip those steps and you get partial refresh at roughly two hundred milliseconds, or around a hundred for panning — paid for in narrowed contrast, faked speckled greys, and ghosting from unfinished pigment, stuck particles and ions piling against capsule walls. No setting escapes the curve.

Handwriting survives only by bypassing the normal drawing path: pen-first makers claim under twelve milliseconds, independent high-speed video puts real note-takers at seventeen to thirty, while ordinary Android drawing apps land between eighty and over a hundred and fifty. Video and smooth scrolling stay bad.

Reflective LCD, by contrast, twists in place in single-digit to low-double-digit milliseconds, runs at video rates, and loses the picture when power stops. Three shapes cover nearly every screen offered: reflective and bistable and slow, reflective and fast, emissive and fast. The devices then sort into four kinds — closed reading appliance, open reading system, general Android e-paper, and pen-first note-taker — with no recommendation made here, because resolution, colour, light, grip and file transfer have not been covered yet.

Three checks on any device you can touch: time one page turn and note whether it flashed; change part of the screen and hunt for remnants; write one line and watch the gap between nib and ink.

Four devices and a price rise

Onyx is shipping the Note Mini C and the Note Air6 C — the latter keeping the Air5 C screen, body, memory and battery with a new processor and software. The fifty-eight gram, four-inch Picco is announced only, due before year end. Amazon's base prices have risen: entry Kindle at $149 with lockscreen ads, $169 without, Paperwhite at $199. iReader's Air 3 Pro ships in China, import only, no local warranty.

Last time we wrote down what you read. Now we look at the thing you would read it on, and we do it before anyone says a brand name, because almost every argument about e-readers turns out to be an argument about one screen and one bargain it strikes.

Here is the bargain in a sentence. An electronic paper screen holds its page without using any power, and it shows that page using whatever light is already in the room. In exchange, it is slow to change. Everything people love about these devices and everything they complain about comes out of that single trade. If you understand it properly, you can pick up any e-reader, new or ten years old, and predict how it will behave before you turn it on.

Start with light, because that is the first difference and the one your eyes notice.

Your phone is an emissive screen. Emissive means it makes its own light and sends it out at you. Behind the glass there is either a lamp shining through a set of shutters, or millions of tiny dots that glow on their own. Either way, the screen is a light source pointed at your face. That is why a phone works in a dark room and fails in direct sun: in the dark it is the brightest thing available, and outdoors the sun overwhelms it.

Electronic paper does the opposite. It sends out nothing at all. It sits there like a printed page and reflects the light that is already in the room back to your eye. That one fact explains a surprising amount. It explains why an e-reader gets easier to read as the light gets brighter, and why it is genuinely comfortable on a beach at noon, which is the one place a phone is useless. It also explains why the screen goes dark when the room goes dark, exactly like paper does, and why these devices need a front light built in if you want to read at night. A front light is a small set of lamps around the edge of the screen that shine across the surface, lighting the page from the front the way a bedside lamp would, rather than shining out through it from behind.

That is already useful for Nadia. Her recorded complaint was not that her phone is hard to read. It was that her phone at night is too bright and too tempting. A reflective screen with a dim front light is a different object from a lamp held at arm's length. The sunny beach half of her record is handled too, and handled by the same mechanism.

Now the second half of the bargain, which is the part that makes these screens strange.

Inside the screen there is a thin layer of millions of tiny clear capsules, each one somewhere between thirty and a hundred microns across — a capsule is well under the width of a human hair. Each capsule is filled with a clear oil, and floating in the oil are two kinds of extremely small solid particles. The white ones are titanium dioxide, the same white pigment used in paint. The black ones are carbon black, which is essentially soot. Both kinds are treated so they carry an electric charge, and crucially they carry opposite charges: the white particles are negative and the black particles are positive.

Above and below this layer there are electrodes. The one in front is transparent, so you can see through it. The one behind is divided up into a grid, one patch per pixel, each patch able to be switched on its own. Now the trick is simple. Opposite charges attract and like charges repel. Put a positive voltage on the patch behind one pixel, and the positive black particles are pushed away from the back and up to the front where you can see them. Light coming into that pixel hits soot and is absorbed, so the pixel looks black. Reverse the voltage and the white particles come to the front instead, scattering the room's light back at you, so the pixel looks white. Hold the voltage for a partial amount of time and you can leave the particles part way, mixed, at some depth in the oil, which is how these panels produce grey — typically sixteen levels of it.

So the picture on an electronic paper screen is not a pattern of light. It is a physical arrangement of pigment. You are looking at soot and white paint that has been shoved into position.

Which brings us to the retained image, and this is the fact that changes what an e-reader is. Once the particles have been pushed where you want them and the voltage is switched off, they stay. There are three reasons, and they are all reasons of very small physics. At these sizes, the short-range stickiness between particles, and between particles and the capsule wall, matters far more than gravity does, so the pigment is effectively pinned in place rather than sinking. The oil has additives in it that get in the particles' way and damp out the random jiggling that would otherwise let them wander apart. And the chemistry around each particle keeps its charge balanced once the outside push is gone.

The engineering word for this is bistable: the screen is stable in either state, with no power needed to keep it there. A page on an e-reader costs electricity only at the moment it changes. Showing it costs nothing. Leave the device on a shelf for a month with a page displayed and that page is still there, not because the battery kept it alive, but because nothing moved.

This is where every battery claim you will ever read about these devices actually comes from. It is not a bigger battery. It is that the screen is, almost all the time, drawing nothing. The power goes on turning pages, on the wireless radio, on the front light if you use it, and on whatever software is running — never on the page itself. That is also your first warning about battery numbers, which we will take apart properly when we get to ownership: a runtime figure is really a figure about how often you change the page and how bright you keep the light.

Now the cost side of the bargain, and it follows straight from what we just said. Changing the page means physically moving material. Solid particles have to be dragged through oil, and that is a matter of fluid drag, not electronics. The physics here is the same physics that makes it slow to stir honey. A particle's speed depends on how hard the electric field pushes it, divided by how thick the fluid is and how big the particle is. The particles have to cross a gap of roughly twenty to forty microns, and at the voltages these panels use, they creep across it. Getting pigment all the way from one side to the other takes somewhere between tens and hundreds of milliseconds of continuous pushing. And because the limit is the thickness of the oil, cold slows it down further — a colder screen has thicker oil and needs longer pushes. No faster processor fixes any of this. You are waiting on liquid.

So when you tap for the next page, the controller does not simply light up a new image. It runs a recipe of voltage pulses, timed frame by frame, and that recipe is where the whole speed-versus-quality argument lives.

Consider the careful recipe first, the one called a full clearing refresh, which is the flash you have probably seen an e-reader do. It has three parts. First it shakes: it applies rapid alternating pulses that jiggle the particles loose from the capsule walls and break up any clumps that have formed, without actually sending them anywhere. Then it erases: every pixel on the screen is driven all the way to full black and then all the way to full white, so that nothing anywhere on the panel remembers where it used to be. Only then does it drive: from that known clean starting position, precisely timed pulses put every particle at the depth its new grey level requires. That erase step, driving the whole panel to black and then white, is the flash. It is not a glitch and it is not the screen thinking. It is the screen deliberately wiping itself to a known state. The whole sequence takes something like four hundred to a thousand milliseconds, call it half a second to one second, and the payoff is a page with clean blacks, clean whites and no leftovers.

Leftovers are the other thing to name, because they are the most common complaint people have about these screens without knowing what they are looking at. Ghosting is when faint remnants of the previous page are still visible underneath the new one: a shadow of the old paragraph, or the outline of a menu you closed. The causes are exactly what you would now expect. If the pushing pulse is cut short, some of the pigment simply does not finish crossing, so a little of the old colour is left near the surface where you can see it. Particles that have been sitting still for a few minutes stick harder than ones that just moved, so the same pulse moves different pixels by different amounts depending on what each has been doing. And over long stretches of driving, stray ions in the oil pile up against the capsule walls and set up their own small opposing field, which tugs particles back toward where they were until a shaking cycle clears it.

That last one is why the recipes have to be electrically even-handed — over a full cycle the pushing has to balance out, as much in one direction as the other. A recipe that cheats on that to save time can leave charge accumulating in the panel, which degrades the image and, over years, is not kind to the electronics behind it.

Now the faster modes. A partial refresh only rewrites the rectangle of the screen that changed, and it skips the shaking and the erasing. It just drives the changed pixels straight to black or white, with no attempt to produce true greys. That runs in roughly two hundred to two hundred and sixty milliseconds. There is a faster mode still, meant for panning and scrolling, which truncates the pulses even harder and accepts that the pigment will not finish its journey — around a hundred to a hundred and twenty milliseconds. And there are middle recipes that compute a direct move from one grey to the next without the inversion flash at all.

Here is the thing to hold on to: these are not better and worse settings. They are positions on a curve, and the curve has no best point. Speed is bought with three specific losses. Contrast narrows, because pigment that did not finish arriving leaves black looking dark grey and white looking slightly dirty. Greys get faked, because a mode that can only do black or white has to approximate a middle tone by scattering black dots in a pattern, which you see as speckle and rough edges. And ghosting increases, because you are skipping the two steps whose entire job was to prevent it. Choose fast and you are choosing those three costs. Choose the full clearing refresh and you are choosing the wait and the flash. A well-made device makes that choice sensibly for each task and lets you adjust it; there is no setting that gets you out of the trade.

Which is exactly why the same screen can be a joy for one person and a frustration for another, and why we spent the first lesson writing down what you read.

Take Nadia's novels. A novel changes the entire page, all at once, roughly once every minute or two. Nothing about that is hurried. She can afford the full clearing recipe every single time: about half a second of wait and a flash she will stop noticing within a day, in exchange for text as clean as the panel can make it. Her reading shape sits at the easy end of the curve. The demands her record actually makes on a device are elsewhere — the dim warm light, the one hand, the library book — not on refresh speed at all.

Now take Owen. His textbook page has a figure spanning the full width, and his working habit is to zoom into part of that figure, drag around inside it, and jump back to a table four pages earlier to check a value, over and over. Every one of those actions changes part of the screen, frequently, and he cannot wait half a second each time. So his device will be running the fast partial modes most of the day, which means he gets the fast-mode costs most of the day: ghosting around the edges of what he panned, greys turning speckly in the middle of a plotted curve, contrast a little flat. His laptop never made him think about this, because a laptop screen redraws sixty times a second without being asked. That is precisely why his current habit maps badly onto electronic paper. It is also why, for him, how a device handles zooming and jumping is not a minor feature. It is the whole question.

And take Beatrix, whose demand is the hardest of the three. When you write with a pen, your eye expects the line to appear under the nib essentially as it happens. If the ink lags behind the tip, writing feels like driving with a delay in the steering. Remember that a full clean refresh takes around half a second: if handwriting went through the normal page-drawing path, it would be unusable. So devices that are serious about the pen do something different. The digitizer under the glass samples the pen position hundreds of times a second. Those coordinates are then routed around the normal software that composes the screen image and sent almost directly to the display controller, which fires one very short black-or-white pulse limited to the tiny box around the pen tip. Afterwards, when you pause, a quieter refresh runs in the background to tidy the stroke's edges.

The numbers tell you how much that engineering matters. Manufacturers of pen-first devices claim under twelve milliseconds from pen to visible ink; independent high-speed video measurements across several note-taking devices put the real figure at roughly seventeen to thirty milliseconds, which is fast enough that your hand does not notice. But the same measurements find ordinary third-party drawing apps running on a general Android e-paper device, using the standard system drawing path instead of that shortcut, landing somewhere between eighty and over a hundred and fifty milliseconds. That is a visible lag. So for Beatrix the question is not whether a device has a pen. It is whether the writing she does goes through the fast path, in the app she would actually use.

There is one more set of tasks worth being blunt about. Animation and smooth scrolling are what electronic paper is worst at, and no firmware update changes that. A screen whose pixels take a hundred milliseconds at best to change cannot show smooth motion, because smooth motion needs a new image every sixteen milliseconds or so. Video on these devices is smeared and stuttery. Scrolling a web page is worse than paging through it. If your reading life is built on scrolling feeds and watching things move, electronic paper is the wrong technology for you, and that is a fine thing to discover from a lesson rather than from a purchase.

Two other screens get confused with this one, and the difference in both cases is mechanism.

The first is a reflective LCD. Like electronic paper, it uses the light in the room: there is a mirror behind the liquid crystal layer, and light comes in, gets modulated, bounces off the mirror and comes back out. So it looks paper-ish in bright light and does not glare at you, and some devices are built around exactly that feel. But the mechanism inside is completely different. Nothing travels anywhere. Liquid crystal molecules twist in place to change how much light gets through, and they only stay twisted while they are being held that way, which means the panel has to be electrically refreshed continuously just to keep showing a still image. It is not bistable. Cut the power and the picture is gone. What you get in return is speed: in-place twisting happens in single-digit to low-double-digit milliseconds, so a reflective LCD can run at ordinary video rates without ghosting. Same use of room light, opposite position on the speed-and-retention trade. In OCDevel's table the Daylight DC-1 is the reflective LCD entry, listed there as a paper-feel device that is explicitly not electronic paper.

The second is the ordinary lit display in your phone or tablet, which is emissive, in-place and fast. Whether it is a backlight shining through liquid crystal shutters or organic dots glowing individually, the state change is electrical or molecular rather than mechanical, and it happens in milliseconds or fractions of one. It needs continuous power for every moment it is showing anything, and it puts light into your eyes. That is why it can play video and why it is the wrong companion at two in the morning.

So: reflective and bistable and slow, reflective and not bistable and fast, or emissive and not bistable and fast. Those three descriptions cover nearly every screen you will be offered.

With that, the market stops looking like a wall of brands and starts looking like four shapes, each one optimised for a different set of the demands we just described, and each one giving something up to get there. What follows is OCDevel's own carving of its comparison table, and I want to be clear before we start: there is no recommendation in this chapter, no price, no short list. The axes that decide between these shapes — resolution in the mode you actually read in, colour approaches, front light, weight and grip, how a document gets onto the thing — have not been taught yet. Picking now would be picking blind.

The first shape is the closed reading appliance. It is built around one store and one purpose: long novels, minimal friction, battery measured in weeks. The software is a stripped-down system of the vendor's own, tied to their catalogue and their cloud. What it gives up is openness — you cannot install arbitrary apps, sideloading other formats goes through the vendor's conversion or not at all, and open library lending across arbitrary library systems is not built in. In OCDevel's table the mainstream examples are the Kindle Paperwhite, a seven-inch monochrome reader on the current Carta 1300 panel at three hundred pixels per inch with a warm and cool front light, and the Kindle Colorsoft, the same size with a Kaleido 3 colour panel. OCDevel calls the Paperwhite line its value default, and separately notes that colour on a Kaleido panel halves the effective sharpness in the coloured parts of the page — a fact we will take apart when we do panels properly.

The second shape is the more open reading system. Still a dedicated reading device with its own lean software, still weeks of battery, but built to read formats nobody owns — EPUB, comic archives, PDF — and to borrow from public libraries directly on the device through the OverDrive and Libby system. It gives up the general app environment, competing stores' protected books, and serious pen work. OCDevel's current top non-Kindle pick here is the Kobo Libra Colour: seven inches, Kaleido 3, physical page-turn buttons, an asymmetric grip meant for one hand, and water resistance rated to survive a dunk. If you are keeping score against Nadia's open questions — direct library borrowing, warm light, one-handed page turns — you can see why that shape is where her sheet is pointing, even though we have not done the work to confirm it.

The third shape is the general Android e-paper device. Full Android with the Play Store, so you can run Kindle and Kobo and Libby and a browser and your own document tools side by side on one reflective screen. That flexibility is the whole point, and it costs two things. Battery drops from weeks to days, because now there is real software running. And every app you install was written for a screen that redraws sixty times a second, so you end up managing refresh modes yourself, trading ghosting against stutter, app by app. OCDevel's examples include the Boox Go 10.3 Gen II Lumi, a ten-point-three-inch front-lit monochrome tablet running Android 15, and the pocket-sized Palma line at just over six inches.

The fourth shape is the specialist note-taker, built around the pen rather than the page: large screens from around ten to thirteen inches, a textured writing surface, and a digitizer and drawing path engineered for the low latency we talked about. What it gives up is one-handed portability, and often the front light, because reducing the optical gap between pen tip and pigment layer matters more to writing feel than reading in the dark does. OCDevel's examples include the reMarkable Paper Pro at eleven-point-eight inches on a Gallery 3 colour panel with its own closed Linux system and no third-party apps, the Boox Note Max at thirteen-point-three inches, roughly A4, monochrome and unlit, and the Kindle Scribe at eleven inches, which puts pen tools alongside the Kindle store. Beatrix's record lives in this shape — and her open question, whether her notes can leave the device in a form other people can use, is a question about software and export, not about writing feel.

So here is what to do with all of this today, and it costs nothing.

Get your hands on any electronic paper device — a shop display, a friend's Kindle, a colleague's note-taker — and run three checks. First, turn one page and count the wait. Not a guess at it: watch from your tap until the new page is fully settled, and notice whether the screen flashed to black doing it. That tells you which recipe the device chose and how long you are giving up per page. Second, do something that changes only part of the screen — open and close a menu, or zoom into a figure and drag around it — then look hard at the background afterwards for faint remnants of what was there before. That is ghosting, and how much of it a device leaves in normal use is the fast-mode cost it decided to accept on your behalf. Third, if writing matters to you at all, write one line of ordinary handwriting at your normal speed and watch the gap between your nib and the ink. Not whether the letters appear, but whether they appear under the tip or trail behind it.

Then go to ocdevel dot com slash ereaders and look at the screen technology column. Every value in it — the current flagship monochrome panel, the previous generation, the two colour approaches, the reflective LCD — is now a thing you can read rather than a code. That column is evidence about where each device sits on the curve we just drew. It is not a verdict, and we are not ready for a verdict.

Four new devices and a price rise

Onyx has released two colour note-takers running Android 16, both shipping now: the Note Mini C, an eight-and-a-half-inch Kaleido 3 tablet with a front light, stylus support and Google Play access, and the Note Air6 C, a ten-point-three-inch notebook that keeps the previous Air5 C body, screen, memory and battery while updating the processor and software base and including the pen in the box. If you already own an Air5 C, read that carefully: the newer model is a processor and software step, not a new screen. Anyone comparing the two should set them side by side on the table's own comparison view rather than assuming the later number is better.

Onyx has also announced the Picco, a fifty-eight gram pocket reader with a four-inch monochrome screen at two hundred and thirty-five pixels per inch, physical page buttons, an adjustable warm and cool front light and sixteen gigabytes of storage. It skips Android for a lightweight system of Onyx's own, using their Tiles tool for direct transfer from a phone or browser. Announced is the word to hold onto — delivery is scheduled before the end of this year, and nothing about it has shipped. Do not put it on a short list yet.

Amazon has raised base prices: the entry-level Kindle now lists at a hundred and forty-nine dollars with lockscreen ads or a hundred and sixty-nine without, and the Paperwhite base model at a hundred and ninety-nine. That is live in retail. If you wrote down a budget before August, that budget is now out of date, and the ad-free step is part of the real figure rather than an extra.

And iReader has launched the Air 3 Pro, an eight-point-two-inch unlit monochrome writing tablet on a Carta 1300 panel at two hundred and ninety-three pixels per inch, with an EMR pen included and a memory card slot. It is shipping in China and reaches anyone else only by direct import, which means no local warranty path. Treat it as a reference point for what that size and panel can be, not as an option.

One page turn, timed, before you read anything else about any of this.