3. Reading the Screen Specification Instead of Believing It
What the panel name is worth
Carta 1200, Carta 1300, Kaleido 3, Gallery 3, reflective LCD — these are manufacturer revisions of a product, not tiers of technology. A newer device does not automatically carry the newer layer: the Carta 1300 generation sits in the Kindle Paperwhite and Boox Note Max, while the previous Carta 1200 still ships in 2026 devices like the Boox Go 10.3 Gen 2 Lumi. Each generation's claim is comparative and measured under fixed conditions — 25°C, standard waveform, light at 45 degrees, front light completely off. Contrast reaches about 20:1 on Carta 1300, mostly from a darker black; reflectance stays near 44 percent across both generations, so a new panel does not give you a brighter page in a dim room.
The three costs of colour
Kaleido 3 laminates a red, green and blue filter mask over an unchanged monochrome layer. Hence the doubled specification every colour device lists: 300 PPI mono, 150 PPI colour — half in each direction, a quarter of the addressable area. Light crosses the mask twice, dropping white-point reflectance from around 40–44 percent to roughly 15–23 percent, which is why a colour reader is a reader you light. Measured gamut lands near 30–40 percent of sRGB: pastel, not tablet colour. And the mask cannot be switched off, so plain text carries a permanent faint diagonal grain. Gallery 3, on the reMarkable Paper Pro at 229 PPI, moves coloured pigments instead and pays in refresh speed.
Distance, measure, and a five-minute test
Pixel density only means something relative to viewing distance: about 60 pixels per degree for 20/20 vision puts the invisibility threshold near 344 PPI at 25 cm and around 195 at 45 cm. Reflowable text has a lever — font size — and one target: 45 to 75 characters a line, 66 ideal. Fixed PDFs have no lever, so screen width does the work; a 6-point figure label on a 7-inch screen shrinks to roughly a millimetre. Two checks follow, run on your own document at your real distance.
Four recent items
The Boox Palma 3, announced 15 September 2026 at about $340, is a Carta 1200 pocket reader — announced, not shipping. reMarkable OS 3.28 recalibrated greyscale tone mapping and rebuilt the EPUB engine. Libby now automates reciprocal lending where agreements exist. And the front-light-free Kindle Scribe reached US buyers in June at $429.99.
Nobody is buying anything today. That matters, because the moment a screen specification gets attached to a purchase, it stops being information and turns into a reason. And a screen number that arrives as a reason is very hard to argue with. So before the requirement sheets, before the short list, before any of that, we are going to take the columns on the comparison page that describe the screen and ask a narrower question of each one: what is this number evidence of?
There are four of them. Screen technology, which names the panel. Pixels per inch, usually written as PPI. Screen size, in inches across the diagonal. And colour, which says whether there is any and, if there is, which kind. Four columns, and every one of them is a measurement someone took under conditions. Strip the conditions away and a bigger number is not a better page. That is the whole lesson, and everything that follows is the same idea worn down into checks you can run.
Start with the panel name, because it is the column people trust most and understand least. On the comparison page you will see entries like Carta 1200, Carta 1300, Kaleido 3, Gallery 3, and on one device, reflective LCD. These are not categories of technology. They are a manufacturer's revisions of a product. E Ink builds the electrophoretic layer we took apart last time — the capsules, the pigment, the electrodes — and when they improve the chemistry or the driving, they ship it as a new number. Carta 1300 is the current mainstream monochrome layer. It is in the Kindle Paperwhite, in the Boox Go 7 and the Go 6 Gen Two, in the thirteen-and-a-third-inch Boox Note Max, in the Supernote A5 X2 Manta, in the PocketBook Q. Carta 1200 is the generation before it, and it is still shipping in current devices: the Boox Go ten-point-three Gen Two Lumi uses it, so does the Supernote A6 X2 Nomad, so does the Meebook M8. Older Carta panels turn up on the small pocket readers, the Xteink X3 and X4 among them.
So a newer device does not automatically carry the newer panel, and a device carrying the older panel is not therefore old. The Go ten-point-three Lumi is a twenty twenty-six device on a Carta 1200 layer. That already tells you the generation label is a component choice, not a verdict on the product.
Now the part that decides whether the label means anything. E Ink states what each generation improves. Carta 1200 against Carta 1000: about fifteen percent better contrast, taking the nominal contrast ratio from around fifteen to one up to roughly seventeen or eighteen to one, and about twenty percent faster response. Reflectance — how much of the light hitting the screen comes back to your eye — was left essentially unchanged, around forty-two to forty-four percent. Carta 1300 against Carta 1200: about twenty percent better contrast, rated near twenty to one, coming mostly from a darker black rather than a whiter white. Response time faster by roughly twenty-five to twenty-seven percent, and integrators who build stylus devices claim up to a thirty percent cut in pen input latency. Reflectance again around forty-four percent, unchanged.
Read those two sentences again and notice something. Every claim is comparative and every comparison is against the immediately previous generation. Twenty percent better contrast does not mean twenty percent better than whatever you are holding. It means twenty percent better than Carta 1200, measured a particular way. And the gains compound in a modest direction, not a dramatic one: the improvement across two generations is in the black ink getting blacker, while the white page stays about as bright as it was. If you were hoping a new panel generation would give you a brighter page in a dim room, it does not. That is not what improved.
Then the conditions, which are where most of the honesty lives. E Ink's optical numbers come from module specifications, and those specifications pin down the room. Ambient temperature twenty-five degrees Celsius, give or take two. Relative humidity fifty percent, give or take five. The panel driven by the factory look-up table calibrated for twenty-five degrees — which is to say, the standard waveform, the standard refresh recipe, not a fast mode. The light source sits at forty-five degrees off the perpendicular to the screen. The sensor sits dead on at zero degrees, looking straight at the middle of the panel through a very narrow aperture, two degrees or less. Measurement is made with a spectrophotometer or a colorimeter against a calibrated white reflectance standard, a block of PTFE or barium sulphate that reflects almost everything. And contrast ratio is defined simply as white-state reflectance divided by dark-state reflectance.
One more condition, and it is the one to hold on to: all of those baseline figures require the integrated front light to be completely unpowered.
Put that together and you have the shape of a real question you can ask of any generation label. Improved compared with what, and measured how. Twenty to one contrast means twenty to one in a twenty-five degree room, on the standard waveform, with light arriving at forty-five degrees, with the front light off, read by an instrument that only accepts light coming straight back at it. You do not read like that. You read at whatever temperature your bedroom is, at whatever angle your lamp happens to be, often with the front light on, and frequently on a fast refresh mode because you like page turns that do not flash. Every one of those departures moves you off the measured condition. Cold slows the pigment. A faster refresh recipe narrows the contrast, which we went through in detail last time. The front light adds light in front of the ink, which lifts the dark state along with the white and therefore lowers the contrast ratio even as the page looks easier to see.
None of that makes the number a lie. A twenty to one panel really is better in the dark state than a seventeen to one panel, and under identical conditions you would see it. What it makes the number is conditional. The same panel generation, in two different devices, in two different rooms, at two different refresh settings, will not deliver the same page. Which is why the panel column tells you what part is inside and roughly what it was rated to do, and tells you almost nothing about whether you personally will find the page comfortable. That second question needs the other three columns.
Colour is next, and colour is where the arithmetic gets interesting, because you can work out the cost yourself from the specification as it is printed.
Last time we built up a monochrome pixel: a little region of capsules whose pigment can be driven white, driven black, or stopped part way for one of roughly sixteen greys. That is all any of these pixels can do. They have no colour in them. So how does a Kobo Libra Colour show you a red cover?
The answer for the most common colour panel, E Ink Kaleido 3, is a colour filter array. A mask of tiny red, green and blue filters, laid out in a repeating pattern, is laminated over the top of the monochrome layer. Nothing underneath has changed. A pixel under a red filter still only goes light or dark. What the filter does is decide which part of the spectrum gets through on the way in and on the way back out. Drive the pixel under the red filter to white and it returns red light. Drive it dark and it returns very little of anything. To make one coloured dot on the page, the panel has to combine several neighbouring pixels sitting under different filters, each contributing one component, and let your eye average them into a single colour.
That combining is why the specification is quoted twice. Every Kaleido 3 device on the comparison page lists the same pair: three hundred PPI for monochrome text, one hundred and fifty PPI for colour. Kindle Colorsoft, Kobo Libra Colour, Boox Go Color 7, Boox Note Air5 C and Note Air6 C, the Palma 2 Pro, the Bigme colour models — all three hundred and one hundred and fifty. One hundred and fifty is half of three hundred, and it is half along both axes, horizontal and vertical. Halve a length twice and you quarter an area. So the colour image is assembled on a grid with one quarter as many addressable points as the text grid. A coloured detail has four times less area to live in than a black detail on the very same screen. The three hundred number and the one hundred and fifty number are not two opinions about the same panel. They are two different grids, and which one you are using depends entirely on whether the thing you are looking at is coloured.
The filter costs light as well as resolution, and for the same reason it costs resolution: everything has to pass through it twice. Light from your room goes down through the coloured mask to reach the pigment, bounces off the white pigment, and comes back up through the mask again. Each pass absorbs whatever that filter is built to absorb. Measured independently, the unlit white point of a native monochrome Carta panel returns somewhere around forty to forty-four percent of the light hitting it. On Kaleido 3, that falls to roughly fifteen to twenty-three percent. That is a loss of something like forty-five to sixty percent of the returned light. To get a Kaleido 3 page looking about as bright as an unlit monochrome page under ordinary indoor light, measurements indicate you need the integrated front light running at something like fifteen to forty percent. Which means a colour device is, in practice, a device you light. Whether that bothers you depends on where you read, and that is the reading-conditions part of your record doing work.
Saturation is the third cost. E Ink's claim for Kaleido 3 is four thousand and ninety-six colours — twelve-bit, sixteen levels per subpixel — and about thirty percent more saturation than the Kaleido Plus generation announced in twenty twenty-two. Independent measurement of the actual gamut puts it at roughly thirty to forty percent of the sRGB colour space, which is somewhere around twelve to fifteen percent of NTSC. A phone screen covers over ninety-five percent of sRGB. So the honest description is pastel. Colours are present, distinguishable, useful — and muted. If you have imagined comic panels looking the way they look on a tablet, adjust that picture down substantially before you make any decision based on it.
And then the one that catches people by surprise, because it affects the pages that have no colour in them at all. The filter mask is physical and it is permanently there. It cannot be switched off for a page of plain prose. Independent evaluations of the Libra Colour, the Colorsoft and the Go Color 7 all describe the same thing: even when the device is rendering pure black text at its full three hundred PPI, the white background carries a faint fine diagonal grain — people reach for words like sandpaper, or the screen-door effect — and looks slightly dingy next to a native Carta 1200 or 1300 panel, with slightly softer letter edges. You are trading a small permanent tax on every text page for colour on the pages that have it.
There is one colour approach that works differently, and it is worth knowing it exists so you can recognise it in the column. E Ink Gallery 3 does not use a filter mask. It puts multiple coloured pigments — cyan, magenta, yellow and white — into the capsules themselves and moves them selectively. Because no mask sits over the pixels, resolution holds closer to native and the light loss is a different problem. On the comparison page it appears on the reMarkable Paper Pro, an eleven-point-eight-inch panel quoted at two hundred and twenty-nine PPI with one hundred and fifty PPI in colour. Note that the base number there is two hundred and twenty-nine, not three hundred; the panel is a different product with different figures, and Gallery 3's own tradeoff is a slower refresh. Different mechanism, different bill.
So: colour is a real gain on pages that carry information in colour, and a real loss on pages of plain text. Which of those you are is not settled by whether you like colour. It is settled by counting what you actually read. Nadia reads about eighty novels a year and essentially nothing else. Her pages are black text on white. A permanent grain over every one of those pages, plus a front light she now has to keep on to get the brightness back, in exchange for colour she would see on a cover thumbnail — that is a bad trade for her reading, and it would be a bad trade even if the colour were free.
Which brings us to the column everybody reduces to a threshold, and the reason the whole chapter exists.
There is no PPI number that makes a page comfortable, and there is no screen size that does either. Comfort comes out of four things acting together: how many pixels there are per inch, how many inches there are, how far the screen sits from your eye, and how big you set the type. Change any one and the answer moves.
Distance first, because it is the one nobody puts in the comparison. Pixel density only matters relative to how much of your field of view each pixel occupies. Ordinary twenty-twenty vision resolves about one minute of arc, which works out to roughly sixty pixels per degree of visual angle — beyond that, you stop seeing pixels and start seeing an image. People with sharper vision, twenty-fifteen or better, get up to around ninety pixels per degree. Run that through the geometry and the density at which pixels become invisible depends on distance in a way that is not subtle. At twenty-five centimetres, close handheld reading, you would need something like three hundred and forty-four PPI. At thirty to thirty-five centimetres, a typical e-reader distance, around two hundred and fifty to two hundred and ninety. At forty-five centimetres — a desk, a lap, a big page — about a hundred and ninety-four to two hundred for twenty-twenty vision, rising to around two hundred and ninety if your eyes are unusually good.
So the same three hundred PPI panel is comfortably past the threshold at forty-five centimetres and slightly short of it at twenty-five. And a two hundred and nineteen PPI pocket reader, like the Xteink X4 Pro, is a different proposition held at arm's length than it is held close to your face in bed. The column cannot tell you which you are doing. Only you can.
Beyond roughly three hundred PPI at forty to forty-five centimetres, the pixel grid has effectively vanished, and what you notice instead is contrast between the pigment and the background, the quality of the anti-aliasing on the letter shapes, and how clear the front-light stack is. Which is exactly where the panel generation conversation comes back in, and exactly why past a certain density more PPI stops being the thing to want.
Now the second lever, and here reflowable text and fixed pages split apart completely.
Reflowable text — an EPUB novel — has no fixed page. The device lays the words out to fit whatever screen and font size you give it, and reflows when you change either. So the screen's job is not to hold a particular page. Its job is to hold a comfortable number of words. Typographic practice has a settled answer for what comfortable means: somewhere between forty-five and seventy-five characters per line including spaces, with about sixty-six treated as the ideal measure. Go past eighty or ninety characters and your eye starts losing its place on the return sweep. Drop below forty and the line breaks come so often that the reading rhythm keeps stalling. In practice, readers on six and seven inch devices tend to settle on a font size that gives them roughly a hundred and fifty to two hundred and fifty words on screen, somewhere around fifty to sixty-five characters across eighteen to twenty-five lines. A printed novel page, for comparison, runs about two hundred and fifty to three hundred and fifty words.
That is the whole legibility question for Nadia, and it resolves gently. She reads in bed with the lights off, one-handed. If the type is too small for her, she raises the font size, and the text simply reflows into fewer words per page. The screen only has to be wide enough that at her chosen size she still lands inside that forty-five to seventy-five character band. A seven-inch screen gives her plenty of room to do that. Push font size high enough on a four-inch pocket reader and you can fall under forty characters a line, and then the page starts fighting her. But there is no threshold to hit and no minimum inch count. There is a lever she controls, and one check to run.
Owen has no such lever, and this is the point of carrying him. He reads engineering papers and two textbooks — roughly eight hundred PDFs on his laptop plus a university library portal — at a desk, in daylight, two hands, about forty-five centimetres out. A PDF page is fixed. The type is baked into the layout at the size the publisher chose. He cannot raise the font size; he can only change how big the whole page is rendered, which means the physical width of the screen is doing for him exactly what font size does for Nadia.
Work it through with real geometry. Academic formats — the IEEE and ACM templates, journals like Nature — put their content in a printable width of about a hundred and seventy to a hundred and eighty-five millimetres. Figure legends, axis labels, subscripts inside those figures are typically set at six to eight point, giving capital letters somewhere between two point one and two point eight millimetres tall. Now fit that page onto a screen. A six or seven inch reader has a usable width of roughly ninety to a hundred and five millimetres, so the full page has to shrink to about fifty or sixty percent. A six point label becomes about three or three and a half point — around a millimetre tall. That is not small, it is gone; he would be zooming and panning on every figure, which for someone whose complaint is constantly jumping between pages, tables and references is the complaint made worse. A ten-point-three-inch panel, around a hundred and fifty-seven millimetres wide in the usual ratio, scales to about seventy-five to eighty-five percent, and a seven or eight point label lands near five and a half to six and a half point, which holds up at forty-five centimetres. Eleven to thirteen inch panels, a hundred and sixty-eight to two hundred and three millimetres wide, render near one to one and keep those six point annotations at the physical size they were drawn for.
And density decides whether a shrunk label survives as letters rather than smudge. At forty-five centimetres, twenty-twenty vision resolves down to about a tenth of a millimetre, which is where that hundred-and-ninety-five PPI figure came from. But resolving a detail and rendering it are separate problems. A five or six point stroke scaled down is something like fifteen to twenty hundredths of a millimetre wide, and to draw that without the letter closing up or the stroke pooling into a blob you want two to three actual pixels across the stroke. That is a three hundred PPI monochrome requirement. Two hundred to two hundred and twenty PPI breaks those labels. And a colour filter layer running its colour grid at one hundred and fifty PPI breaks them too — which is why, for Owen specifically, a colour device is not a neutral addition. His figures are frequently coloured. Colour-coded traces at one hundred and fifty PPI, with the labels on those traces at six point scaled down, is the exact case the colour grid cannot serve.
So Nadia's verdict from these axes is that almost any modern seven-inch panel will do, and the decisive question lies elsewhere. Owen's verdict is that screen width is close to the first thing he must solve, and density under it, and that colour is a cost rather than a feature. Same four axes, opposite readings.
Beatrix is a third reading again, because she barely reads. She writes — meeting notes, diagrams, contract annotations — on her knee, in indoor office light. Word count per page is meaningless to her. Her screen question is about writing area: how much surface there is to put a hand and a diagram on, at what distance, with the page at whatever size her handwriting naturally comes out. Her legibility problem is not reading text someone else sized; it is whether her own diagram fits without her shrinking her writing to make it fit. Pixel density matters to her through the pen line, which is where that claimed reduction in stylus latency on the newer panel generation earns a look — with the same question attached, of course. Compared with what, and measured how.
Here is what to do with a device you already own, today, and it takes about five minutes.
First, the reflowable check. Hold the device where you genuinely read it — in bed, on the sofa, wherever, not held out for inspection — and set the font exactly where you actually set it, not where you think a sensible person would. Open a novel. Count the words on a full page. Then count the characters across three prose lines and average them. If you are under forty-five characters a line, the screen is too narrow for the size you need and every page is breaking your rhythm. If you are over eighty without a column layout, your eyes are doing extra work finding the start of each line. Somewhere in between and the screen is not your problem.
Second, the fixed-layout check. Open the widest, densest diagram or PDF you own — a multi-panel figure with axis labels, not a clean one-column page. Fit it to the width of the screen, crop the margins if your reader lets you, and look at it from your real reading distance without zooming, without panning and without turning the device sideways. Find the smallest label on it: a tick mark, a subscript, a legend entry. Can you read it? That single yes or no is worth more than every number in the PPI column, because it is a measurement taken on your document, your eyes and your distance.
The screen-related columns at ocdevel.com/ereaders are there to be interrogated with those two answers in hand. Screen technology tells you the panel and therefore which comparative claim applies. PPI tells you the monochrome density, and the colour column tells you whether a second, quarter-sized grid is hiding underneath it. Screen size tells you how much a fixed page will have to shrink. The site's own total score is weighted, and it says so: front light, text clarity, weight, document transfer, price, battery and physical controls feed the score directly, while colour, screen size, RAM, stylus and audio stay visible and filterable without adding points for general reading quality. That is a set of weights chosen for ordinary book reading. Owen's reading would weight screen size enormously and colour negatively. Beatrix would weight the stylus into the scoring that OCDevel deliberately leaves out of it. A weighted score you can see the weights of is evidence. Read it that way.
Nothing has been recommended here, and no device has been named as a choice. The buying decision comes several steps later, after access and handling and ownership, and it will be made from requirements, not from panels.
What changed recently
Four things worth your attention, and the first is an announcement rather than a shipping capability. Boox announced the Palma 3 on the fifteenth of September twenty twenty-six, at around three hundred and forty dollars, marked coming soon. It is a six-point-one-three-inch monochrome pocket reader at three hundred PPI on a Carta 1200 layer, with dual-tone front lighting, an aluminium frame, six gigabytes of memory, a hundred and twenty-eight gigabytes of expandable storage, Android 16, and support for an active stylus. Note the panel generation against what we just did: this is the older monochrome layer in a new device, which is a component choice, not a flaw, and not an upgrade over a Carta 1300 reader you might already own. Announced is not shipping. If you are tracking it, the concrete action is to wait for units in hands and for someone to time a page turn and look for ghosting, the two checks from last time.
Second, reMarkable shipped operating system version 3.28, announced on the fourth of September and rolling out globally in early September, across the Paper Pro, the reMarkable 2 and companion hardware. Two parts of it touch what we covered today. They recalibrated the greyscale tone mapping and the contrast curve, which is a firmware change to how tones separate on the panel you already have — a reminder that contrast is not purely a hardware property. And they rebuilt the EPUB formatting and rendering engine with updated typography. If you own one, the action is direct: open a novel you know well after the update and run the character-per-line count, because a new text engine can change your line measure without you changing a setting. There is also a new circular selection tool that pulls a region out of a PDF straight into a notebook, which is the kind of thing worth testing on an actual figure rather than trusting on a release note.
Third, book access rather than screens. Between the eighteenth of August and the second of September, OverDrive shipped automation for reciprocal lending agreements in Libby, live now across library consortia in North America and elsewhere. Where partner library systems have an agreement, patrons no longer have to register and link each secondary card by hand; Libby discovers the partner systems and gathers them under a Partner Libraries menu, and you borrow and send loans through your primary home card. For anyone whose reading depends on library availability — Nadia's whole supply, and Owen's portal is a different animal — the action is to open Libby, look for that Partner Libraries entry, and check whether a title your home branch never has is now reachable. This is conditional on your library actually holding a reciprocal agreement, so the check is the only way to know.
And a smaller note on the same theme as the panel discussion: the front-light-free Kindle Scribe, announced in twenty twenty-five, reached United States buyers in June, at sixteen gigabytes and four hundred and twenty-nine dollars ninety-nine. Same eleven-inch three hundred PPI writing surface as its front-lit sibling, same four hundred grams, with the light removed. Before anyone treats an unlit panel as the purer reading experience, run today's second check in the room you actually read in at night. An unlit panel returns only the light already there. If the answer is a lamp, that is a decision you are making, not a spec you are avoiding.