Comment by derefr | Hacker News Reader

Comment by derefr | Hacker News Reader

Parent Full thread

So, why exactly is it that color e-ink displays try to make every micro-capsule capable of switching to any of the supported colors; rather than laying (groups of) CMYK micro-capsules out in a subpixel-like arrangement, and then wiring them to the substrate such that each of the CMYK "bitplanes" forms an independently-addressed display signal?

Sure, it'd look like badly-registered offset printing, but that only matters for small up-close displays, not large far-away displays. (And you could fix the registration with a precisely-cast diffuser layer, convolving each subpixel-cluster.)

If this was for reading rather than imagery, and you wanted to have true blacks, you could also just step one level forward in e-ink technology, and have the colored micro-capsule groups be just dual-tone (the capsule's color plus black), giving you {C+K, Y+K, M+K, K} bitplanes.

Are e-ink display manufacturers just imprecise in the way they deposit the capsules into the panel, making it impossible to address individual capsules? (If so, that seems like something that could be solved pretty easily with modern photolithography processes, e.g. etching onto the backing electrode a grid of "buckets" for individual micro-capsules to fall into.)

Yes, modern e-ink capsules are not precise. In fact, they're not intended to be.

If you look at an e-ink display under a microscope you'll see that the "pixels" are more like "shards" of variously-shaped cells.

https://twitter.com/sqfmi/status/1178018736684580865

The "pixels" are made of the actual addressing grid that creates the currents that attract the pigment in the capsules.

This is part of what ends up giving the appearance of a higher-resolution display when rendering text, as you get some softer edges "for free." (And it's what makes the display look 'blurry' for some viewers.)

So, to your question, there's currently no attempt with the technology to precisely place the colors on the display. The capsules are just spread all over, and the grid does the work.

This makes me sad. Based on what you're describing it's sounding like my dream of having eReaders that are anywhere near sharp enough to do visually rich, full color, documents justice (e.g. nature photography, comic books and graphic novels, well illuminated manuscripts) will never come to fruition. Barring some kind of transformative breakthrough, of course.

https://upload.wikimedia.org/wikipedia/commons/thumb/a/a6/Co...

Rods in eyes aren't on a regular grid anyway, why should it matter how the pixels are distributed as long as they cover the plane and are mostly uniform?

Because of what was mentioned way up in the original question: The capsules not being precisely arranged means that they can't be individually addressable. Meaning that you can't use the trick that non-epaper screens typically use where each "pixel" is actually a tightly packed cluster of individual red, green and blue sources that are individually controlled.

So instead, each cell is capable of displaying any color, and they've got a more complicated refresh mechanism that they use to "write" to them.

And the upshot of that is listed in the specifications from the linked page: The refresh time on this display is 15 seconds. Which is far, far away from where it would need to be in order to be a practical option for color e-readers.

Except you can do multiple colors, by introducing differently charged pigments inside the capsules:

- https://www.eink.com/assets/img/technology/GIF_Cup_E.gif

- https://www.eink.com/assets/img/technology/GIF_ACeP_E.gif

Refresh rate was always an issue, but it is gradually improving. For static reading it's not a problem.

I have the 2.7” 3 color (b) hat, and it is very slow with stock code. As far as I know, and from the looks of the Python code, 2 full buffers are transferred for the colors and then a process of wave stimulating the fluids into arrangement are how it works, very conservatively and slow. There has been work done to the firmware to allow partial updates fast enough for animation, but they should be used sparingly to avoid temporary or permanent burn-in, they say.

> Meaning that you can't use the trick that non-epaper screens typically use where each "pixel" is actually a tightly packed cluster of individual red, green and blue sources that are individually controlled

If there was a way to mix the inks and have them 100% separate by the grid on demand, it could be possible.

Who designed these ‘eye’ things anyway, sounds like they took a lot of engineering shortcuts and amassed a metric boat-ton of technical debt

Let it thus be observed that a truly intelligent designer understands how to live comfortably with technical debt.

If I was all-knowing and omnipotent, I'd be OK with that in my code as well.

Isn't tech debt a sign of not being all knowing and omnipotent?

There's always tech debt. It starts to day you build something and ends when it stops being used. How many thousands of python 2.7 programs had "no" tech debt when developed bit now have tons. Tech debt isn't binary, it's a sliding scale.

Seems like a classic-style omnipotent god could write anything conceivable in some eternal Lisp dialect that he maintains personally. If he’s really capable of violating thermodynamics he should be able use knowledge of the future to ultra-waterfall the design and implement it flawlessly thereby completely avoiding accumulation of technical debt.

It’s fun to imagine Jesus returning in 30xx AD but this time instead of absolving the world of its accumulated sin he wipes away the accumulated technical debt of 1.5 millennia worth of corporate IT

If you were "all knowing", you would've known that Python 3000 would not be backwards compatible and waited until it is released (assuming that you don't have a deadline, such as 7 days).

Even if you had a deadline, you'd know all tech debt and could anticipate it beforehand.

The joke is (and it is a joke, mind you) that an omnipotent and omniscient entity could have written a perfect program in, say, 2005, and it would have no tech debt now, because it's never needed to change since it was written to anticipate every need for the program in the first place.

Hell, since the writer was omnipotent and omniscient, the code was even built with the ability to perfectly modify itself when the company changed the DB on September 12, 2011 at 3:24 in the morning.

You mean all the Python programs started before 2008, right? Because that's when it was announced that they were sunsetting Py2.

So any projects started since 2008 that were written without Py3 compatibility had known tech debt they were choosing to take on.

Yes. Why does it matter than the particular example I chose had a cutoff date 2008? It's the same issue with any language, program, platform, OS, etc. Even if you implement the latest stable version of any of these things for any project, tech debt begins immediately. In fact if you're dealing with established enterprise software, you know right away what the roadmap & minimum end-of-life is from the day it's released. (as side topic, that's part of why mediocre enterprise software is often chosen over newer better options: because predictability is valued over the newest/best/fastest that may be subject to disruption if the vendor goes away or gets acquired-- something I've seen happen a few times. A vendor gets acquired, we get a "Hey Great News!" message from them along with an "Oh by the way our new parent company is transitioning everyone off our platform to their own completely different offering over the next 6 months, hope you have the capacity/bandwidth to deal it #sorry-not-sorry"

> over the next 6 months

My point is this is a very different situation from the Python 2->3 transition.

If you have 12 years' notice to make some changes, I don't consider that technical debt: that's the normal price of maintaining software in a changing world. It's not "debt" that costs "extra interest" to pay back for questionable choices made previously.

No. Tech debt, like money debt, can be taken on in purpose to be able to take advantage of a short term business opportunity.

The joke is (and it is a joke, mind you) that if you're all knowing and omnipotent, there's no such need to take on tech debt. Because you will have already known about the short term business opportunity and coded it (without tech debt since you also know all of the requirements, pitfalls, and corner cases perfectly) before it was needed.

I don't know.

Let me introduce you to the fall of man.

"How did we justify this whole 'cancer' thing being SEV-1? I really don't see the urgency."

It's what happens when your rookie engineer leans too heavily on generative design instead of doing their homework. :P

Product management kept changing the requirements!

colour printing is often made from randomised dot screens, so long as they are small enough its fine.

I think the solution is a dual eink and oled display. Overlay a “transparent” oled on top of an epaper screen and use the oled to display colour images and video while using the epaper for static monochrome regions.

I’m not sure that current “transparent” oled technology is transparent and thin enough though.

This sounds really similar to https://www.mobvoi.com/us/pages/ticwatchpro which is a Smartwatch with a dual layer display, one of them is e-ink, the other is an AMOLED display.

Do a search for hybrid eink LCD display; Apple and others were looking into this 10 years ago; the Pixel Qi's 3qi displays were apparently manufactured and used in at least some products, then Apple applied for a more complex patent as well. As far as I can tell, both systems seem to use a layered display.

https://www.pcworld.com/article/165724/bye_bye_kindle_netboo...

https://arstechnica.com/gadgets/2011/04/apple-exploring-hybr...

eink should die already. There are other options rather than eink, microLED and some of the newer low refresh LCDs (Clearilk comes to mind) should be much better. https://www.displaydaily.com/article/press-releases/clearink... Lenovo is interested and made a thinkpad with eink so it shouldn't come from a weird no name company in the future.

Do the others ones not consume power when you don’t need to change the display, are legible in daylight, or otherwise not emit light?

These are unique advantages to e-ink as far as I’m aware.

It doesn't matter for a stationary screen

It certainly does depending on your application. Grocery stores use them for prices where you wouldn't want a glow and low power means you can run off a battery. If I integrated into my walls as a smart device I would want a subtle design without a glow. These are just two examples off the top of my head. The application matters.

Also where did this stationary context come from?

People are asking about using it as a monitor in the comments. Grocery stores can do wireless power but the comments are complaining about them being too big or small.

You can easily have a non backlit black and white display, like the one on an f91w or LCD clock.

The contrast ratio will never make it good for text. I have used the so called high res ones and I would take a 720p LCD any day. They are trash.

As a reflective display, you want to optimise for the percentage of useful light reflected back to the user. Switching the white (or black) out for another colour will necessarily reduce the contrast ratio of the display, which is of course not ideal when you don't have much light to work with in the first place (as it's a reflective display after all). That's why reflective LCD displays were pretty much a disaster, since the LCDs require a polarisng filter which absorbs a large portion of the light, and reflective colour LCDs more so, because that multiplies with the effect of the colour filters.

Of course, you could simply make a matrix of tri-tone capsules, with white, black and colour. Then you have the contrast ratio and the colour space advantages. I don't know why this hasn't been done, I'm not an expert on e-Ink!

Colored e-ink displays are discussed at around 8:00 in this video; there are probably better sources, but this is the one I'm familiar with, so I figured I'd link it: https://www.youtube.com/watch?v=MsbiO8EAsGw

The microcapsules are not individually wired up to anything, so I don't think it's feasible to make each color channel independently addressable. Instead, a grid of wires behind the capsules are either positively or negatively charged, in order to attract or repel the charged colored particles in each capsule. There are only 2 charges, but by using different sizes of particles for different colors, you can finagle it so that the color you want ends up on top.

If they could address capsules individually, that would also increase the possible resolution of a black-and-white display; they'd do it if they could. But since the mechanism is about attracting and repelling charged particles, there's probably a pretty hard limit on how much you can increase the resolution before there is interference between neighboring pixels.

> And you could fix the registration with a precisely-cast diffuser layer, convolving each subpixel-cluster.

Would that still look like paper, though?

Although nobody implemented this yet, I'm guessing someone did patent it already ...

Reply on news.ycombinator.com