There are some sentences in this that are technically vague enough to pass, but I don't think are strictly speaking correct, and I believe will likely lead to a mistaken understanding:
> modern displays don't paint the image line-by-line (...) They light up each pixel simultaneously, refreshing the entire display at once.
The entire screen area is lit all the time now, yes, but refresh still typically happens line by line, top to bottom [0], left to right [0], for both LCDs and OLEDs. It's a scanning refresh, not a global refresh (sadly).
You can experimentally confirm this using a typical smartphone. Assuming a 60 Hz screen refresh, recording in slow motion will give you enough extra frames that the smartphone camera also likely operating in a scanning fashion (rolling shutter) won't impact the experiment. On the recording, you should see your screen refreshing in the aforementioned fashion.
[0] actual refresh direction depends on the display, this is for a typical desktop monitor
I was glad they at least mentioned how IPS (PLS) and VA differ from older TN.
But you're right both LCD and OLED refresh a stored voltage on the cell (or caps) on a roughly line by line (OLED can easily be 5 clocks on the GIP to cancel internal transistor offset voltages).
I was mostly annoyed that they didn't mention the circular polarizer on OLEDs. Although there is discussion of going to color filters with Quantum Dot OLED, the circular polarizer is what makes the blacks so black on mobile OLED devices.
Also, didn't really mention pentile RGGB sub-pixel pattern which is dominant in mobile OLED (which is more than 50% of devices). Now they're moving to "tandem" stacked OLED for higher brightness and lower current density, but no latteral sub-pixel pattern.
Maybe slightly off topic but I was surprised to discover that my glasses with photoreactive lenses (SpecSavers ‘Reactions’) are actually circularly polarised but only when they go dark. I originally thought they didn’t work because they didn’t interact with another pair of polarised sunglasses (no changing brightness as I rotated one lens in front of the other) but later noticed that my phone screen with its circularly polarised IPS screen was almost black in bright sunlight… until I took my glasses off.
There were a few things I personally found lacking as well, albeit they're fairly minor.
Regarding CRTs, at the vector CRTs section, they mention "they were mostly monochrome and so the phosphor dots could be tightly packed" - this is not true either I believe, monochrome CRTs had a uniform phosphor coat on the inside, no subpixel patches. I'd have also liked if they delved a bit into the decay times of the various phosphor chemistries used for color CRTs, and how they compare to LCDs and OLEDs. It's an entertaining comparison, grounds motion performance related discussions really well.
Regarding LCDs, I missed the mention of multi-layer LCDs, especially since they bring up tandem OLEDs.
Regarding OLEDs, now that you mention, the subpixel layouts were left unaddressed.
Regarding quantum dots, I missed both the mention of QDEL as a somewhat promising future contender, and the mentioning of the drawback of their typical implementation. External light also provides them with energy to activate, which I believe is at least partially the cause behind the relatively poor black levels of QD-OLEDs in environments with significant ambient light (+ something about it not being possible to put a polarizer in front of them?)
I was also generally expecting a more in-depth look by the title, would have loved to learn about the driving electronics, maybe learn about why OLEDs aren't ran anywhere near as fast as their full potential (I'd assume throughput limitations), etc. Overall, it basically only covers as much as my own enthusiast but not in-the-area self gathered over the years too.
Indeed, one feature of active-matrix (and even passive-matrix) displays is that it needs only m + n signal lines to address a pixel in an m + n display. To change the color of a pixel, a signal goes out over the lines corresponding to the row and column of the addressed pixel, selecting it; and then another signal is transmitted over another line to actually change the value of that pixel. In this scheme, it would be impossible to address all pixels simultaneously. Nor would you actually want to, as this would require millions of control lines to drive the display!
Assuming a 60Hz refresh rate, does it take ~16ms (± the vblank inteval) for a complete cycle from top-left to bottom-right? Or does the scan happen faster than that (with something else being the limiting factor on overall refresh rate)?
Yes, the refresh cycle takes ~16.6 ms. There's another "point" chasing behind the refresh "point", that will be where the panel's response time has finished catching up with the refresh. In between these points, the pixels are slowly morphing from one color to the next. On LCDs, the area between these two points is quite sizeable, definitely more than a few lines, sometimes even hundreds of lines. On a 1080p 60 Hz display, just 1 ms of response time corresponds to 64.8 lines (6% of the screen) being constantly in flux, for example.
The difference between LCDs and CRTs in this regard then, is that on a CRT you only ever got light during that chase section. The initial state is full darkness, and the final state is full darkness too. It's a pulse.
I don't know how common it is now, but a lot of high-resolution LCDs with dual LVDS interfaces were essentially two separate panels, with one lane feeding the top half and the other the bottom half.
Not only was the initial diagram all/explaining, but the "pop"-"pip" on zoom-unzoom of the image was just as nice as playing with a sheet of bubble wrap.
Wow, and that ruler on the right side, even with the sound.
So crisply done. I wish if Dan writes textbooks for all science & math books for High School students. World would be a better place for those who struggle to understand academics.
Adding my congrats as well. The combination of well-written explanations for the semi-technical layperson combined with clear, intuitive graphics is a powerful instruction platform.
This appears to be a lovely project. I wish the author all possible luck and success. I haven't joined a mailing list in a very long time, but I sure did in this case.
CRT displays are one of those analog technologies that are arguably much cooler than their digital successors. Think – a literan raygun, a particle accelerator, inside your monitor, creating the image you're looking at.
True – but on the other hand, it was "only" a few million elements, and very large ones, compared to, say, the DRAM chips of the time. Monitors certainly make the engineering feat more tangible, though!
It’s certainly one the bits of tech that would be an insane pitch if invented today: “Yes, please stick your face in front of this particle accelerator. I assure you, it’s completely safe. You may experience some visual side effects but that’s completely intended.”
It's much cooler if you haven't used it. That mass of a 19in was crazy. Then you had all of those that had the high pitched noise. And if you were unlucky you had to degauss them. Cool concept, but in practice the digital successors are better.
"I was there" too and somehow this is not my reaction at all.
Everyone knows the obvious reasons we don't use crts any more.
It's true but it's a most uninteresting observation that only cares about practical aspects. Practical aspects matter, none of my own desks has a crt, but they do not define life itself.
Those facts do not at all invalidate the point about the desirable aspects which have been lost, or the fact that the merely interesting and remarkable aspects are interesting and remarkable.
The desirable and/or interesting and remarkable features of a crt are still cool, impressive, fun, desirable, even though we all voluntarily choose to use something else basically everywhere we want a screen because of the practical reasons that just happen to overwhelm.
Trust me, I remember. Cool does not equal convenient. Once a friend and I dragged three extra CRTs to a demo party just so we could put them side by side and write a program that displayed random scrolling messages. Controllable via an IRC bot, even! Today you could do that with a single ultrawide…
CRTs are still slightly magical to me. The image doesn't really exist. It's an illusion. If your eyes operated at electronic speeds, you would see a single incredibly bright dot-point drawing the raster pattern over and over. This YouTube video by "The Slow Mo Guys" shows this in action: https://youtu.be/3BJU2drrtCM?t=190
That slo-mo video is somewhat misleading, though. The phosphor glows for a good while, so there is a reasonable chunk of the image that's visible at any given time.
The problem in that video is that the exact location the beam is hitting is momentarily very bright, so they calibrated the exposure to that and everything else looks really dark.
The phosphor still drops off very quickly [0][1][2], roughly within a millisecond. That’s why you would need a 1000 Hz LCD/OLED screen with really high brightness (and strobing logic) to approximate CRT motion clarity. On a traditional NTSC/PAL CRT, 1 ms is just under 16 lines, but the latest line is already much brighter than the rest. The slow-motion recording showing roughly one line at a time therefore seems accurate.
When I learned how TV worked at the beginning of television history, I found it super cool that the camera and all the TVs across the country had their scanning beams synchronized. That camera was driving your TV, almost literally.
I only recently found out that the tech to save images wasn’t invented, so they couldn’t display a revolving logo between shows. So… so the BBC had a permanent real-life logo with a permanent camera in front of it.
So yes, any image was extremely ephemeral at the time.
PS: Apparently it’s called a Noddy, it’s a video camera controlled by a servomotor to pan and tilt (or 'nod', hence the name Noddy): https://en.wikipedia.org/wiki/Noddy_(camera)
There is some persistence in the pixels/phosphor though so it's not a complete illusion. But yes, your eyes are integrating over the frame. There is also interlacing...
I read something interesting recent but I'm not sure if it's true or not. That as you age your integration frame rate decreases.
To me the magical part about CRTs is color. I don't quite understand how the shadow mask works. Like, yeah, there are three guns, one for each color channel, and the openings in the mask match their layout, and somehow the beam coming out of each gun can only ever hit its corresponding phosphor dots. Even after being deflected. But... how? Also, wouldn't the deflection coils affect each of the three beams slightly differently?
It's parallax, basically. The pigment dots and mask holes are positioned such that when you look from the perspective of the "red" electron gun (*), you only see red pigment dots. Move a couple cm to the "blue" gun and the parallax shift now makes you to see only blue pigment dots instead. Or from the other direction, no matter which "red" dot you stand at, you only see the "red" gun through "your" hole.
The exact sizes, shapes, and positions of the pigment dot triples (and/or the mask holes) are presumably chosen so that this holds even away from the main axis. Also, the shape of the deflecting field is probably tuned to keep the rays as well-focused as possible. Similarly to how photographic lenses are carefully designed to minimize aberrations and softness even far from the optical axis.
(*) Simplifying a bit by assuming that the beam gets deflected immediately as it leaves the gun, which is of course inaccurate.
Each hole in the shadow mask acts as a pinhole camera, giving an inverted image (in electrons) of the three guns. All three beams get bent nearly the same amount, but yes there is some distortion which is traditionally corrected for by a set of convergence coils and corresponding circuit with knobs for static and dynamic convergence [0]. A pain to adjust, BTW.
For me it was the opposite. Learning how a monochrome CRT requires no mask sort of destroyed my world view of what a display had to have. pixels(even the quasi pixels as found in a color CRT mask) were not actually required or present.
As a result monochrome terminal text has this surprising sharpness to it.(surprising if you are used to color displays). But the real visual treat are the long persistence phosphor radar scopes.
I have to take issue with the usage of the terms "pixel" and "subpixel" with regards to CRT. CRTs do not display discrete pixels. They display discrete scanlines, each one made up of a smoothly varying voltage across the line (and thus resolution is a function of both the DAC in the display device in the case of systems that generate a digital signal and then convert it to analog for display, and the hardware inside the CRT monitor). Also, there is no mapping between any "pixels" represented within that varying voltage and the separate color phosphor dots.
Even "digital RGB" isn't digital in terms of the CRT. It's only "digital" because each color channel has a nominal on and off voltage, with no in-between (outside of the separate intensity pin). However, the electron gun still has a rise and fall time that is not instant.
Displays didn't truly become digital for the masses until the LCD era, with DVI and HDMI signals. Even analog HD CRTs could accept these digital signals and display them.
I don't think this is an entirely fair characterization either. Note that everything I lay out here is just based on accumulated information gathered over the years due to vague interest, I haven't worked on or with CRTs (did use them though).
Monochromatic CRTs were well and truly resolution agnostic, there were legitimately no pixels or subpixels or anything similar to speak of. That said, the driving signal still had to be modulated to produce an image, and so it's not magic either. You can conceivably represent [0] all the available information in them using just 720 samples per line, which is exactly why DVDs had that as their horizontal resolution (720 pixels).
This story changes a bit though with color CRTs, where you did have discrete sets of patches of different phosphor chemistries called triads. There was absolutely a fixed number of them on a glass, so you could conceivably consider that as the native resolution for that given display, with each triad being a pixel, and each patch being a subpixel. The distance between these was the aperture pitch, much like how you have a pixel pitch on a typical flatpanel display.
The kicker then is that as you say, there's no strict addressing. From what I understand there were multiple electron guns scanning across the screen simultaneously, only being able to hit the specific color they were assigned, but the patch they were hitting wasn't addressed, they just scanned across the screen like the single electron gun did in monochromatic CRTs. You'd then get resolution invariance by just the natural emission spread providing you with oversampling / undersampling without any kind of digital computational effort. It's not really true resolution independence like with the monochrome ones, I'd say. I even recall articles where they were testing freshly released CRT monitors, and discussing how sharp the beam was, resulting in what kind of resolution adherence.
[0] an earlier version of this comment said "extract from" here; for various reasons you might already know, that's a different thing, and would not actually be true.
It still basically should be, so long as well-designed sites give you the "small screen"/mobile layout.
Even apart from that, a lot of laptops still have 1280x800 as the default resolution, and that's only double the width of 640x480. Honestly, I'd actually be more worried about OS and browser chrome eating up the space than websites themselves being unusable.
I'm much more interested in the hardware driver. This thing gets digital encoded input, has to decode it, and then multiplex it to 8 million pixels. 60 times a second. Being able to hit at least 4 million different levels (talking about 4K 60fps 12 bit color).
The input is roughly serial, so it takes a massive serial to parallel conversion.
But the output of the scaler chip is still serial, just now it's guaranteed to have the same dimensions as the panel itself, and includes whatever OSD overlay might be active, and any gamma/contrast/whatever adjustments. I believe in some cases, this is also where dithering takes place, to take a 6-bit panel and try to give it 7-bit (or, yikes, 8-bit) color depth by PWMing pixels that have intermediate values.
Look at the connector pinout of the panel itself. There's only 50 pins or so, and a lot of them are grounds. Whether the scaler-to-panel format is eDP, or LVDS (FPD-Link), or V-by-One, it's all still differential serial lanes at that point.
Around the perimeter of the panel, then, are the actual TCON and row/column driver chips, bonded right to the ITO traces on the glass, flip-chip-on-glass style. These have an outrageous number of pins, and directly connect to the gate (row) and source (column) traces. It's here that the serial becomes parallel, and the next stop is the transistors themselves (hence the MOSFET signal terminology of gate and source) in the individual pixels.
Older displays would have basically a bunch of serial-to-parallel register chips with each one's SO connected to the next one's SI. I ran across a fasincating video of replacing a bad chip in such a display, which happens to be gas plasma so the voltages involved are also pretty high too:
Yeah, with an electron beam it's clear that it's a continuous signal and a single beam is being controlled, but how do you drive millions of individual digital pixels all at once, how does the signal get routed correctly to each one if them?
It was 50/50 for me as well but the screen source code is fairly readable and if I remember right eerily over-commented for Unix code! The function names actually make sense.
I can appreciate these articles as they are but I personally don’t like them. They are junk food level of infotainment to me. Something I’d find on a Wikipedia summary section that covers general points.
A CRT - to name one - is a device whose actual understanding will challenge people in profound ways. To ask “how does a screen even work?” and to begin to answer this question will require a bit more than a summary form of “thing goes from point A to point B”. The history of this discovery is a stack of books and in and of itself is fascinating - the experiments and expectations and failures and theories as to why and how. I suppose I just expect more of the site. The illustrations are nice. Oh and my moniker is just a coincidence.
Readit News
> modern displays don't paint the image line-by-line (...) They light up each pixel simultaneously, refreshing the entire display at once.
The entire screen area is lit all the time now, yes, but refresh still typically happens line by line, top to bottom [0], left to right [0], for both LCDs and OLEDs. It's a scanning refresh, not a global refresh (sadly).
You can experimentally confirm this using a typical smartphone. Assuming a 60 Hz screen refresh, recording in slow motion will give you enough extra frames that the smartphone camera also likely operating in a scanning fashion (rolling shutter) won't impact the experiment. On the recording, you should see your screen refreshing in the aforementioned fashion.
[0] actual refresh direction depends on the display, this is for a typical desktop monitor
But you're right both LCD and OLED refresh a stored voltage on the cell (or caps) on a roughly line by line (OLED can easily be 5 clocks on the GIP to cancel internal transistor offset voltages).
I was mostly annoyed that they didn't mention the circular polarizer on OLEDs. Although there is discussion of going to color filters with Quantum Dot OLED, the circular polarizer is what makes the blacks so black on mobile OLED devices.
Also, didn't really mention pentile RGGB sub-pixel pattern which is dominant in mobile OLED (which is more than 50% of devices). Now they're moving to "tandem" stacked OLED for higher brightness and lower current density, but no latteral sub-pixel pattern.
Regarding CRTs, at the vector CRTs section, they mention "they were mostly monochrome and so the phosphor dots could be tightly packed" - this is not true either I believe, monochrome CRTs had a uniform phosphor coat on the inside, no subpixel patches. I'd have also liked if they delved a bit into the decay times of the various phosphor chemistries used for color CRTs, and how they compare to LCDs and OLEDs. It's an entertaining comparison, grounds motion performance related discussions really well.
Regarding LCDs, I missed the mention of multi-layer LCDs, especially since they bring up tandem OLEDs.
Regarding OLEDs, now that you mention, the subpixel layouts were left unaddressed.
Regarding quantum dots, I missed both the mention of QDEL as a somewhat promising future contender, and the mentioning of the drawback of their typical implementation. External light also provides them with energy to activate, which I believe is at least partially the cause behind the relatively poor black levels of QD-OLEDs in environments with significant ambient light (+ something about it not being possible to put a polarizer in front of them?)
I was also generally expecting a more in-depth look by the title, would have loved to learn about the driving electronics, maybe learn about why OLEDs aren't ran anywhere near as fast as their full potential (I'd assume throughput limitations), etc. Overall, it basically only covers as much as my own enthusiast but not in-the-area self gathered over the years too.
The difference between LCDs and CRTs in this regard then, is that on a CRT you only ever got light during that chase section. The initial state is full darkness, and the final state is full darkness too. It's a pulse.
Wow, and that ruler on the right side, even with the sound.
One of the nicest pages I have been on.
And the landing page... https://www.makingsoftware.com/
It just keeps on giving.
https://ciechanow.ski/archives/
Each individual pixel is driven by a transistor and capacitor that actively maintain the pixel state? Insane manufacturing magic.
Dead pixels used to be a big problem with LCD displays. Haven’t thought about that in at least twenty years.
Everyone knows the obvious reasons we don't use crts any more.
It's true but it's a most uninteresting observation that only cares about practical aspects. Practical aspects matter, none of my own desks has a crt, but they do not define life itself.
Those facts do not at all invalidate the point about the desirable aspects which have been lost, or the fact that the merely interesting and remarkable aspects are interesting and remarkable.
The desirable and/or interesting and remarkable features of a crt are still cool, impressive, fun, desirable, even though we all voluntarily choose to use something else basically everywhere we want a screen because of the practical reasons that just happen to overwhelm.
The problem in that video is that the exact location the beam is hitting is momentarily very bright, so they calibrated the exposure to that and everything else looks really dark.
[0] https://blurbusters.com/wp-content/uploads/2018/01/crt-phosp...
[1] https://www.researchgate.net/figure/Phosphor-persistence-of-...
[2] https://www.researchgate.net/figure/Stimulus-succession-on-C...
So yes, any image was extremely ephemeral at the time.
PS: Apparently it’s called a Noddy, it’s a video camera controlled by a servomotor to pan and tilt (or 'nod', hence the name Noddy): https://en.wikipedia.org/wiki/Noddy_(camera)
I read something interesting recent but I'm not sure if it's true or not. That as you age your integration frame rate decreases.
The exact sizes, shapes, and positions of the pigment dot triples (and/or the mask holes) are presumably chosen so that this holds even away from the main axis. Also, the shape of the deflecting field is probably tuned to keep the rays as well-focused as possible. Similarly to how photographic lenses are carefully designed to minimize aberrations and softness even far from the optical axis.
(*) Simplifying a bit by assuming that the beam gets deflected immediately as it leaves the gun, which is of course inaccurate.
[0] https://antiqueradio.org/art/RCACTC-11ConvergBoardNewRC.jpg
As a result monochrome terminal text has this surprising sharpness to it.(surprising if you are used to color displays). But the real visual treat are the long persistence phosphor radar scopes.
In a sense, all vision is.
Even "digital RGB" isn't digital in terms of the CRT. It's only "digital" because each color channel has a nominal on and off voltage, with no in-between (outside of the separate intensity pin). However, the electron gun still has a rise and fall time that is not instant.
Displays didn't truly become digital for the masses until the LCD era, with DVI and HDMI signals. Even analog HD CRTs could accept these digital signals and display them.
Monochromatic CRTs were well and truly resolution agnostic, there were legitimately no pixels or subpixels or anything similar to speak of. That said, the driving signal still had to be modulated to produce an image, and so it's not magic either. You can conceivably represent [0] all the available information in them using just 720 samples per line, which is exactly why DVDs had that as their horizontal resolution (720 pixels).
This story changes a bit though with color CRTs, where you did have discrete sets of patches of different phosphor chemistries called triads. There was absolutely a fixed number of them on a glass, so you could conceivably consider that as the native resolution for that given display, with each triad being a pixel, and each patch being a subpixel. The distance between these was the aperture pitch, much like how you have a pixel pitch on a typical flatpanel display.
The kicker then is that as you say, there's no strict addressing. From what I understand there were multiple electron guns scanning across the screen simultaneously, only being able to hit the specific color they were assigned, but the patch they were hitting wasn't addressed, they just scanned across the screen like the single electron gun did in monochromatic CRTs. You'd then get resolution invariance by just the natural emission spread providing you with oversampling / undersampling without any kind of digital computational effort. It's not really true resolution independence like with the monochrome ones, I'd say. I even recall articles where they were testing freshly released CRT monitors, and discussing how sharp the beam was, resulting in what kind of resolution adherence.
[0] an earlier version of this comment said "extract from" here; for various reasons you might already know, that's a different thing, and would not actually be true.
I was thrilled when my computer let me choose a resolution of 848x480, and it worked perfectly.
Back in those days, the web was usable at that resolution.
Even apart from that, a lot of laptops still have 1280x800 as the default resolution, and that's only double the width of 640x480. Honestly, I'd actually be more worried about OS and browser chrome eating up the space than websites themselves being unusable.
The input is roughly serial, so it takes a massive serial to parallel conversion.
Look at the connector pinout of the panel itself. There's only 50 pins or so, and a lot of them are grounds. Whether the scaler-to-panel format is eDP, or LVDS (FPD-Link), or V-by-One, it's all still differential serial lanes at that point.
Around the perimeter of the panel, then, are the actual TCON and row/column driver chips, bonded right to the ITO traces on the glass, flip-chip-on-glass style. These have an outrageous number of pins, and directly connect to the gate (row) and source (column) traces. It's here that the serial becomes parallel, and the next stop is the transistors themselves (hence the MOSFET signal terminology of gate and source) in the individual pixels.
Older displays would have basically a bunch of serial-to-parallel register chips with each one's SO connected to the next one's SI. I ran across a fasincating video of replacing a bad chip in such a display, which happens to be gas plasma so the voltages involved are also pretty high too:
https://www.youtube.com/watch?v=6W3H5wOy5sY
DVI (and thus older HDMI) being essentially "VGA that skipped Digital to Analog conversion" you're riding the beam, including porches.
A CRT - to name one - is a device whose actual understanding will challenge people in profound ways. To ask “how does a screen even work?” and to begin to answer this question will require a bit more than a summary form of “thing goes from point A to point B”. The history of this discovery is a stack of books and in and of itself is fascinating - the experiments and expectations and failures and theories as to why and how. I suppose I just expect more of the site. The illustrations are nice. Oh and my moniker is just a coincidence.
Dead Comment