Basic 4K vs HD vs SD Question

[Miggy Sanchez]

I am going around in circles trying to understand what 4K means in a way that I may be able to explain it to an 8 year old. I don't fully understand it, as I have only used it 4K footage once. The questions may sound dumb to some, but it's the way I learn. The questions and answers I guess are inter-related, but please bear with me as I break this down. Here goes:

1. Is a 4K frame of video the same size as a HD frame of video?

2. Does a 4K frame of video simply have a greater pixel density than a HD frame of video? That is, the 4K frame is not bigger than a HD frame, it just packs more pixels?

3. What generally happens (use of simile/metaphor will do, rather than a overly technical description) when you downscale a 4K frame of video to a HD frame if I am editing in a 4K timeline? Do the 4K pixels simply merge/mash together to create the HD frame?

As you will guess, I am in a state of confusion. Any ideas/thoughts/suggestions etc. much appreciated. I have read many articles, watched many tutes, but the penny has not yet dropped :-)

Cheers.

[Ron Evans]

I think your use of the word frame may be confusing you in trying to compare. It is all pixel density. How it is viewed depends on the display and how that manages the pixel density it is being presented with. Same for editing timelines. Let us use a 4K display to start with and let it display in a 1:1 format. The 4K video will fill the display, the HD video will only fill 1/4 of the display and the SD video will just fill about 1/12 of the display, depending where you are in the world ( PAL or NTSC ). If we now use a HD display then there are too many pixels of 4K for the display , the HD fills the display and the SD video still only fills a small part of the screen. To make all these inputs fill the screens in all cases the display will interpolate the input to fill the screen. On a 4K display anything other than the 4K input will have to create more pixels. For a HD display it can combine groups of pixels of the 4K input ( if it is able to do that ) and create pixels for the SD input.

4K on a HD timeline will be resampled and create a very detailed HD image in my mind the real value of shooting in 4K as this can be used to reframe a shot as well as pan and zoom into the image.

Ron Evans

[Miggy Sanchez]

Ron, thanks for input.

I get a sense from your explanation that the 4K image is bigger than HD? Is that fair to say?

A related question from your post:

Are all pixels created equal in size?

[Ron Evans]

Depends what you mean by bigger. The pixel density is more. If you are counting pixels then of course it is bigger. UHD 3840x2160 compared to 1920x1080. 4 times as many pixels. If you compare image on a 55" TV then the pixels will be 1/4 the size on a 4K TV compared to a HD TV. So the image will have very fine detail. The whole point of the move to 4K TV's. NO, pixels are definitely not all the same size. That is the difference.

Ron Evans

[Seth Bloombaum]

4k is bigger than HD, yes, but,
A pixel doesn't have a particular size until it is rendered on some device.

Pixels are of an arbitrary physical size in the video data - they're just some information, defined below, but, to focus on size for a moment:
A 4k 65" TV will display the same number of pixels over a bigger screen area than a 4k 55" TV. If you got up close with a magnifying glass you'd see that the pixels on the larger TV are bigger.

What is a pixel (px) then?
Pixel is short for picture element, the smallest division of a picture whose color can be defined. In a world without video compression, each pixel gets 24 bits of color information, 8 bits each of Red, Green and Blue data. In the real world things are quite a bit different, because moving lots of data, and video has LOTS of data, over the air or onto an SD card is difficult and expensive. That's one of the reasons why a camera/storage system that will capture Raw video at high color resolution is so costly.

Here's what 8 bits of digital data looks like:
01101001
That much data can determine how much Red is in a particular pixel, out of 256 possible levels. 24 bits gets you enough data a full uncompressed pixel. (In 8-bit color; there are spendy systems that go higher)

We need to talk about pixel dimensions:
4k UHD, the most common video/TV format (there are other cinema formats), is 3840 px wide, and 2160 px tall. If we multiply it out, each frame, each 1/30th of a second sample of the scene, is 8,294,400 pixels in total.

Imagine a grid of dots 3840 wide, by 2160 high. Multiply the width times the height to get the total pixels.

4k = 3840x2160 = 8,294,400px
FullHD = 1920x1080 = 2,073,600px
HD (720p) = 1280x720 = 921,600px

As you can see, 4K has roughly 4 times the resolution as FHD, measured by total pixels.
FHD has roughly twice the resolution of HD-720p.

Those shooting and editing 4k face a question: "I shot this in 4k, what should I distribute it in?"
For TV playback the answer is usually FHD. Most viewers have FHD TVs, and 4k Blu-Ray isn't common, though it is available and prices continue to come down. If you can burn it...

So, in a 4K to FHD conversion 3/4s of the pixels will be thrown away in some fashion. Different editing or compression software use different methods to resample the video.

For Youtube and other online distributions, 4k is possible, but most potential viewers can't see it for reasons of bandwidth or display specs. In practice 720p and 1080p are the leading formats on video sharing sites.

What resolution will your camera shoot? What can you edit? What will your distribution support? What can your viewers' displays show? The answers may all be different.

[Miggy Sanchez]

Ron and Seth, thank you for taking the time to respond in such detail. OK, getting a more complete picture.

To check my comprehension so far, and before the next round of questions (ha ha):

1. A pixel has NO inherent physical size. A pixel could well be described as a digital storage unit for colour and luminance information/data/values etc. Explaining it to an 8 year old, I could use terms such as "pixels of light and colour information", or " pixels that contain light and colour information".

Is that correct?

2. At the play-out stage, a pixel's size is determined by the playback device, such as a 55 inch TV monitor, or the Handycam's flip screen monitor, or the iPhone 7's screen etc. Explaining it to an 8 year old, I could say that "the 55 inch TV monitor tells the pixels what size they should be", or that "the 55 inch TV gives orders to the pixels to present themselves at a certain size".

Is that correct?

3. At the recording stage--of converting light into its digital analogue-- a pixel has NO inherent physical size. Rather, the pixel can contain more or less luminance/colour information depending on the physical size of the sensor and its photosites. Thus, a full frame sensor and 1/3 inch sensor capturing a FHD image are equal in terms of pixel density, but NOT EQUAL in terms of quality.

Is that correct?

Again, much respect for the input.

[Seth Bloombaum]

1. Yes, you have that right.

2. The explanation to the 8 y.o. might be that the TV maps a pixel's information to a particular area of the screen... rapidly!

3. You wrote:
At the recording stage--of converting light into its digital analogue-- a pixel has NO inherent physical size.
Right.

Rather, the pixel can contain more or less luminance/colour information depending on the physical size of the sensor and its photosites.
Here I think you're talking about the angular width/height of the scene captured by a pixel? That is, a pixel will represent a small portion of the scene, depending on focal length and distance to subject. We do need to make a shift to degrees to get specific about how much, eg. if a lens passes a 90-degree wide look to the sensor, and the sensor is 1920px wide, that is about .047 degrees of image width per pixel. So, in addition to the sensor and photosites, what the lens projects onto the sensor varies with focal length.

Thus, a full frame sensor and 1/3 inch sensor capturing a FHD image are equal in terms of pixel density, but NOT EQUAL in terms of quality.
Leaving pixel density out of it for a moment, the projection the lens creates is constant for a particular focal length. But, your 1/3rd inch sensor sees a smaller part of it than a full frame sensor. This is behind what we know as "crop factor". A smaller part of the image is in effect a narrower, more telephoto view.

Back to pixel density, it's more of a camera engineering/design concern. As *video* shooters, we're more concerned with sensor size and pixel dimensions. When Canon figures out a way to go from 20 to 23 megapixels on a sensor the still photogs see that benefit directly, in video we're standardized on pixel dimensions, so we care only if the sensor, image processing, and recording allows a jump from, say, FHD to UHD. Perhaps that's too much of a simplification, as a sensor with relatively more photosites can be downsampled to a video standard, but now we're really into very small benefits that may be hard to see.

Then there's quality - but a full frame vs. a 1/3" sensor, both in FHD, technically can have close to the same quality. For the aesthetics that have developed, we know that a good full frame sensor gives us a kind of linearity of the image, with wall-to-wall color consistency, and, depending on lens aperture, much shallower depth of focus... all these are qualities we've learned to appreciate. Qualities that we associate with film in cinema. I myself am loving the large-sensor look, though I have concerns about shallow depth of focus on *every* shot.

[Miggy Sanchez]

Thanks again Seth! Always exciting to get a detailed reply.

I just wanted to clear up what I meant when writing: "the pixel can contain more or less luminance/colour information depending on the physical size of the sensor and its photosite".

OK. I now clearly understand that when recording/capturing a video image "a pixel has NO inherent physical size. It could well be described as a digital storage unit for colour and luminance information/data/values etc."

Following from that, what I meant was: Are the 2,073,600 pixels on a full frame sensor bigger in physical size than the 2,073,600 pixel sites on a 1/3" sensor?

In short, I am tentatively hypothesising that: given the same lens, at the same focal length, the full frame sensor would generally yield more colour/luminance data than the 1/3" sensor owing to its bigger pixels/photosites.

Or am I way off track here :-) ?

Again, many thanks.

[Ron Evans]

It is actually a little more complicated as sensors do not always have the same number of photo sensors as the pixels that are derived for recording or output at the interfaces. The sensors often have a lot more photosites than is described as pixels. The output pixel density is interpolated from the photo sites in the sensor by the processor in the camera depending on the output requirements. Some have less photosites ( like some of my Sony's ) and a lot have a lot more photo sites than the output pixels. Cameras that take stills as well as video often have 4 or 5 times the sensor sites than is needed for video. So sensor overall size is not always a guide to low light performance or picture quality. A 1/3" sensor with just 2K sites will have bigger photosites than a larger sensor with 40M sites. Its the physical size of the sensor photo sites that matter not the overall size of the sensor.

For the 8 year old it may be easier to say that a 4K TV of any size has 3840x2160 individual lights that can be any brightness or colour. A HD TV has 1920x1080 lights etc. Showing a HD or SD video on a UHD TV then the TV will have to guess what the extra pixels have to be to make up the 3840 x 2160. Called upscaling. Displaying 1:1 like is possible on a computer the HD image will only be 1/4 of the UHD display for example. Using digital zoom on a camera will give an idea of how a poor upscale will look.

Ron Evans

[Dave Blackhurst]

Miggy -

You're mixing apples and oranges... (image) DISPLAY and SENSOR (image acquisition) design are two entirely different balls of cheese....

Sensor SIZE (such as 1/3", 1/2", 1" 4/3, FF, etc) is merely a way of designating the physical measurement corner to corner (sort of like how TV's are marketed - by how far it is from one corner to the opposite corner). Since we're talking 8 year old, we'll leave out the whole thing about how the sensor measurement is not the actual physical measurement... you're confused enough already. And that's before we get into oversampling, line skipping, and on and on and on.... Ron touched on sensor design a bit, and as he says it gets a "little more complicated"...

Sensor engineering and design is not an "8 year old" level discussion - there are so many aspects to how a sensor is constructed, and so many patents that you're not going to grasp it in a few (even well intentioned) internet questions and answers.



As far as DISPLAY, it's "somewhat" simple, there are standards laying out how many."pixels" are required to market a panel as SD,HD, FHD, 4K, etc. That said, there are different approaches to how a given display is constructed (RGB, RGBW, RGBY to mention a couple I've seen), several different display standards, and many approaches to backlighting.

Generally, you can take a tape measure on a TV and it will be a little less than the marketed size, corner to corner, and depending on the marketed resolution, it will have a certain number of pixels vertically, and horizontally, which will determine how much detail can be resolved (more pixels, more information/detail).


Once you get into HOW each one of those pixels is captured, displayed, processed and so on, you're WAY off the 8 year old path.... There are people here who have a far greater grasp on it than I do, but I'm trying to get you away from "excited" and into "realistic" - this simply isn't typical "8 year old" stuff, and unless you're prepared to study, read patents, and settle in for a steep learning curve, you're going to get real lost real fast....

[Miggy Sanchez]

Originally Posted by Ron Evans

A 1/3" sensor with just 2K sites will have bigger photosites than a larger sensor with 40M sites. Its the physical size of the sensor photo sites that matter not the overall size of the sensor...

For the 8 year old it may be easier to say that a 4K TV of any size has 3840x2160 individual lights that can be any brightness or colour. A HD TV has 1920x1080 lights etc...

Ron, thanks again. Post is informative and well-written. Always exciting to move one step forward.

I am getting a clear picture now that there is no neat correspondence between photosites, sensor sizes and pixel output.

Does it follow that skilled cinematographers/videographers consider the physical size of the photosites when considering what cameras to use/purchase/rent for the job at hand?

Originally Posted by Dave Blackhurst

...the sensor measurement is not the actual physical measurement.

Just seeking clarification here: physical measurement of what exactly?

Originally Posted by Dave Blackhurst

As far as DISPLAY, it's "somewhat" simple, there are standards laying out how many "pixels" are required to market a panel as SD,HD, FHD, 4K, etc.

Is "pixels" not the correct term? Is the term marketing/consumer jargon?

And is the standard for a FHD display always 1920x 1080 pixels? Or are there variations?

Originally Posted by Dave Blackhurst

...unless you're prepared to study...

That is the exciting part of coming on here I guess and asking "well-intentioned" questions. Thanks for your contribution.

[Ron Evans]

Simplistically, I stress simplistic, the size of the sensor will govern depth of field and the actual size of the photo sites will govern how the camera performs in low light. The bigger the photo site( actual physical measurement ) the more light it can gather. Again this is simplistic. So an S35 sensor with less photo sites than a 1/4" consumer camera will have much better low light performance and a shallower depth of field. Yes this is a choice users make. There is no direct correspondence between sensor size, photos sites and pixel output. They are all factors in the design of the sensor and the camera electronics. Bigger sensor gives shallower depth of field, physically larger photosites give more low light performance, electronics define the needed outputs. This a description of my NX5U camera with Clearvid sensor which has less photosites than its 1920x1080 output. https://pro.sony.com/bbsc/assetDownloadController/nxcam_hxrnx5u_v2454.pdf?path=Asset%20Hierarchy$Professional$SEL-yf-generic-153708$SEL-yf-generic-153751SEL-asset-186892.pdf&id=StepID$SEL-asset-186892$original&dimension=original

As Dave says displays are governed by names. HD can be 1280x720, the original HD displays, but Full HD is always 1920x1080 pixels. Current 4K TV's are 3840x2160 but Cinema 4K is actually 4096x2160.

Ron Evans

[Seth Bloombaum]

Originally Posted by Ron Evans

...So an S35 sensor with less photo sites than a 1/4" consumer camera will have much better low light performance and a shallower depth of field. Yes this is a choice users make. There is no direct correspondence between sensor size, photos sites and pixel output. They are all factors in the design of the sensor and the camera electronics...

Emphasis added.

Ron sums up the engineering issues rather nicely!

I'd say that *some* users evaluate cameras on the design details and engineering. An equally valid measure, and one that other users see as primary, is the look of the camera's imagery, and performance in the user's typical workflow.

The more engineering-oriented among us see those design details as predictive of the look and performance, but it doesn't always follow.

For example, and at the risk of offending some, I used more Sony cameras than anything else through the early and middle of my career. They were accepted standards in the markets I was in. I never had a complaint, they worked well, had easy-to-understand menus and controls, and results were good *and* predictable.

I've switched camps. LOVE the look of Canon's cameras. I may be more engineering-oriented than some, but, "if the image looks good, it IS good" is a little anti-engineering, or perhaps alternative is a better word than anti.

All this is personal opinion. To be sure, what Ron wrote about the relationship and potential benefits between photosite size, sensor size, down-or-up-sampling is all true, completely factual. But I see it only as a *possible* indicator of performance in my hands for my purposes.

This is born of experience. I remember in the HDV era that Sony was the obvious leader. They'd pioneered DV, and HDV was the next step, built on DV tech. Sony's tech. All the Sony HDV cams were 1440x1280, which through the tricks of non-square pixels became FHD in the edit.

Along came JVC with the HD100. We could argue the merits of this camera, but it was a 1280x720p cam always. CLEARLY the engineering says that's of lower quality than HDV or FHD, right? Except for one little detail - the imagery looked GREAT. And it was all-progressive all the time. One of the most underrated cameras of its time, in my opinion.

Another example - low budget shooters went gaga for the Canon 5D mark II, the camera that sparked the dSLR video revolution. In its day, it was so filmic compared to so many other cameras. Why was that so? A big part of the story-myth-meme was because it was a relatively huge monster of a sensor (in its day, compared to common consumer and broadcast cams). The second wave of dSLR cameras with Super35-ish size sensors was OBVIOUSLY of lower quality - the sensors were smaller, after all! And Panasonic's video dSLRs? Don't talk about them, the sensors are even SMALLER!

Wait just a minute here. How did we lose sight of the look in this evolution of what everyone knows about sensors? Was bigger always better? That's what we thought we knew, and you could darn well see precise measurements in millimeters - what could be clearer than that?

Any sensor of any size doesn't exist by itself as the sole predictor of quality. The in-camera image processing, the settings the user has selected, the in-camera compression for storage, the lensing, the lighting, annnd ::: the storytelling... these are all contributors to quality.

I could go on, but you get the idea. Quality is as-perceived. It's an aesthetic experience of conformance, extension, or rejection of what we've learned about beauty through a lifetime of experiences.

Train your eye along with your intellect.

Rant over.

[Steven Digges]

Wow, what a great thread! It is guys like Ron and Seth that make this forum such a valuable place. Even though I knew what they were explaining before I read this thread I could not have explained it that well in a million words. They are succinct, and spot on. My hat is off to them both, AGAIN.

And, Seth wrapped up the tech talk with what matters the most: What looks good and works for you??? Video is a left brain and right brain endeavor. Going too far in ether direction means you will get cut by the pendulum! As in most things in life, balance matters....

Great thread you two!

Kind Regards,

Steve

[Paul R Johnson]

I've always found a useful example to tell people to consider are the huge LED screens we see at all the pop concerts. If you walk up to them you can see the individual LEDs in their clusters. On some of them, these individual LEDs are bang next door to the next one, but in some, the next LED cluster has a gap between them - so in two screens exactly the same size, one has far more individual elements. Viewed from a distance, the number of pixels, and the 'quality' of them are pretty important. Have a fiddle with your computers screen display options - your screen background picture can easily be changed from low res to high res. See if you can see the real quality differences. If you can't, then it's probably your monitor just can't manage it. You know there are more pixels on the screen, you know the image you have picked is very good quality, but you can't see them. Or maybe you can?