The term 4K is as confused as HD. With 720i, 720p, 1080i, and 1080p all referred to under the umbrella term of HD, the precedent of multiple formats sharing the same marketing name continues with 4K. Currently, there are 4K, Quad Full HD, 4K UHDTV and Academy 4K, which we shall now try to explain.
4K is a rough approximation of the pixel width of the screen, which is nearly 4,000 pixels wide ‘ not tall. The decision to begin using the horizontal measurement instead of the vertical, as with 720p and 1080p, appears to have just ‘happened,’ with no real thought or explanation. The most common 4K measurement is 3840 x 2160 ‘ exactly twice the dimensions of a 1080p screen (1920 x 1080).
3840 x 2160 retains the 16:9 (1.78:1) aspect ratio that is common to most TV screens. However, there are other iterations that chose 16:10 (1.90:1) and measure 4096 x 2160, as well as a slightly wider than 16:9 3996 x 2160. Academy 4K is a resolution that is used in cinemas. It uses a 4:3 aspect ratio, which measures 3656 x 2160, but is really pushing the envelope at almost 10% off of the 4,000 pixels that would logically constitute the 4K moniker.
So the term Ultra HD would probably be a better fit for the 4K family, but 4K is already a powerful marketing buzzword. CES 2014 had a heavy focus on the latest ultra large TVs from the leading manufacturers, as well as ‘affordable’ 4K sets that came in at the $1,000 mark. 4K benefits are found in the largest of screens, where the individual pixels in 1080p begin to become apparent to the naked eye. More pixels on the same size screen means smaller pixels overall with more pixels per inch, and on large screens, usually over 60 inches, 4K can make a visible difference to the picture quality.
However, the much hyped picture quality is dependent on the quality of the source material. Upscaled 1080p content will only look as good as the software that handles the upscaling, and the incumbent manufacturers may have a better grasp on that than the newer and cheaper manufacturers. As such, the ‘more affordable’ 4K sets may have a worse picture quality when displaying 1080p content than the native 1080p panels, which would somewhat defeat the purpose of buying an Ultra HD TV in the first place.
The standard that defines 4K is worth examining in detail, as it will help to clarify the strains that are placed on the codecs and file sizes that are an inevitable part of the discussion of the state of 4K today.
The ITU-R Recommendation BT.2020, or Rec. 2020, was posted in August 2013, and defines UHDTV as 3840 x 2160 (4K) and 7680 x 4320 (8K), a 16:9 aspect ratio, and the use of square pixels. It specifies frame rates of 120, 60, 59.94, 50, 30, 29.97, 25, 24, and 23.976 frames per second. 60 frames per second is broadly considered to be the native format, although the issue instigates fierce debate among videophiles on internet forums.
The ITU-R specified bit depth can be either 10 or 12 bits per sample. Rec. 2020 also uses ‘real colors instead of imaginary colors’ to show the color space on screen without needing dedicated color conversion circuitry. Rec. 2020 can display 75.8% of the CIE 1931 color space, whereas the HDTV Rec. 709 could only display 35.9%. The extra color equates to an increase of 1-bit per sample for 2020 more than the 709 HD spec.
Rec. 2020 allows RGB and YCbCr formats with 4:4:4 color, and 4:2:2 and 4:2:0 which use chroma subsampling to save on file and signal sizes. These numbers are occasionally confusing, but refer to the data that is both recorded and converted by editors and codecs.
The first number in the 3-digit sequence refers to Luma (Y), which represents the brightness in an image, and governs the achromatic (without color) portion ‘ the black and white brightness. Chrominance, or chroma, is the part of the signal that conveys color, and comprises the last two digits (Cb and Cr).
As the human eye is much more sensitive to changes in brightness than color, lots of the data that reproduces color can be discarded or simplified without the difference in quality being noticed. This is why the Luma value is always 4, as it is the most noticeable aspect of the picture quality.
In an example matrix of 4×4 pixels, 4:4:4 records the respective Y Cb and Cr values for all 4 pixels. Each pixel displayed outputs Y, Cb and Cr. 4:2:2 omits the color values for every other pixel in its rows, alternating Y pixels with Y plus Cr and Cb. 4:2:0 blends the two by recording the Y and Cb value in every other pixel and the Y and Cr in alternate pixels. This results in separate rows of Cb and luminance with rows of Cr and luminance. When the screen is viewed from a distance, the rows blend together and the pixels cannot be distinguished ‘ providing a color image that looks as if it had not been compressed.
The 4:2:0 technique, chroma subsampling, is used in JPEG encoding as well as H.264 and H.265 (HEVC). It is one of the ways that encoders try to save on space. Another method involves reducing the frames per second, also known as the refresh rate. The source material may be recorded in 60fps, but most televised content is currently broadcast at 24fps. By halving the number of frames per second, you have significantly reduced the data that needs to be sent regarding pixel changes.
Halving the number of frames per second, say to 30 from 60, will not save exactly half the data that needs to be transmitted. Encoders already use complicated techniques to reduce the data that is sent, by describing changes to the pixels already on screen instead of resending the entire frame’s pixel positions. The encoder can recycle the previous frame’s information if the pixel has not changed, or shift the pixels across the screen if the screen is projecting a slow moving landscape or panning shot. These techniques reduce the amount of data needed to display the frame, but fast action cuts require the most data. Encoding techniques have already managed to efficiently manage the data that needs to be sent.
Many viewers can’t tell the difference between 30fps and 60fps, thanks to the persistence of vision in their retinas, and 24fps is an entrenched standard in Hollywood that is unlikely to change any time soon ‘ although certain films have experimented, such as The Hobbit and Avatar. Some reviews objected to the picture quality of those movies, complaining that they looked more like a soap opera than a movie. The 48fps format reduces the motion blur on screen, compared to 24fps, and some critics felt that The Hobbit looked liked a low-budget soap opera ‘ which are usually recorded at 60i in America.
It is also worth noting that HDMI 1.4, the most common current HD display output, will only display 4K at 24fps, and so it is likely that this will impose something of a limit on the uptake of 4K content, as consumers will have to upgrade their media players or set tops to watch 4K at the higher frame rates for the best quality picture ‘ the main selling-point of 4K in the first place.
So recording or converting the source material at 4:2:0 will save space, as will reducing the frame rate, but the biggest space saving of all comes via compression. The newest generation of codecs, HEVC and VP9, will be tasked with encoding and decoding 4K video as the format gains mainstream popularity. However, some 4K content is available now, and the sizes of the files will give some insight into the challenges that these codecs will face.
TimeScapes is ‘the debut film from award-wining cinematographer and director Tom Lowe. The film features stunning slow-motion and time-lapse cinematography of the landscapes, people, and wildlife of the American Southwest.’ It claims to be the world’s first movie to be sold to the public as a 4K file, and the sizes in which it is sold are quite something.
The DVD download (720 x 480) is an mp4 file encoded by H.264, with a bit rate of 3-7Mbps that measures 1.4GB. The Full HD download (1920 x 1080) is an mp4 file encoded by H.264, with a bit rate of 10-20Mbs that measures 3.4GB. A special 30-inch monitor version (2560 x 1440) is a H.264 mp4 with a bit rate of 15-30 Mbps that is 6.2GBs. The first of three 4K formats is encoded in H.264 at 3840 x 2160, is an mp4 file with a bit rate of 65Mbps and is 21GB. It is delivered on a USB drive, thanks to its huge file size. Now brace yourself.
The 4K ProRes 3840 x 2160, is a .mov file, encoded by ProRes at a 10-bit sample depth, at a bit rate of 370Mbs, and measures 120GB in size. It is sent to buyers on a hard drive. The 4K Cineform (4096 x 2304) is an .avi file, encoded by the ‘near-lossless’ 4:2:2 Cineform codec, at a bit rate of 370Mbs, that clocks in at 330GB. It is also sent on a hard drive, and comes with a warning that you will need a very fast HDD in your system or a RAID array, as well as a very capable CPU, in order to play the video. These last two options cost $200 and $300 respectively.
Although we suspect that the price is largely due to the cost of external hard drives, with a considerable mark up aimed at those consumers who want to finally stress their 4K hardware, it is not the numbers in the price that are problematic ‘ it is the numbers found in the file sizes.
Comcast currently imposes a 300GB soft cap on their broadband customers, where every 50GB after the allowance incurs a $10 charge. AT&T imposes the same type of costs on their customers, but after 150GB per month on DSL lines, and 250GB per month on its U-Verse broadband. Charter’s data caps vary between 100 – 500GB and Time Warner said in 2011 that imposing data caps on their customers was ‘inevitable,’ but never detailed what it imposed. As it stands, a user would most likely breach their soft cap, if they wanted to download the 4K Cineform of TimeScapes.
We have some figures available for the average use of data in broadband households. Comcast said that in 2008 the average home used 2.5GB in a month, but that the figure was between 8-10GB per person per household per month in 2012. The FCC said that the median cable broadband user used 28GB a month in mid 2012, and that there were no data caps at all before 2008 on US broadband networks.
It is important to note that it is not entirely the speed of the connections that is the issue for 4K ‘ it is the size of the content and limits of the data caps. Netflix announced at CES 2014 that it would be streaming its House of Cards in 4K via HEVC compression that required a 15.6Mbs connection. Many US broadband services offer those kinds of speed, but at 15.6Mbps, a 90-minute film will consume roughly 8GB of data.
Netflix’s figures for average speeds of its current SD and HD streams had Google Fiber in first place with 3.69 Mbps, Cablevision Optimum in second with 2.85Mbps, Cox in third with 2.62 Mbps, with Verizon’s FiOS down in sixth with 2.11Mbs, AT&T’s U-Verse in ninth with 1.75, Comcast in 14th with 1.63 and Clearwire in last place (17th) with 0.96Mbps. These numbers are measures of speeds at primetime usage, and are ‘well below the peak performance due to many factors.’
Netflix recommends 3Mbps for DVD quality, 5Mbps for 720p, 7Mbps for 1080p and 12Mbps for 3D. It will no doubt soon be adding the 15.6Mbps required for 4K. Its average speeds show that many users are not viewing its content at HD quality at peak times, but if 4K is the future of TV then its users will want to use the full potential of their TVs and not be hampered by the capacity of their broadband connection, so perhaps it had better start offering downloads.
A GfK report from September 2012 said that the average Netflix user watched 5.1 TV shows, and 3.4 movies a week via the service ‘ roughly 8 hours in total. Some back-of-the-envelope calculations will highlight the problem that looms for US Netflix users who want to use 4K.
If the average Netflix user watched 8 hours of 1080p content in a week, they would consume 2.8GB per hour for a total of 22.4GB per week, based on Netflix’s data usage advice and assuming their streams stayed in HD quality for the whole time. This average user would consume over 90GB of data a month, just through Netflix. That’s near the limits of some current data caps, and possibly over many when you start totaling their entire broadband consumption.
If the HD figures were replaced with the back-of-the-envelope 4K figures we saw above, then the 8 hours of HEVC encoded 4K would consume around 42GB of data, for a monthly usage of about 160GB. This is certainly in data cap territory, and that is for the average Netflix user. A heavy user could quite easily watch double that amount of content in a month for 320GB, and it must be remembered that families can easily increase their rates of consumption if family members watch Netflix independently of one another.
On this basis, it looks likely that ISPs who impose data caps will have to raise the cap levels or consider removing them, as their customers consume more and more data. 4K, and soon enough 8K, will stress the networks harder than Netflix’s figures suggest ‘ as the average peak time stream speeds are much slower than the speed necessary for a 1080p or a 4K stream. The network infrastructure, particularly the last-mile connections will be a key area of investment for ISPs as customers move towards streaming more HD video content into their homes.
Net Neutrality is an issue that looks to be at the forefront of developments, with sponsored data and certain data cap exceptions making the news recently. Comcast garnered controversy and outcry when it decided not to count the data usage of its Streampix service against its customers’ data caps, with critics claiming that this was a slippery slope towards an age where ISPs grant privileges to websites who have paid up, or discriminate against those websites that have not signed an agreement with them. Comcast’s behavior does appear to be a tacit admission that data caps do in fact influence their customers’ behavior ‘ and refrain from using services like Netflix because of the threat of soft cap fines.
We feel that in the wake of the FCC’s recent legal loss, where it was judged that broadband was not a ‘common-carrier service’ and therefore not under the FCC’s remit, something of a standoff is in place in the US. The carriers don’t want to provoke the government to introduce net neutrality legislation, and yet they aren’t currently breaking the law by effectively sponsoring their own media content on their networks above rival content, as well as putting it on other networks. It is clear that currently the balance of power is leaning toward ISPs and their media divisions.
In the shadowy world of Internet Tier 1 network transit agreements and the deals struck between content delivery networks (CDNs) and ISPs, there may be a lot of room for content providers to push unlimited or generous amounts of 4K content to the home. If consumers don’t have ready and easy access to 4K, either through OTT, VoD or broadcast, then 4K will not reach the heady-heights that CES 2014 would have you believe.
Another problem that lies in store for 4K, is the rival 8K that is already in development. Japan’s NHK is already focusing on 8K (7680 x 4320), and a good portion of the London Olympics was recorded in 8K. According to Digital Trends, 8K TVs might be available in ‘just four years,’ which begs the question of whether 4K will simply be a stepping stone on the way to 8K, and end up like 720p that was quickly overtaken by 1080p.
4K is a promising technology for consumers who want to buy very large TVs. It has a strong future in digital signage, where units like BrightSign’s 4K1240 will allow 4K to reach large audiences as part of visually striking advertising in public spaces. For many consumers, these eye-catching adverts will be the only place they see 4K for the next few years. As we have seen, the pressures on LCD panel prices, the quality of consumer broadband connections and the issue of current data caps may hamstring 4K in its first few years ‘ when 8K may come snapping at its heels if the current multi-year time-to-market predictions for 4K are accurate.