Logically it would make sense to capture it directly at 640x480 (or better yet, 648x486). There's no actual horizontal pixel count specified in the NTSC baseband video specification, because analog video is transmitted as a sequence of 486 continuous horizontal lines. The horizontal resolution of the video is limited only by the FCC TV broadcast specification, which imposes a 6MHz bandwidth for the entire channel (both audio and video). Because an actual physical SD TV screen has an actual physical aspect ratio (as measured by a ruler) of 4:3, it makes most sense to have video capture cards capture directly to 640x480 pixels. Actually it would make even more sense to capture to a resolution of 648x486, to keep the 4:3 aspect ratio while also keeping the original number of vertical lines (there are actually 3 lines above and 3 lines below the normal 480 found in video captures, and the top line and bottom line are actually only half-lines in that the top line starts about half way into the image and the bottom line ends about halfway into the image). The horizontal resolution of a displayed image on an analog TV with CRT is limited only by the resolution connected to the composite video input of the TV (how many pixels wide is the CCD chip of the video camera), so you could theoretically have a video source that is 1000x486 and it should display properly on an analog SD TV with just exceptionally high horizontal resolution (objects looking very sharp in the horizontal direction, such as very sharp vertical edges on objects).
If you are talking about the limits imposed by the NTSC and FCC for broadcast TV https://antiqueradio.org/art/NTSC%20Signal%20Specifications.pdf then then you need to limit the video baseband bandwidth to 4.5MHz. A 4.5MHz bandwidth could be captured by an analog to digital converter sampling at a rate of 9MHz. The line rate of video in color NTSC video is approximately 15734.26 lines per second. With a sample rate of exactly 9MHz, you have approximately 572 pixels per line (including both image content and H-blanking period). On each line, 83.5% of the line is image (the remainder being the H-blanking period). This gives you an image content line width of 477.62 pixels. As pixels are discrete, the final image will need to be 478 pixels wide. So the proper resolution for capturing broadcast TV, would be 478x486 (which is almost a perfectly square image, and would need to be stretched horizontally on playback to fit the 4:3 aspect ratio).
If you are talking about digitized VHS tape (which has a reduced bandwidth, even when compared to broadcast TV), then the screen width is only 320 pixels, so it's actually quite a bit taller than it is wide, so would need to be highly stretched in the horizontal direction to fit the 4:3 aspect ratio. My source of info on the width of a VHS tape frame is from https://www.mediacollege.com/video/format/vhs/ and it gave it in TV Lines resolution, a resolution measurement which is described here https://en.wikipedia.org/wiki/Television_lines on Wikipedia. So basically a VHS tape has a horizontal resolution of 240 TVL, which means that in a portion of the screen that's as wide as the screen is tall, how many individual alternating 240 alternating black and white vertical lines (120 of each) can be seen. This is equivalent (on a display using pixels) to 240 pixels wide in a portion of the screen as physically wide (as measured by a ruler) as it is tall. Since the aspect ratio of the screen is 4:3, that means you have 240/3*4=320, for a width of 320 pixels across the full screen. Of course that doesn't take into account the overscan portion of the image. In the broadcast standard I was talking about, 478x486 pixels is for the full image, and that includes pixels in the overscan region. There's more than 320 pixels wide in a VHS tape stored image, as 320 is only the visible screen portion. According to https://build.com.au/what-overscan the overscan on a TV can cut off up to 5% of width and height of an image. If the on screen portion of the image is 320 pixels wide, then the actual full width of the image could be as large as approximately 336.84 pixels. So to properly digitize a video on VHS tape, you would need to capture it with a width of 337 pixels, for a full image size of 337x486. Again, this would need to be drastically stretched horizontally on playback to fit the correct aspect ratio of 4:3.
As you have noticed, none of these assessments I've done on various digitized video sources, have shown that there should ever be a width of 720, or that the correct aspect ratio of the video should be 3:2 (which is the aspect ratio of 720x480). I also don't know why 480 is the standard height for capturing NTSC video, as 480 is not the correct number of lines for a video frame, and the difference of 6 lines high is only 1.23% of the video height (the height being 486 pixels). 1.23% is less than any amount of overscan that might exist (which is 2% to 5%), so I don't know why the standard is to crop off the upper 3 and lower 3 lines of the image during video capture, as overscan would crop off a lot more than that.
Can someone here explain how the resolution of 720x480 became the standard for digitizing NTSC video?
+ Reply to Thread
Results 1 to 30 of 42
Last edited by Videogamer555; 11th May 2022 at 15:10.
not sure what is your goal, also not sure if you intentionally (or not) ignore some signal processing aspects and/or technology background associated with video signal processing. Anyway i strongly recommend to use worldwide standard reference https://www.itu.int/rec/R-REC-BT.601/ .
Btw this perhaps could be good topic for discussion somewhere before year 2000 but now overall world is focused on UHD, 8k, and i think most interesting for future TV standards https://www.bbc.co.uk/rd/projects/high-frame-rate-tv - very good lecture for today https://www.bbc.co.uk/rd/publications/whitepaper282
Forgot to add link to paper that answers your questions. https://tech.ebu.ch/docs/techreview/trev_304-rec601_wood.pdf
Last edited by pandy; 11th May 2022 at 16:40.
Last edited by Videogamer555; 11th May 2022 at 17:21.
have been decided upon based on broadcast standards of that time and on interoperation between them. These numbers are as good as it gets. Number of scanlines is fixed across the board, number of unique samples in a scanline is fewer than 720, in case of VHS or Hi8 much fewer.
You cannot use 640x480 instead of 720x480 because:*
- you video may have proportions other than 4:3, for example 16:9 or 14:9
- 720x480 usually includes overscan, so it does not represent a 4:3 or 16:9 frame anyway, it is slightly wider
Well-designed user-friendly software usually has a wizard-style UI that guides you through the process: "Is it a standard or widescreen video? -> Move the wireframe on this preview to select the part of the video you want to capture -> Do you want to make a DVD? -> Do you want to upload to YouTube? -> Do you want to retain the original image rate? Etc." Bad software takes the most common and least objectionable path. And then there is software that puts a user in charge, this the kind of software that members of this forum use.
YouTube accepts video with non-square pixels if PAR is correctly specified in the header. Too bad that many people either forget to do it, or don't know about non-square pixels, or they use containers that do not clearly define how to set PAR, so they end up with videos that are stretched. But since you are here, I am sure you are not one of these people
*You can, but then you may need to re-sample again.
I think I wrote too much as usual. To address the original quote: no, you will not lose anything by capturing into 720x480.
720x480 is broadcast standard - you can't capture less than BT.601 standard offer per se - 486 lines in NTSC are present only in pre 601 tmes - then you may loose some information if you trying to capture it with 601.
Oversampling has nothing or not much with synchronization - using ADC to acquire whole video seem to be waste (as 8 bit ADC usually offer sub 8 bit accuracy - try familiarize with for example ENOB .
Oversampling allow to deal with some signal processing limitations - there is many befits for using oversampling - extreme case are so called 1 bit converters (99% or more for nowadays audio ADC/DAC) - with massive oversampling such as 64x times or more with help of 1 bit you can reach 16 bit resolution and sometimes also accuracy. Nevertheless oversampling is your friend.
If capture sample rate and playout sample rate are equal then you have 1:1 mapping there is no distortions in terms of aspect ratio.
Also aspect ratio before directly addressable pixels type of displays (i.e. LCD, PDP, OLED etc) i.e. in CRT times was less stable - there is many factors affecting display aspect ratio on CRT's and usually error around 10% was common - also DAR in CRT times was heavily dependable on scene brightness.
CRT was never perfect display from videophile purist perspective - many factors involved - you can't ignore limitations of the technology in those times - cameras using tubes https://en.wikipedia.org/wiki/Video_camera_tube - and shared some limitations commonly with CRT's.
601 allowed to content exchange between Europe and USA - it was beneficial for both sides - and this is key for 601. It tried to address problems encountered earlier and thanks to 601 we have modern digital video technology.
Never mind, but it seems we are going to repeat what has recently been addressed and argued here, and discussed in many similar threads over the past decade(s).
Everyone is free to re-invent the wheel and develop his private digitization standards wrt sampling rates and "resolutions" though, one each tailored for NTSC, PAL, SECAM, Video8, Hi8, VHS, SVHS, laserdisc ... and so on, and eventually try to find the support of the industry .
The Rec.601 was a breakthrough in finding a "common denominator" with the 13.5MHz sampling rate which has subsequently been accepted for analog SD video digitization, even though it is some overkill for certain low resolution sources with their "approximate" resolutions or spectral bandwidths.
Last edited by Sharc; 12th May 2022 at 05:11.
Everyone is free to re-invent the wheel and develop his private digitization standards
This topic is poor trolling IMHO. Such topics should be closed as soon as possible without any discussion.
Overscan is something a display (TV) does. It's not related to the signal itself.
Blanking is what's left of the signal if we subtract active image, sync pulses and color burst. In other words, the front and back porches as well as most of the vertical sync area are blanking (altough the latter can be used to carry data such as closed captioning in 60 Hz systems and teletext in 50 Hz systems).
According to ITU-R BT.601, 720 covers the duration of 53.333 Ás of the analog signal, which means it carries some blanking (part of the front and back porch) for both 50 and 60 Hz systems. This is the commonly observed ~16 pixels in total at the left and right sides of a 720 capture.
The math is 53.333 Ás x 13.5 MHz = 720 px
Skiller, I guess this is what I meant. It is interesting that on the second picture there is a 40T area (16+24 or 8+32) that corresponds both to the analog blanking and to video data (but is excluded from digital blanking).
Last edited by ConsumerDV; 12th May 2022 at 13:21.
As professional video equipment usually use 10 bit video then all video samples are sent in 10 bit (8.2) format - 8 bit equipment ignore (truncate) fractional part of 8.2 format) - raw video speed is 270 Mbits due of this.
Overscan is unrelated to analogue and digital video standards - it is not covered either by BT.470/BT.1700 or BT.601.
The 480 height seems arbitrary though. I mean there are 486 total lines that include at least some image signal (the top and bottom only have half of a line of image, with the other half on each of them being blank). If you assume between 2% and 5% of the image is lost to overscan (I read those percentages on another website), then that leaves you with 476 to 462 lines that would be visible on a real CRT TV. So even if your capture specs were intended to simulate overscan, by capturing only the non-overscan part of the video (cropping out the overscan part), then the image height should be between 462 and 476 lines. So the numbers I'm looking at here are 486, 476, and 462. None of these are 480. I don't see any reason for a height of 480, other than the fact that very old CRT computer monitors (which did not suffer from overscan like CRT TVs) tended to actually be exactly 480 pixels heigh (there were physically 480 rows of pixels on those old CRT computer monitors).
480 seems to be less about the specification of the analog signal, and far more about keeping the digitized copy of the NTSC signal compatible with the CRT computer monitors that existed at the time that the NTSC digitization technique was standardized (making it so that on playback, the video would exactly fill the height of the computer monitor). We have monitors with a lot more vertical resolution than 480 now though, so there's no reason for any modern analog video capture device to not capture all 486 lines.
overscan. Remaining 90% is action-safe area. Title-safe area is usually 70%. Some think that these recommendations should be adjusted or completely rescinded for digital TV, but for analog TV they are as I wrote above.
Today i see commonly that even reputable TV stations don't care about overscan (some news tickers are partially cropped on TV's incapable to do 'just scan') - this is common trend, every generation of engineers (except single cases) seem to be stupider than preceding one. I seriously doubt if today something like NTSC or PAL could be created from scratch - seem monochromatic TV is maximum for modern engineers... Sad but true...
Using neat numbers simplify design, make it possible (feasible from technology perspective) - fractions are difficult to deal, require additional error processing and there is no benefits from it - so if you can loose 6 lines and cut complexity by half then it is obvious decision - it cost less, has no practical impact on quality - imagine using 4164 DRAM circuits with 250ns R/W cycle - you need 8 of such IC's to create 64KiB memory where you can store array 256x256 8 bit pixels so not even single field of your composite video, 250ns R/W cycle means that you can maximally transfer 4MB per second i.e. you need introduce memory interleaving to get faster transfers and memory interleaving means you need more memories so practically you need to use not 8 IC's but 32 of them to create 512x512 array capable to transfer at least 16MBps - 32 DRAM's means you need large PCB and additional electronics it means it will be power hungry and definitely not for consumer use.
At the beginning of 90's DCT require dedicated, specialized IC capable only to perform 8x8 DCT on 8 bit integer with speeds like 15MPixel per second... you need few of such IC's to build video JPEG realtime codec...
Every engineering decision made in 70's/80's/90's there is strong foundation of math and technology limitations from those times... Imagine that DCT as foundation for every modern video codec doesn't existed before 1972 (and it took almost 10 years to adopt mathematical theory behind DCT to practical implementations).
Those 6 lines are meaningless especially that in 99% cases there is no useful video there.
Last edited by pandy; 15th May 2022 at 04:50.
All of your questions are answered in a book I recommended to you in October.
It does make sense that consumer capture devices stick with the consumer 480 resolution, while the professionals who won't be troubled by the "garbage" at the top of the signal can use professional 486 capture devices. The fact that 720x480 captures reveal VHS head-switching noise is already confusing enough for consumers.
Been meaning to add some examples to my old thread, so here you go: https://forum.videohelp.com/threads/359690-Geek-fun-with-VBI-Macrovision-DVD-recorders...es#post2656567
Last edited by Videogamer555; 15th May 2022 at 16:09.
Don't use blanking lines for anything except the reserved values (e.g. CC, VITC, WS, program guide). So not just anything else added there.
Don't put "essential" image info anywhere except safe area. So, nothing important around the edges.
Use the most common pro standards for universality of interchange. So 486 is not expected to be as common as 480.
Above all else: anything you/we are capturing off of consumer formats (vhs, etc) would almost never be considered "masters", unless that is the only format on which it exists (e.g. camera raw footage from home video).
VBI is not video so to parse VBI use dedicated HW.
Going with your way of thinking then for archival copying you need to store also VBI area i.e. stop recording video as video but record samples with sync pulses and remaining structure - for example Macrovision - it should be archived or exactly opposite - Macrovision shall be stripped from source?
Where is border in accurate copy? At sync level, video samples? content?
Last edited by Sharc; 16th May 2022 at 05:14.