technical - which 480 of 486 lines are captured?

[jmac698]

Hello,
this is very technical and I don't really expect an answer, anyhow does anyone know which 480 of the 486 lines of NTSC are captured by capture cards? That's something I've never been able to figure out, and I've even read several standards.
One standard talks about corresponding middle of the line in analog and digital, and I found some references to the way lines were numbered, but couldn't figure it out.
i.e. , line 21 with CC is certainly not part of the lines that are captured.
Thanks.

[juhok]

Blackmagic Decklink captures 486 lines so the answer in this case is..all lines? :) My ex card (PDI Deluxe) had offset of about 8 lines upwards (adjustable using BT8X8 Tweaker) so that was definetely not right nor standard. I guess there's many ways to skin a potato.

[zilog jones]

One standard talks about corresponding middle of the line in analog and digital, and I found some references to the way lines were numbered, but couldn't figure it out.
i.e. , line 21 with CC is certainly not part of the lines that are captured.

That line 21 would be out of the full 525 lines, including the vertical blanking interval.

[lordsmurf]

It really varies.

[jmac698]

My capture cards seem pretty consistent, and of course they only produce 720x480 files. There must be a standard for this, else how could bt601 have been defined?
There's no way left for me to even produce 486 lines, no modern device will play them back.
The only way I could hope to measure this is using an old VCR recording, counting lines of some frame on a real analog TV (using vertical hold to scroll up enough to see), and then comparing it to a digital capture.
Can you think of any other way of producing a true 486 line signal?
I wonder if I can pinpoint which of my recordings where the last 6 lines disappeared...I don't think TV is using them anymore, since they are just rebroadcasting satellite signals in the last analog channels I have.

I'd like to know what file format your blacklink uses, as some file formats can't create a 486 line video....

[Ghitulescu]

Can you think of any other way of producing a true 486 line signal?

Sure, you design yourself an aquisition card that transform the RF it gets into a series of bits. Then program a software to interpret those bits. How hard can this be?:p
Because any capture card would have to obey the standards if the manufacturer wants to market it as NTSC (or PAL or SECAM). End of story.

[zilog jones]

DScaler might let you change exactly what lines it is capturing. I know with my SAA7134 I can make the VBI visible, but I am usually only caputuring PAL so there's no missing video lines like with NTSC.

[juhok]

I'd like to know what file format your blacklink uses, as some file formats can't create a 486 line video....The specs are available at their site (http://www.blackmagic-design.com/products/decklink/) (4:2:2 raw YUV or MJPEG etc.). Here's a 5 frames sample (http://www.siluriformes.net/doom9/Decklink_NTSC_sample_720x486_YUV.avi) recorded from NTSC D-Threater player idle screen.

[Emulgator]

A very good source is here, written by Chris Pirazzi,
someone who helped Apple to get the video thing solved:

http://lurkertech.com/lg/video-systems/

It seems to suggest that the majority of NTSC 13.5 MHz sampling capture devices start
to capture with video signal line 21 and so discard 6 bottom lines of the 486 analog lines.
Devices may vary. Once arrived in digital NTSC you cannot have these 486 analog lines anymore.
Other devices may start capturing 3 video lines later and end sampling 3 lines before end to keep the picture center.

The "discard 6 bottom lines approach" would be:
Start to capture with video signal line 21 (being second off the top picture line, first of bottom field)
through 260, leaving lines 261, 262, 263 unsampled,
nothing during blanking (video signal lines 264-282,)
then again 283, being top picture line and first of top field - 522, leaving lines 523, 524, 525 unsampled),
thus 480 line sampling would effectively have to stop
before it reaches the bottom 6 lines of the 486 picture lines present in analog NTSC video.

But best is to actually try and look at the captured output,
because this approach actually shifts 720x486 analog picture center for the digital 720x480 result.
The existence of Closed captions informations in line 21 would suggest
that the "discard top and bottom 3 lines" approach should be more useful,
but here I am in PAL land, so on thin ice with my thoughts...

[zilog jones]

Discarding the bottom lines would make the most sense as there's more likely to be head switching noise there from certain analogue tape sources.

[jmac698]

I found some information on this, a programmers guide to mpeg2 (or video), can't find link now.

He suggests that in 2000 or so onwards, a de facto standard has been to start at the beginning and lose the last 6 lines. He gives a line numbering as below, however this doesn't correspond with the numbering of the previous poster:


line field content
283 2 first line of active analog video
21 1 line2, US CC (EIA608) channels 1&2
284 2 line3, CC line 2, channels 3&4
22 1 line4
285 2 line5
....
404 2 line243, end of top half of picture
142 1 line244, start of 2nd half (of analog standard)
...
263 1 line486

thus, you can tell what your card is doing by seeing if there's CC flickering lines at top. They should be the 2nd and 3rd lines if you digitizing is centered properly.
He suggests that digitization is normally not centred properly, being 3 lines too high.

And I thought of another way to tell, a modern soundcard with high sample rate (96KHz) should see 6 samples per line of video.
Through oversampling of a still frame and a lot of work, you could eventually digitze the whole hidden line with your sound card...
that is by waiting to sample those 6 samples at different phases, and then solving for unknowns.. similiar to CT scanners. Crazy idea though :)

All references are to NTSC.

[lordsmurf]

I'm curious. Why does it matter which 6 are missing?

[ChiDragon]

Some info in case anyone else stumbles upon this topic in a search. As lordsmurf said, it varies -- not just device by device, but even from one driver version to the next using the same hardware in one case I noticed. However...

I've been slowly reading Charles Poynton's Digital Video and HD Algorithms and Interfaces. As part of a diagram of 525-line video, he mentions which lines SMPTE RP 202 says to use for MPEG-2. I think it's worth noting, though, that this standard from 2000 is recent compared to BT.601.

ATSC A/54a refers to SMPTE RP 202:
The interlaced formats are based on ITU-R BT. 601-5; the progressive formats are based on SMPTE 294M. If the input is based on ITU-R BT. 601-5 or SMPTE 294M, it will have 483 or more active lines with 720 pixels in the active line. Only 480 of these active lines are encoded. The lines to be encoded should be lines 23–262 and 286–525 for 480I and lines 45–524 for 480P, as specified in SMPTE Recommended Practice RP-202, “Video Alignment for MPEG Coding.”

So that's what studio hardware is supposed to be doing, and it also seems to be the most common arrangement among the capture devices I have used.

The DV standard captures the same lines from field 1 but one line earlier from field 2: lines 285-524. (Bottom-field first.)

[TheSkiller]

But I struggle to translate this into a scenario where I have a 720x486 frame (say, a computer rendering) and need to crop off 6 lines for 480 lines distribution (say, DVD). Which lines do I crop to replicate the behavior of studio hardware? Also, is analog video captured with 486 lines typically TFF or BFF?

[raffriff42]

It's not mentioned in Rec.601 (http://www.itu.int/rec/R-REC-BT.601-7-201103-I/en), as that standard digitizes everything, including H- and V- blanking.
The best I could find was this excellent article, excerpted below. It's a complex issue:

Programmer's Guide to Video Systems - Lurker's Guide (http://lurkertech.com/lg/video-systems/) - lurkertech.com The Standard-Def Pixel Debacle (http://lurkertech.com/lg/video-systems/#debacle)

The Result: Chaos
The result was that, for many many years, different brands and models of video input hardware would give you images with:

different widths and heights, or even worse,
the same width and height that represented a different region of the underlying video signal

Because many software engineers were not familiar with video, they would assume that if they have an image of a certain size, say, 640x480, that it must line up exactly with all other images of that size...

Here are some hints for different situations:

Video Input and Output devices: You need to know how your hardware device will line up your pixels in memory with the real video signal coming in or going out. You may have to contact the manufacturer to find this out.

DVC (DV/DVCPro/DVCAM/1394/Firewire Video, aka Blue Book, aka IEC 61834): You are in luck. The DVC compression standard was shockingly ahead of its time. Someone on the DVC design committee actually thought about video, and wrote down in the DVC spec exactly how the compressed DVC data lines up with standard-def video signals, exactly what the pixel aspect ratio is, and exactly how fields are handled.

System 525-60

Field 1: line 23 to line 262
Field 2: line 285 to line 524

System 625-50

Field 1: line 23 to line 310
Field 2: line 335 to line 622

[ChiDragon]

Also, is analog video captured with 486 lines typically TFF or BFF?

As raffriff42 alluded to, there's no standard for which lines a 720x486 frame contains: https://cpcweb.com/blackmovie/

It's easy to deduce if you capture something that already contains line 21 captioning, but it varies with each setup.

But I struggle to translate this into a scenario where I have a 720x486 frame (say, a computer rendering) and need to crop off 6 lines for 480 lines distribution (say, DVD). Which lines do I crop to replicate the behavior of studio hardware?

If it's a fully-CGI image that fills the frame, I don't see how it would make a difference. It's when you're processing from an external source that it matters.

[TheSkiller]

Well, it's more like an academic question, sure.
It's not exactly that I care which lines need to be cropped, I just want to keep the original center of the image centered so that after a 486 -> 480 -> analog conversion the image is still perfectly centered according to studio standards.

Of course, this is assuming there is a standard for where to put the digital 480 lines when converting to analog. :rolleyes:



@ Thanks for this excerpt.
So, if line 21 of the analog signal is the first line to carry image, starting our first digital line at line 23 would mean we are cutting off 2 lines at the top and 4 at the bottom for a 480 lines result. Sounds good to me.

[jmac698]

This has made a difference to me. I was doing some tests with producing and recording test signals, and each loop through analog moved the screen up 2 lines, so keep that centering :/

[jmac698]

I just remembered why this made a difference, and it still does. I was making a video with the framenumber embedded as a digital code in video (white lines denoting a binary number, like Close Captioning), and I found that when I captured the video back, the lines were gone. This is because the top line in the encoding was not captured by the card. Put it this way, the card captured 2 lines lower than expected. So I had 2 black lines at the bottom, and the top 2 lines were cut off. It matters how many lines capture cards will be offset, so that I can place my binary code at a line that will always be captured (and I didn't just want to put it in the middle of the video, that's a bit distracting).

I used this for my glitch analyzer. http://forum.doom9.org/showthread.php?p=1462931

[ChiDragon]

The lowest I have seen is the ATI 750 USB, which stupidly starts capturing at line 24, continues to the bottom half-line 525, and then ends outside of the picture with a line of superblack + garbage data for chroma.

[pandy]

SMPTE 170M-2004
available for example as part of the http://www.itu.int/rec/R-REC-BT.1700-0-200502-I/en

[ChiDragon]

What's the relevance to this thread? The title of 170M is "Composite Analog Video Signal — NTSC for Studio Applications" and it just describes the properties of analog. The numbers 720, 486, 483, and 480 don't appear.

[pandy]

What's the relevance to this thread? The title of 170M is "Composite Analog Video Signal — NTSC for Studio Applications" and it just describes the properties of analog. The numbers 720, 486, 483, and 480 don't appear.

Are you capturing digital data or analog signal? If this is analog signal then analog signal need to fulfill SMPTE 170M.

Notation how video lines are numbered you can find in multipple sources for example http://forum.doom9.org/showthread.php?p=1664603#post1664603

btw 720 pixels is digital not analog. Particular description for digital video you can find in two standards ITU-R BT.601 and ITU-R BT656 .

https://www.itu.int/rec/R-REC-BT.601-7-201103-I/en
https://www.itu.int/rec/R-REC-BT.656-5-200712-I/en

[ChiDragon]

I still don't get your point. The analog standards don't specify which lines should appear in a 480-line digitized signal.

[Ghitulescu]

Analogue standards do not indeed describe how a digitisation would occur (like Herr Benz describing in 1900 the OBD2 interface).
Yet they describe the middle of the image, and what lines are included in the overscan.
Even if such standards exist, this doesn't prevent the manufacturers to ignore them, partially.
In practice, since the video programmes of analogue era were manly stored on tape, there have been lots of distortions down (mainly head switching noise) and only a few up (mainly colour issues). So even if the standards would prescribe which lines are blanked and which not, or where the center of image is located, some manufacturers might choose to crop less from top and more from bottom (to better hide the head switching noise). This posed actually no issue in the practice, as the camera (and the director) always had the main subject within the safe zone or safe area, which was generously set.

[drmpeg]

Just FYI, the Blu-Ray specification states that for 525/60 video, line 23 to line 262 of field 1 and line 286 to line 525 of field 2 will be encoded. For 625/50 video, it's line 23 to line 310 of field 1 and line 336 to line 623 of field 2.

Ron

[pandy]

I still don't get your point. The analog standards don't specify which lines should appear in a 480-line digitized signal.

How's that? I'm reading spec and i see clearly where SAV and EAV with bits F and V are set... Not sure about SMPTE - is there any different standard for digital SD video in US different than 601 and 656?
--
Answering myself, yes, indeed SMPT has series of various standards covering for example 768x486 video, sadly i must say they are not available freely.

Which lines should be captured (or rather how capture window should be positioned across video frame) probably can be indirectly taken from SMPTE RP 27.3-1989 (also not free) .