is this true? resolution myths?

[dvd_maniac]

got this from http://www.dvdrhelp.com/forum/userguides/98177.php
I capture with ati aiw 9700 pro from digital cable and directv sat. I use s-video and if this is true I should only cap at 352x480 if I want to author dvd with it?
Now, it is a perfect opportunity to clarify a myth here: There is no way to grabb from any analogue source (including TV broadcasts) and with the use of "S-Video in" to get a usefull resolution beyond 384 X 576/480.
The reason for this, is that in analogue technology, there are no horizontal lines like the digital world, but horizontal DOTS. 720 dots in a raw, creates what we call a line. With the analogue transmissions, if a dot is missing in the raw, there gonna be no replacement, just a null dot, a nothing. The lenght of the raw, gonna be always the same with or not some dots. For example:

A line with all the dots: ...........................................................
A line with some dots missing : ... ............. ............. ..................... ...

In theory all types of connections/cables are capable to carry a picture of CCIR-601 resolution from a source to a screen, but there gonna be loses in the dot raw line like those in the example. Those loses are more for composite, less for S-Video and even less for Component and RGB. It is not easy to expain it further in a simply post, maybe a future article
With S-Video, those loses are about 1/5 of the video in theory. In practice and mainstream cables/connectors, is more, about 2/5 or the original info.
So you end up with 384 horizontal usefull dots X 576/480 vertical lines. And becase in digital world only the usefull info counts, any grabb with the use of S-Video beyond 384 X 576/480 is a fake!
If you want source on higher resolutions, you can experiment with DV, DVB transmissions (when you grabb direct the stream with the use of special hardware), DVD - Video (ripping) and source which came from Component adaptors or RGB ones. The full Euroconnector cable (known more as SCART) includes sometimes the RGB info, but a Full SCART to SCART cable required , with high quality parts and those cables costs a lot!. Also, the difference ain't easy noticable on a TV, but is very noticable from the encoders!
Overall, don't expect higher resolutions beyond 384 X 576/480 with the use of S-Video! It is practicly impossible to have them! And because 384 X 576/480 ain't a supported resolution in any optical media, go straight for 352 X 576/480.

[Kika]

Not realy.
Try it by yourself: Capture the same Video at 704x480 and 352x480. Encode them without any resizing/filtering. Which one is the sharper one?

He is right. Max. Resolution of S-Video (NTSC) is arround 384x480. But you have to capture a higher resolution to get every horizontal Line (not Scan-Lines!). Horizontal Lines (VHS has 240, S-VHS up to 400) aren't Pixels and can't be handled like Pixels. To be correct: You have to Capture double the amount of Pixels than the Video-Type provides in (horizontal) Linies.

[dvd_maniac]

I tried both, but did not encoe them. I capped directly to mpeg-2. both at 720 and 352x480.
I really couldn't tell the differnce. I have one moreestion though.
Should I be able to lower the bitrte yhalf if I cap 352 intead of 720 and expect the same quality since it is only grabbing around 352 anyways? I ask instead of trying because my eyes are not the best and want another opinion.

[ppera2]

I can't agree with things mentioned in that article. Even with composite connection is very visible more sharpness in higher resolutions. Aprox. over 640 I don't see difference. For me. it looks that they talk about S-VHS and not S-Video connection.
According to that, that active line time is around 52 microsec. and max freq. by broadcasting is 5-6 MHz, horizontal res. can be over 500 pixel. Better analogue (studio) eauipment has more resolution (768, declared).
However for TV is irrelevant is connection composite or S-VHS - it is multiplexed in broadcasting, and what use of that, that it's later separated?
Advantage of S-video is mostly in that, that no interference between chroma and luma. Actually it's real useful for: TV out and for capturing from S-VHS VCR.

Back to resolutions: it's always better to capture in bigger resolutions, because capture itself will be more precise (oversampling). Loss, mentioned in article will be always lower if capture in higher res. Later you can resize to desired and to source best fitting res.
Of course that for lower res. you can use lower bitrate by capture.

[vhelp]

Hi guys..

Another important fact is this...
If your goal is to make DVDs, then you want to make them w/ the highest
resolution you can go, w/out taxing your given setup, and limiting
yourself to 384x !!

I do make encodes w/ 352x480 from VHS, but am now beginning to see
an actual difference when I use 704x as my minimum.

But, don't expect to see better results if you capture at 384 and
then raise it to something like 704. That's why its better to just
bake two pies in ONE oven, and capture at highest your given system
will allow you to w/ out such issues as framedrops and others nonsense
things, ie, 704 or 720. Then, if you want, you can bring things down
a little like 352x480 for instance.

Also, note that for VHS movies that are in WideScreen format, you gotta
capture it at highest you can go.. and for me, that 720. Then, I'll
bring it down to just 704, and widescreen encode it, and it looks great :)
in my AD-1500 player, and in widescreen view on my cheap 13" tv.

So, for the most part, and for most people who's goal is highest
resolution for a DVD project, 704 would be minimum (for me anyways)

Have a great day all,
-vhelp

[FredThompson]

Consider the source. DVDRHelp has a number of less-than-logical guides. If you're truly interested in knowing what is possible over analog signals, there are far better sources for information.

With ANY digital sampling, you will have far better results if you sample at a far greater frequency than the analog signal. Best is 4x or higher, 2x is still fine. To illustrate, an audio CD has 16 bits of resolution but truly professional ones are mastered at 96 and downmixed as the final operation. (OK, well, there IS a form of post-processing which can add apparent punch but that's a little different.)

If you capture at the highest rational size and depth, do your processing, then downmix for the final product, the results will be superior to those where the sample is made at the final product resolution.

[vhelp]

I consider vcdhelp's guides and factoids to be as honest and helpful
in many areas :)

But, once in a while, you find a bump on the road, and you can't help
to wonder :rolleyes: what gives.

So, what I'm saying, is I'm a person who tipically punches through
what logic and math might say (or prove) and find myself going against
the grain :p hence, the dissbeleaf that you gotta capture at lower
res cause it's not "mathematicaly" or "logically" sound to capture at
higher res. I believe the opposite. You capture at the highest res
your given setup will allow, and adjust to something comfortable w/
your final goal ie, CVD or DVD etc.

So, the rule of thumb is, in this case, go against the grain :p
-vhelp

[dvd_maniac]

my sysrem can handle 720x480 even in huffy avi capture. my problem is trying to fit as much video as possible onto dvd-r without quality loss. I record my favorite tv shows and do not want to waste too many blanks but want them too look nice.
ps does anybody here capture with ati radeon and encode with CCE?
my huffy avi's come out squashed and scrambled. And don't know how to import mpeg-2 files.
could use some help.

[FredThompson]

It's a huge discussion but what you describe sounds like 2 things: the noise of capture increasing your storage bandwidth, the need for proper filtering.

If filtering is done properly, it reduces the random inconsistencies of noise. Done improperly, it makes the image cartoonish. By way of comparison, satellite MPEG2 transmissions are typically less than 1G/hour of material, higher for premium source like specialty movie channels, pay-per-view, etc.

I have an All-in-Wonder Radeon but don't capture with it anymore. Maybe things have changed with the more recent models. That one did not have proper sampling. Now I use DV passthru or a hardware MPEG2 encoder at a very high bitrate.

DVD handles all kinds of resolutions and bitrates. TANSTAAFL so there comes a point where you'll have to decide which is cheaper, your time to fiddle with the captures or the media. I, personally, prefer to use a higher bitrate to keep detail under the theory that storage costs and space will continue to decline so it's better to do it "right" than to do it over, if "doing it over" is even possible.

[SeeMoreDigital]

There's are many 'trade off' decisions you may have to make when capturing.

Capturing 4:3 images is not so bad. But things can get a little awkward when trying to capture 16:9 images.

As you know, an NTSC 1.77:1 (16:9) image viewed on a 16:9 TV, actually starts it's life as an 1.50:1 image. So in order for the image to perfectly fit a 16:9 TV it has to be (horizonally) anamorphicly stretched, by around 20%.

PAL 1.77:1 (16:9) images start their life as just an 1.25:1 image. So in order for this image to perfectly fit a 16:9 TV it has to be (horizonally) anamorphicly stretched, by over 40%!

So as you can see the PAL image has to be stretched twice as much as the NTSC image!

Never the less, some people may find it interesting to conduct their own 'horizontal pixel' stretch tests. Just to see how far they can put up with a streched pixel before they can't take it anymore!

352 pixels is about enough for me!

Cheers

[Piper]

Originally posted by dvd_maniac
my sysrem can handle 720x480 even in huffy avi capture. my problem is trying to fit as much video as possible onto dvd-r without quality loss. I record my favorite tv shows and do not want to waste too many blanks but want them too look nice.
ps does anybody here capture with ati radeon and encode with CCE?
my huffy avi's come out squashed and scrambled. And don't know how to import mpeg-2 files.
could use some help.

I also cap at 720x480 (huffyuv with ATI AIW) with my destination being dvd. What I do is filter the sourc, delete commericals and resize down to 352x480 and create a second filtered source file which is also huffyuv encoded. Then using the second file, I encode to MPEG-2 Half D1 DVD (TMPGEnc) and burn it to a dvd+rw when I'm finished. Using this method, I cannot honestly tell the difference between the original source and the dvd. In fact if it's filtered correctly, it can actually look a bit better. I find 352x480 is a great resolution for TV caps.

If it's a particularly good episode, I'll encode it (second source) to DivX and dump it on a CD-R. The quality is not quite as good as on the DVD, but works well for my purposes. Of course if DVD's were as cheap as CD-Rs, there would be no question. :-)

Sorry, I can't help with CCE.

[Kika]

I find 352x480 is a great resolution for TV caps.

Yes, it is. But you can expect better results on capturing 704x480, because Half-D1 is based on this Resolution and not on 720x480.

[Wilbert]

Some issues:

@ppera2,

Back to resolutions: it's always better to capture in bigger resolutions, because capture itself will be more precise (oversampling).
I don't understand this. Sampling happens when converting the analog signal to a discrete one. The number of samples is always the same (NTSC: 910, and PAL: 1135; from page 13 data sheet Bt878). After sampling, the clip is resized to the resolution which you used when capturing (with some cards, overscan is included).

In other words: if you capture in bigger resolutions, the capture will be more precise, but not due to oversampling. It is just because you are downsizing to a less smaller resolution.

@Kika,

Horizontal Lines (VHS has 240, S-VHS up to 400) aren't Pixels and can't be handled like Pixels. To be correct: You have to Capture double the amount of Pixels than the Video-Type provides in (horizontal) Linies.
I also don't understand this. I guess you are talking about the Nyquist frequency here. This Nyquist frequency stuff doesn't imply you have to double the amount of pixels, wrt the number of active scan lines. Like I said above the number of samples is not related to the capture resolution.

Nyquist sampling means the following: "In order to recover all Fourier components of a periodic waveform, it is necessary to sample more than twice as fast as the highest waveform frequency v." (ref: www.mathworld.com)

On the same page of the same datasheet, they talk about sampling: "UltraLock^TM operates at CLKx1 although the input waveform is sampled at CLKx2 then low pass filtered and decimated to CLKx1 sample rate".

Where CLKx2 (CLKx1) is the 8*Fsc (4*Fsc) clock signal. This is indeed Nyquist sampling.

[ppera2]

Originally posted by Wilbert
Some issues:

@ppera2,

I don't understand this. Sampling happens when converting the analog signal to a discrete one. The number of samples is always the same (NTSC: 910, and PAL: 1135; from page 13 data sheet Bt878). After sampling, the clip is resized to the resolution which you used when capturing (with some cards, overscan is included).
In other words: if you capture in bigger resolutions, the capture will be more precise, but not due to oversampling. It is just because you are downsizing to a less smaller resolution.


Interesting... So, BT chip performs some resizing during capture.
In that case it's worth to capture in final resolution. In res. what is moduo of 1135-some pixels, if I understood it well.

[FredThompson]

2x isn't enough, you need at least 4x to accurately sample. The theorem is correct but not accurate. 2x could conceivably sample on the 0 zero as a wave is going up and coming down, the result would be nothing.

[Kika]

@Wilbert

I also don't understand this. I guess you are talking about the Nyquist frequency here.

No, i'm not talking about the Nyquist Theorem here. First we have to make clear what we are talking about: Horizontal Lines. They aren't Pixels. One Line is the Time for "drawing" a complete Sine-Waveform (argh, my english-problem again). It is the periodic Time of the Waveform. So a Line can start Black and end up White (or anything between them). Between the Mixima/Minima there can be any shade of greey.
240 Lines means, you can have 240 complete Waveforms per Scanline.
You have to capture 2 Pixels per Line to get the start- and endpoint of the true signal (the sine wave). If you capture only one Pixel per Line, you get the AVG of the Waveform - and that's never white or black, it is always a shade of greey.

Sorry, i can't descripe it better in english :(

[FredThompson]

Originally posted by ppera2
Interesting... So, BT chip performs some resizing during capture.
In that case it's worth to capture in final resolution. In res. what is moduo of 1135-some pixels, if I understood it well. Only if you don't plan to filter and you trust the internal rescaling to be proper...

[FredThompson]

You mean "sine" wave. Sinus is different, the most common usage is the passage from your nostrils to your throat ;)

[Kika]

Um, i was way to slow... ;)

@Fred

OK, corrected this sinus-sine-thing... :)

And yes, BT-Chips are acting like that. They are trying to eliminate this greey zone between the Max. and Min. Point of the Waveform. This results in a sharper Picture.

Oh, and the more Horizontal-Lines, the better the sharpness - in 2 ways! More Lines = more Informations AND more Lines = more sharpness of one Line, because the Lines are "smaler".

[Piper]

Originally posted by Kika
Yes, it is. But you can expect better results on capturing 704x480, because Half-D1 is based on this Resolution and not on 720x480.

Interesting you should say that. While I haven't experimented with capturing at 704x480, I have with 640x480. If anything I've noticed fewer artifacts with a 640x480 source than a 720x480 source (albeit ever so slightly softer).

I'll give 704x480 a shot. It makes sense, horizontal being x2 Half-D1. Thanks for the suggestion.

[FredThompson]

Another option is to capture at 720x480 then crop to 704x480 before resizing. Most broadcasts aren't centered and if you keep the black edges you'll be averaging them into "good" data. The other aspect (there's a dangerous multiple-meaning word for a video processing forum) is you'll have more "good" data and a true halving of resolution of the actual image you want.

[Piper]

Originally posted by FredThompson
Another option is to capture at 720x480 then crop to 704x480 before resizing. Most broadcasts aren't centered and if you keep the black edges you'll be averaging them into "good" data. The other aspect (there's a dangerous multiple-meaning word for a video processing forum) is you'll have more "good" data and a true halving of resolution of the actual image you want.

That's a good point. I've been doing precisely that with my VHS captures which are often way off center. The difference was that I wasn't cropping to exactly 704 before resizing.

There is of course the next issue of which resizer to use. :-)

[ppera2]

Originally posted by FredThompson
Only if you don't plan to filter and you trust the internal rescaling to be proper...

Filtering can be performed while not changing resolution.
Trust ot not the internal rescale - we can't avoid it.

Btw. I dl.-ed datasheet, and it's true about fixed sample.
http://www.linuxmedialabs.com/lml33doc/Bt878.pdf

[FredThompson]

Sigh...capture at as high a resolution as possible, then filter, then resize.

Filter before resizing so you have as much data as possible for the filtering. Resize after that with a method which you know is high quality to reduce distortion as much as possible.

[ppera2]

Yes, that's all right. But, when quality of TV video is already poor, it has not much sense. Other thing is time and resources required for max. quality capturing and postprocessing.

In any case, I go on digital SAT capturing from now (actually not real capturing). Quality is far superior, no noise etc.
And it's often in DVD format already. Cards are pretty cheap now...

[Kika]

Another option is to capture at 720x480 then crop to 704x480 before resizing.

Not on BT-Cards. They don't do a proper Scaling to 720. All you realy get is the same Pic like in 768 or 704.

[FredThompson]

Originally posted by ppera2
Yes, that's all right. But, when quality of TV video is already poor, it has not much sense.Did you even read the first line of the first post?Other thing is time and resources required for max. quality capturing and postprocessing.When did the original poster say this was a problem?
In any case, I go on digital SAT capturing from now (actually not real capturing). Quality is far superior, no noise etc.
And it's often in DVD format already. Cards are pretty cheap now... Irrelevant. You're in Europe, the first poster is in America using DirecTV and digital cable. DirecTV is not DVB.

I have DirecTV and have done capture at 720x480 and converted to both CVD and SVCD resolutions. SVCD looks better.

Seems you lost track of the reason for the thread.

[ppera2]

Originally posted by FredThompson
Did you even read the first line of the first post?[b]When did the original poster say this was a problem?

Beg your pardon???? This was reply to you, not to thread starter, who is not anymore in discussion. New discussion started after 'discovering' how BT878/9 works.


Irrelevant. You're in Europe, the first poster is in America using DirecTV and digital cable. DirecTV is not DVB.

I don't know why you react so. Should I talk here only strictly things related to first post? You lost track about how discuss on such forums.
Anyway, maybe you and others in USA should start think about how record digital video without D/A - A/D spoiling.

I have DirecTV and have done capture at 720x480 and converted to both CVD and SVCD resolutions. SVCD looks better.

I agree :)

Seems you lost track of the reason for the thread.

Reason for the thread is to learn something, to exchange opinions. Not to talking to people where live, and because of that not to do this and that.

Btw. your whole often repeated opinion about 'illegal' resolutions by capture is pretty senseless. I can capture in lot of horizontal resolutions, without any black border left and right. And how I will scale it is irrelevant for capture (and it's quality) itself.

[Kika]

Btw. your whole often repeated opinion about 'illegal' resolutions by capture is pretty senseless.

It isn't


I can capture in lot of horizontal resolutions, without any black border left and right.

Still not understanding? That's exactly the problem.
In 704/720/768 you always get EXACTLY the same Picture, only the PAR is different.


And how I will scale it is irrelevant for capture (and it's quality) itself.

And that's not true. It's a difference to Resize 2 Pixels to one or 2.045 to one...

[ppera2]

Originally posted by Kika
...
Still not understanding? That's exactly the problem.
In 704/720/768 you always get EXACTLY the same Picture, only the PAR is different.

Source is same, but digital interpretation is not same! How can be same 704 with 768 pixel???? AR is not so important here. I can correct it by postpro. or by playback.


And that's not true. It's a difference to Resize 2 Pixels to one or 2.045 to one...

Right. But you don't know what is my target res. And we stated here that BT 878/9 does internal resize. So better capture in moduo of internal resolution - if you can. However, I don't see difference in captures at 720 or 704 with BT 878.

[Kika]

Source is same, but digital interpretation is not same! How can be same 704 with 768 pixel????

That's a point we had before. I wrote Picture, not Pixels. All three Resolutions do show you the same Picture. The difference is the amount of Pixels and their AR (the PAR).

But you don't know what is my target res
At last, we talked about Half-D1-Resolution.
And then, there IS a difference between capturing 704x480 and resize to 352x480 and capturing 720x480, cropping to 704x480 and resize to 352x480.

If you goal is to capture as many Pixels as possible, then 768x480 may be a good choice. But for Half-D1, the best you can do is capturing exactly double the amount of Pixels needet.

[jggimi]

Please try to calm down, people, and be as polite as possible. Thank you.

[ppera2]

I don't think that Kika and me have some heavy discussion here. There are only slight differences, and as I see we start to closing.

Just to say that by capturing we always have same picture - but it's, I think very bad formulation. Better to say video. And we can't never get exact 'picture' by digitalising, just closing to it. Difference is not just amount of pixels and AR, but what is most important: that at higher res. we go closer to 'exact picture'.

[Wilbert]

@Kika,

One Line is the Time for "drawing" a complete Sine-Waveform (argh, my english-problem again). It is the periodic Time of the Waveform.
I guess you are right, but I don't understand it. Maybe you can clear it up?

The frequency of the waveform is 4.43 MHz (pal), thus the period is 2*pi/4.43 = 1.42 ms.

But I guess you can also calculate the period in the following way: The duration of a frame is 1/25 = 40 ms, and thus 40/576 = 0.0694 ms per line. Or should I take the total number of scan lines, thus 40/625 = 0.064 ms.

Why are the two periods (1.42 and 0.0694/0.064) not equal?

[ppera2]

Wilbert:

Time of waveform by analogue video should be calculated somehow so:

We have max bandwith of video signal, and it should not be mixed with freq. of chroma carrier. By CCIR (PAL) TV it's about 5 Mhz. So we can calculate that one period is 200 nanosec long. By VHS bandwith is much lower, and on other side, by studio tapes is higher.
One horizontal line is 64 microsecond long, but we have horizontal blank of about 12 microsec, so can calculate that active part of line is about 50 microsec. Dividing it with 200 nanosec we get 50x5 periods. This 250 period per hor. line means 500 changes, or 500 pixels. Sampling freq. must be at least twice as bandwith, what is explained many times. But as much higher, more accurate.

Your first calculation is btw. total wrong, you don't need any pi and chroma carrier freq. here. Right is that 40/625.

As I see Kika's explanation is very confusing, but one line is not for writing 'complete Sine Waveform' It can write much more periods, as I explained at begin, depending from source (capability). Plus, sineform is just theoretical, real video is random. '240 lines' (VHS, I guess) should be for vertical lines, but it's better say just 'pixels'.

[Kika]

@ ppera2

One horizontal line is 64 microsecond long, but we have horizontal blank of about 12 microsec, so can calculate that active part of line is about 50 microsec.

52 microsec. ;)
And yes, that's the same way i calculate it.

As I see Kika's explanation is very confusing, but one line is not for writing 'complete Sine Waveform' It can write much more periods

Maybe my explanation was a little clumsy. I talked about horizontal Lines, not the Scan Lines. At VHS-Bandwith, each Scan Line can include up to 240-250 horizontal Lines, and each horizontal Line is one periode. Guess that's more clear, isn't it?

Plus, sineform is just theoretical, real video is random.

Right. But Sineforms are used as a measurement for the horizontal analogue Resolution.

@jggimi
I agree to ppera2. That's not a heavy discussion or a flame war.

[ppera2]

Originally posted by Kika
@ ppera2
Maybe my explanation was a little clumsy. I talked about horizontal Lines, not the Scan Lines. At VHS-Bandwith, each Scan Line can include up to 240-250 horizontal Lines, and each horizontal Line is one periode. Guess that's more clear, isn't it?


Why you say that Scan Line include up to 240-250 horizontal lines.... ?

Scan line is a horizontal line. It can include pixels. I don't see about what bandwith you talk - VHS or TV? Why say that each hor. line is one periode? Hor. line can include many periods depending from bandwith. It's nothing more clear, I think :)
Is this some direct translation from German? I doubt it, but you can write it te mo in German, i can uderstand it.

[Kika]

Why you say that Scan Line include up to 240-250 horizontal lines.... ?

Because that's what the Specs are saying about that.
But if you don't like those term, you can say it this way:

A Scan Line includes up to 250 Lines (VHS). But that's not clear.
Way 2: A Scan Line includes up to 250 periods of a Sine Waveform (VHS).

Don you know some technical Manuals from Camcorders or VCRs? They are "talking" about things like "Resolution, better 400 Lines". These Lines aren't the Scan Lines, because PAL have to have 576 and NTSC 480 Scan Lines, right?

And that's also where the Myth about VHS-Resolution started. The Manuals of most VHS-VCRs are talking about 220 to 240 Lines - and most people are still beliving, these were Scan Lines. But analogue Resolution has (Horizontal) Lines per Scan Line as it's measurement.

[Wilbert]

@Kika,


Because that's what the Specs are saying about that.
But if you don't like those term, you can say it this way:

A Scan Line includes up to 250 Lines (VHS). But that's not clear.
Way 2: A Scan Line includes up to 250 periods of a Sine Waveform (VHS).

And that's also where the Myth about VHS-Resolution started. The Manuals of most VHS-VCRs are talking about 220 to 240 Lines - and most people are still beliving, these were Scan Lines. But analogue Resolution has (Horizontal) Lines per Scan Line as it's measurement.

Interesting! Do you have some references/books which explains this stuff a bit (ie. line = one period of a Sine Wavefrom, etc.)?

How many lines does a scan line include, when talking about tv's?

[ppera2]

I see. Those lines are actually vertical lines - resolution better than xxx VERTICAL line. Vertical is often not accented. And you wrote it like horizontal, what is of course wrong.

[ronnylov]

All this talk of horisontal lines of resolution and pixels and oversampling is a little bit confusing. But I found this link:
http://members.aol.com/ajaynejr/vidres.htm

They talk about the circular rule. Quote from this link:
"When measuring "lines of resolution" for TV sets or monitors, a reference distance needs to be specified. Traditionally, video resolution is measured across the largest circle (not an ellipse) that fits in the space we are talking about. All the technical books on video say this, although using different words. On a standard TV screen such a circle would span 3/4 of the screen width. For U.S. HDTV (16:9 aspect ratio) the circle would span a little over half the screen width. Since the circle exactly fits the screen height (assuming no overscan), the phrase "picture height" stands for a distance equal to the diameter of the circle and is also used when talking about horizontal resolution. The circle rule also applies to film.

Under this rule, if a standard (4:3) TV set can reproduce 800 dots alternating black and white across the screen width using an input signal representing same, it is said to have 600 lines of horizontal resolution. However if the source material had just 330 lines of resolution you will see just 330 lines of resolution. A 16:9 aspect ratio TV that can reproduce just 800 dots across its screen would have 450 lines of horizontal resolution."
End of quote.

Should I interpret this as if VHS has 240 lines of reolution this would correspond to 240*4/3= 320 dots of alternating black and white across the screen on a traditional 4:3 TV? So even if I capture at a higher resolution the real resolution is not higher. But some sources can have "over 500 lines of resolution" (like DVD-players). This would correspond to over 667 dots of alternating black and white. This sound correct because if you count the overscan area this may leave around 670 "visible pixels of horisontal resolution".

I also found this link:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/750

The table is a good reference. There also some formulas and stuff.

Then we had the original question of the capturing via S-VIDEO cable. If somebody could tell the bandwidth of this cable connection we could calculate what resolution it corresponds to. If it's higher than the bandwidth of the video source then the source is the limiting factor.

Since the table show that NTSC maximum limit is 338 lines of vertical resolution and 451 horisontal active pixels it seems that half D1 destination resolution is not enough to get all the details from the broadcast. SVCD resolution seems to be enough. But if your destination is DVD then you can't choose anything between 704 and 352 so 704 pixels horisontal resolution seems to be the right choice. But all sources are not of maximum quality. VHS source is fine also with half D1 resolution on the final destination format because maximum VHS resolution is 240 lines, which would correspond to 320 "pixels of visible horisontal resolution".

PAL maximum limit gives 403 lines of hor. res. and 538 "total horisontal active pixels". So you need to capture with higher resolution than 538x576. 704x576 seems to be a good choise as both capture and destination format.

I'm not sure how the overscan area should be taken into account. Maybe 480x480 is what you need for full NTSC and 576x576 for full PAL broadcast quality? Anyway the conclusion is the same - 704 pixels of horisontal capture resolution should be enough but half D1 may not show all the details of the original broadcast.

But if VHS-like resolution is good enough for your liking, then by all means choose the half-D1 DVD format as destination.

EDIT: I have read the original post again and he claims that S-VIDEO connection in theory removes 1/5 of the NTSC bandwidth. If that is true then you may end up with 384x480 just like he said. Then analogue video capture in NTSC is not much better than VHS recordings? I normally capture PAL video so that may be the reason I find that half-D1 is worse than full D1 from my analogue captures?

[ppera2]

ronnylov:

This approach with not involving whole hor. screen is by me very wrong.
And it's usual to consider whole width when talk about of res.

Also, as I see, you confused vertical and horizontal lines too:
"PAL maximum limit gives 403 lines of hor. res. and 538 "total horisontal active pixels". " - what this should mean???

It's not NTSC which limits bandwith or res. It's not connection which limits them. Limiter is max bandwith (freq.) by current system.
VHS has lowest (it's 25 or more year old).
Perhaps some good NTSC DVD player or digital SAT recevier has much bigger bandwith than TV programm.

[ronnylov]

Originally posted by ppera2
ronnylov:

This approach with not involving whole hor. screen is by me very wrong.
And it's usual to consider whole width when talk about of res.

Also, as I see, you confused vertical and horizontal lines too:
"PAL maximum limit gives 403 lines of hor. res. and 538 "total horisontal active pixels". " - what this should mean???

It's not NTSC which limits bandwith or res. It's not connection which limits them. Limiter is max bandwith (freq.) by current system.
VHS has lowest (it's 25 or more year old).
Perhaps some good NTSC DVD player or digital SAT recevier has much bigger bandwith than TV programm.

Read the links in my original post. The circle rule indicates that translation from "lines of resolution" on a display to computer pixels of the whole width is made by multiplicate with the 4:3 aspect ratio. I'm not confusing horisontal and vertical resolution. But if the circle is a circle and not an ellipse then the complete width of the display will not be included because the TV does not have 1:1 picture size. The lines are traditionally measured on a width equal the hight of the picture. And each line is either white or black (this is how the lines of resolution traditionally are measured). So if the lines of resolution is 403, then you can se maximum 403 white and black dots on the the same width of sreen as the height of the screan (normally measured in the middle of the picture). OK, this dots may be translated into horisontal pixels of a digital image. But the screen is wider than it is tall. So you need to multiplicate with 4/3 to get the horisontal number of pixels of the complete visible width that you can see on the 4:3 screen. 403*4/3. But you can't see all the pixels, some are hidden in the overscan and you cannot measure hidden pixels on a TV-screen. Thats why I add 7 %. Maybe I'm totally wrong?

403 horisontal lines within the circle on a 4:3 PAL TV:
403*4/3=537.3333
I guess there is some rounding when the table say 538 pixels.

OK, now the method was measuring on the visible picture on the TV. So let's say there is 7 % hidden invisible overscan area we need to compensate - 538*1.07=576. OK this indicate that 576x576 is the maximum analogue broadcast quality by the current analogue PAL broadcast system. I don't know if this corresponds to the maximum bandwidth in the system. Maybe 538x576 is max including the overscan area. I'm not sure if the lines are measured on a real TV or on the source signal.

But that does not meen that my receiver and capturing card and everything is capable of capturing all the details of the complete signal. I was thinking of what is the maximum quality that can be broadcasted on current analogue system and what resolution is needed to create a comparable digital video, and anything higher than that would be overkill when you digitize it. But then the DVD specification has only a few valid resolutions that is best to stick with if the destination is a DVD.

I have a DVB-C card also. Some channels send mpeg2 stream with full D1 resolution. But then I just record it digitally and does not involve any DA/AD conversions. So analogue capture is only interesting when capturing from an analogue source, otherwise try to record it digitally directly if possible. Also if you want to copy a DVD then do it digitally, don't capture it analogue.

[ppera2]

I still say that this aproach where looking circle res. is good just for adding more confusion in all this thing, what is obviously hard to understand for many people.

I'm start to be bored of all this thing. Resolution is just one aspect of capturing. My biggest problem by capturing TV programs is noise, and some artefacts appearing when use diverse compressors like Mpeg2, Mjpeg by compressing so noisy video. Dealing with resolution is simple in compare with noise problem.

[Piper]

Originally posted by ppera2

I'm start to be bored of all this thing. Resolution is just one aspect of capturing. My biggest problem by capturing TV programs is noise, and some artefacts appearing when use diverse compressors like Mpeg2, Mjpeg by compressing so noisy video. Dealing with resolution is simple in compare with noise problem. [/B]

Perhaps this is best moved to a different thread, though I'd be interested to see what you're using in your scripts to handle denoising. Even with as much researching as I've done, I've yet to use anything but Convolution3d(preset="movieHQ") on a consistent basis. The exception may be GuavaComb (in conjunction with C3D) which I've only started to use but I'm very impressed with it so far.

[Wilbert]

Btw, I just found this very great thread: http://forum.doom9.org/showthread.php?s=&threadid=34122

[Kika]

Yes, it's a great thread. ;)
Didn't know that Ookami linked to Karls and my older explanations. And the Thread itself makes things clear, i think. But there is still a problem:

VHS and it's VERTICAL resolution. In theorie, VHS can save all the Scan Lines of PAL or NTSC. In Practise it's limited by the Mechanic of the VCR (maybe someone can translate "Kopfspalt" to english, it's a problem with crosstalking while reading the signals from the tape)...

[FredThompson]

@Wilbert, head over to neuron2's forum to see this argued ad nauseum.

[ppera2]

Originally posted by Piper
Perhaps this is best moved to a different thread, though I'd be interested to see what you're using in your scripts to handle denoising. Even with as much researching as I've done, I've yet to use anything but Convolution3d(preset="movieHQ") on a consistent basis. The exception may be GuavaComb (in conjunction with C3D) which I've only started to use but I'm very impressed with it so far.

We have big space to experiment with many postprocessing filters, but I think that is most important to avoid errors on first step - capturing. My experiences are that if cource is clean, most of 'capture' codec will do job well. If there is too much noise, compressors will start to make artefacts, especially by motion.
For such situations huffyuv is perhaps only solution (except uncompressed). So, big and fast HD is what we need in first place.

[SeeMoreDigital]

I, like alot of people, have kept up with this thread to see how it's been progressing for a while now!

I think we all have to admit that when it comes capturing analogue video from a TV type source it's no as straight forward as it could be.

We should also, all be aware, that NTSC and PAL capturing don't follow the same rules and therefore the chipsets respond to the input signal in different ways.

I live in a PAL area. And my prefered method of capturing (for home use anyway) is either DV cam pass thru' or very high bitrate (9000+kbps) Mpeg2 capturing using the ADS USB 2.0 device (my thanks to Fred for this tip).

However, at the end of the day the ADS is the easiest device to setup and use and 'way' cheaper than buying an DV camcorder. Sure editing is more difficult to do with Mpeg2 but you get used to it!

Cheers

SeeMoreDigital (ex tele-cine operator)