TV Resolution

[Nelson37]

Using mathematical calculations to "prove" which video looks better is like using math to "prove" which cherry pie tastes better. While this may point you in certain directions, it is ultimately irrelevant.

Factors such as how your particular card or software performs the resize, and amount of noise in the original signal, application of IVTC, etc. can throw all these calculations out the window.

I have done numerous TV caps at a variety of resolutions and bitrates. For testing, I cap a repeating program, burn to disk, and then play synchronized with the repeat broadcast, switching between disk and cable with my amp, and compare. I have used various bitrates, generally increasing as I increase cap resolution. Typical values were 352x480 @ 1.8, 480x480 @ 2.1, and 720x480 at 3.5 to 4.0. Bitrates are average VBR, Max values set to .3 to .5 higher. To illustrate my point, these bitrates are indeed somewhat low UNLESS you use real-time IVTC, which interestingly enough works much better using 720x480. Results of real-time comparisons in A-B switching between encoded disk and actual broadcast? CVD looks noticeably worse, SVCD is close, and full D1 is indistinguishable from original source. Myself, wife, daughter, son, brother, and two friends all agree (different test clips at different times).
Also note that the differences are not as noticeable at all times and on all test clips. They are most obvious on hi-action and hi-color scenes.

This is viewed on a 27" TV. Source is usually Digital Cable. Analog cable shows significantly less benefit from higher resolutions, though it is still noticeable. SVHS VCR, using S-video cable, is too close to call although the enhanced performance of the real-time IVTC at the higher res makes the difference in filesize. BUT a bad tape does not IVTC well, so those I do at CVD res.

Also note that "better" means different things to different people. Is a sharper picture with occassional barely noticeable pixelation better than a blurrier picture with no visible pixelation? How about filesize?

[trevlac]

people will often take the full resolution source (720 wide) and resize directly to 352. .... This is also NOT the correct way to do this particular resize (at least according to most TV engineers).:

This will also throw off the aspect ratio a tad. Crop to 704-->resize is better for AR also. SVCD on the other hand should be 720-->480 to keep a proper aspect.

***Back on topic***
NTSC DVD resolution is 480 vertical. ITU601 NTSC digital resolution is 486 vertical. Small difference but can be important during resizing or field order/deinterlace operations.

ITU 601 is 720x480 for NTSC.

Some may say NTSC analog is 711x486 (it's probably a range but 486 is a good V number). But, even if you ment that:

I don't see why this matters. All 'catpure' devices I have ever heard of crop off the 6 extra lines. As long as they don't do it in odd numbers (like 1 from top and 5 from bottom), I don't see the issue because you only ever have 480 on the PC.




@Nelson37

It's good the hear someones test results. Thank you.

I kinda liked the calculation from Cornucopia. I think is was ment to show the relationship between compression and resolution, not as a way to get the 'right' answer. It clearly did have a factor for taste. This shows taste matters alot in picking the answer.

[Cornucopia]

ITU 601 is 720x480 for NTSC.

Some may say NTSC analog is 711x486 (it's probably a range but 486 is a good V number). But, even if you ment that:

I don't see why this matters. All 'catpure' devices I have ever heard of crop off the 6 extra lines. As long as they don't do it in odd numbers (like 1 from top and 5 from bottom), I don't see the issue because you only ever have 480 on the PC.


I kinda liked the calculation from Cornucopia. I think is was ment to show the relationship between compression and resolution, not as a way to get the 'right' answer. It clearly did have a factor for taste. This shows taste matters alot in picking the answer.

No, ITU-601 D1 resolution IS 720x486 for NTSC. I just finished reading it from an ITU publication.
What you and others here and a lot of other people are getting this mixed up with is the difference between real "broadcastable" D1 and the DV format (720x480) or the DVD format (720x480). Even many hardware manufacturers get it wrong...or cut corners.

Don't get me wrong, there's usually nothing wrong with cutting corners, but when you want that little extra something to sweeten the picture, (in terms of math, etc.) it can make some difference.

Other aspects of 601 are different form DV or DVD, but are not as easily confused. For example, color space sampling.
601/D1 NTSC is supposed to be 4:2:2 (Y:U:V)
DV NTSC is 4:1:1
DVD NTSC is 4:2:0

BTW, there ARE companies that do make full-frame digitizing cards. Who are they? Just some no-name companies like AVID, Media100, Sony, Panasonic, Quantel, Matrox, etc. Of the 7 digitizing boards at my disposal right now I have 3 that do D1 720x486, 2 that do DV 720x480, 1 that does the old MJPEG 640x480, and 1 MPEG2 (for DVD) 720x480. Yet, I would certainly agree that most prosumer cards are based around the DV spec, and so are 720x480, even if they do more than just DV.

I also say Analog NTSC is ANALOG! It doesn't have pixels at all! Being analog, it is a continuous waveform. It has lines (525 NTSC, 483+2 halves being active/visible) but in-between-on any particular line-it's just a smooth transition. The idea of pixels is completely digital (though borrowed somewhat from the TV monitor's shadow mask/holes idea). It's a result of the discreet sampling, using the Nyquist theorem, on the raw total usuable video bandwidth.
You can generalize with an estimate or rule of thumb, but you can't use generalized numbers to support a position you're trying to persuade people with.

And yes, my calculation idea was mainly meant as an analogy so as to bring both sides of this continuing arguement to some form of agreement.

Scott

[SatStorm]

You are overanalysing stuff, just for the fun of it!
That is what I believe for this treat!
Anyway!

In short terms, when you are NTSC and you use CCE to encode, SVCD gonna look better CVD. If you use TMPGenc to encode, CVD gonna look better SVCD!

This is a practical rule and has it's roots of how the resizing is done.

- TMPGenc resize itself any source, using the bilinear method. The problem is that it does the best possible job only if the source is cropped D1 (704 x 480/576).
- With CCE SP you always resize before you feed the encoder. And most CCE users use avisynth for this, the best frameserver ever! So, by default, they use the best resizing method possible and this small secondary thing indeed does the difference!

On an old post, Adam mention about resizing and the importance of it when you encoding, so I did almost a year tests about it, so I have a very strong opinion about it!

So, when you use CCE SP, SVCD looks a bit sharper, because SVCD has a better horizontal framesize.
When you use TMPGenc, CVD looks better SVCD, because TMPGenc built in resizing method, do a terrible job on SVCD and a better job to CVD.
Expecially if your source is 704 x 576/480, TMPGenc gonna convert to CVD (352 x 576/480) perfect!
But if you use an external resizer to TMPGEnc (a step only few do when they use this program) then SVCD is a bit sharper.

An other approach for the subject, is let say that CVD looks smoother SVCD and on the eyes of most people that makes SVCD sharper.

But sharpness ain't always for the best you know...

This sharpness is a plus when the source is a DVD or DV. But if your source is an analogue VHS/SVHS/LD/Aerial (satellite, cable, terrestrial) transmission, this "smoothness" of CVD is like a "natural" noise reduction filter.
I can say the same for DVB transmissions, which very rare are beyond 544 x 576 in Europe. TMPGenc resizing method "smooths" the picture when encodes to CVD, and some macroblocs looks less than using SVCD.

About those antialising problems:
Most alising problems with CVD is because of the wrong resizing method also. And most negative judged CVDs, use to be the ones produced by programs like DVD2SVCD for "backing up" DVDs to CDRs. This program is well known for SVCD backups, not CVD ones! Only recent versions do a decent job with CVD and that because now they have to deal with DVDRs, and we all know the DVD standard framesizes

Speaking of PAL, you have to resize bilinear any source to CVD (if you don't capture direct to this framesize) and then encode with TMPGenc. Using an average of 2000Kb/s you have excellent results for movies, terrible for music videos and home movies overall. For "perfect" results, or you use VBR with an average of 3000kb/s or CBR about 3500 (if the source is filtered) or 4000 (when you capture and encode "as is")

The only software realtime mpeg 2 program I believe works best, is WinDVR 3.0, which includes the best on the fly resizing I ever saw for 1/2 D1 captures by the way. A realtime mpeg 2 captrure at 352 x 576 @3500kb/s looks amazing IMHO. Even a capture at 2700 has only few macroblocks on extreme situations with this programs. Far better power VCR IMHO, and no tweeking is needed...

Anyway, before I turn of topic, my opinion is that if your source is analogue NTSC (TV transmissions for example), then CVD gonna look better SVCD overall. More smooth. And if you have a good TV set, then you might notice SVCD a bit sharper, if you encode it with CCE from a source, captured from an almost perfect TV transmission or LD.
With VHS sources, the "smoothness" of CVD is a plus, because it works as a soft filter of the analogue noise.

[energy80s]

Some interesting statements in this thread! I work only in PAL and use either 352, 704 or 720 as a horizontal base. SVCD's aren't that compatible with standalone DVD players - they can have problems re-sizing the image (same goes for Kwags KVCD standard) whereas 352, 704 and 720 are industry standards. As for broadcast frame sizes, the analogue standard is 702x576 active picture elements and the digital is 720x576. However a lot of material is edited in 4:3 and transmitted "pillerboxed" in a 16:9 image. Some broadcasters (not the BBC) have actually dropped the 720x576 res down to 704x576 for digital as they believe that no benefit is to be had from adding the extra 16 lines. For resizing down to CVD or half D1 if your source is 720 wide, crop off 8 pixels left and right off the image. If its 704 wide, do nothing. TMPG will easily cope with 704 - 352 resizing but has more problems with 720- 352 or indeed anything - 480.

[trevlac]

1) No, ITU-601 D1 resolution IS 720x486 for NTSC. I just finished reading it from an ITU publication.
...

2) BTW, there ARE companies that do make full-frame digitizing cards. Who are they? Just some no-name companies like AVID, Media100, Sony, Panasonic, Quantel, Matrox, etc.

3) I also say Analog NTSC is ANALOG!

1) You are right! Kind of.... I went back and read thru Rec. ITU-R BT.601-5. I can find no reference to Active Lines. 480 nor 486 are mentioned. You may be mixing in info from ITU-470, although I'm not sure 486 is valid from that either.

2) Ha! You caught me. AVID/Quantel.... It's my hobby. Buying those to capture CATV is a bit overkill. Alas, It is unlikely I'll ever own/use equipment like that.

3) Analog may be analog, but giving up on comparing is not useful. For example, if you take a 52.666 microsecond line and sample it at 13.5MHz, can you say it is 711 pixels? Also, If you say CATV is 330 TVL and DV is 540 TVL, can you say CATV is 330*4/3 = 440 (pixels) and DV is 720 (pixels) on a 4:3 screen?

Is there any valid way to compare resolutions and frame sizes of analog and digital and translate it to pixels?

I much appreciate your insight.


@SatStorm

Great info. Too bad it's hard to dig it out of posts from a year ago. It's great you still hang around here and help out. Thanks.

However, It is my understanding that analog capture devices sample at 1 rate and then resize the samples to whay you want, before it even gets to pixels let alone before it gets out of the hardware.

I say this because people with an ATI card are most likely getting 704 sample/pixels regardless of what size they ask. If they wanted 352, the chip resizes the samples to that. Same with 480 if that is an option. This resize is probably just as good as any software.

I wanted to say this for people who do not cap at full frame and then resize in software.

Anyway, before I turn of topic, my opinion is that if your source is analogue NTSC (TV transmissions for example), then CVD gonna look better SVCD overall. More smooth. And if you have a good TV set, then you might notice SVCD a bit sharper, if you encode it with CCE from a source, captured from an almost perfect TV transmission or LD.
With VHS sources, the "smoothness" of CVD is a plus, because it works as a soft filter of the analogue noise.

Great opinion:

For CATV SVCD has more resolution, CVD smoothes out noise. All depend on your source.
For VHS CVD = more smooth.

I'd like to add that VHS source has less luma resolution than CVD so no point in doing SVCD.

I figure VHS resolution this way:

TVL = (2 * active scan line in microseconds * BandWidth)/AR
VHS TVL = (2 * 52.666 * 3MHz) * 3/4 = 237

VHS = 237x486 TVL or 240x486 if you like the general number given.

To go back to pixels on a 4:3 screen (the way 352x480 CVD is measured), simply remove the AR adjustment. VHS in pixels is 237*4/3 = 316x486 pixels.

I'd really like to know if you or Cornucopia think the math is valid. I basically asked the same thing above by saying NTSC broadcast is 440x486 pixels.

Sorry if anyone thinks this is off topic, but I wanted to give some basis as to why SVCD vs CVD matters for NTSC broadcast where as for VHS it does not. The original topic is after all 'TV Resolution'


@energy80s

Ah yes. 480 is not DVD. Too bad for anyone who has or is making SVCDs. Also, too bad if you want to take DBV and put it on a DVD. It's a shame they didn't give us something in the middle like 480 or even 544.

[Heavensent]

svcd will look sharper than cvd (at the cost of more macroblocks)
if given the same bitrate.
the best you can do is capture at 720 X 480,
and if you dont care about moving it later to dvd than convert to svcd.
if the movie has high action, the better option will be cvd (given the same bitrate) because it will have less macroblocks
(at the cost of lost sharpness).

dont capture 480x480 no matter what anybody tells you.
try it out and you will see that the high resolution capture (720 X 480)
will allways make a better svcd / cvd at the end.

[trevlac]

svcd will look sharper than cvd (at the cost of more macroblocks)
if given the same bitrate.

I couldn't agree more. Why? Because if the source has more resolution, svcd supports more resolution. If it has less like VHS, no point in going higher. Also, if you want to remove noise, CVD may be better like statstorm said.

I'm beating the drum about this because if people (including myself) understand why, they can make the right decision in their case.

the best you can do is capture at 720 X 480,
...

dont capture 480x480 no matter what anybody tells you.
try it out and you will see that the high resolution capture (720 X 480)

Why? Well here I don't agree/disagree because you don't say why so people can make the right choice.

Here is some interesting things I believe based upon reading spec documents from various capture cards, and from testing. It especialy applies to you if you have a BT8x8 based card. Which you list in your profile. In the following, you could replace BT8x8 with all, but that may be too bold.

The following may be a bit off topic, but If the original poster was caping directly to mpeg, he is caping at 352 or 480. Not 720, so it applies.

BT8x8 cards pull the exact same amount of samples from the scan line regardless of what size you ask. They then crop/resize these samples based upon the driver and what you ask. They then give you pixels.

For example, my BT878a/BTwincap combo always samples and crops to 712 (13.5MHz) samples. IF I ask for anything other than 712 (say 480), It resizes to that value and gives me pixles.

So, If you think your resize is better than the chip/drivers, then a full 712 is better. For a specific case, 368 is much better than 352, because the chip applies a filter to get to anything less than 368. So I would never cap at 352. However, 480 does not seem to be any worse (for < 480 source) than 712.

So basically I'm saying a 720 cap has it's aspect off on your card, the card already did a resize, and it is probably no better than 480 for source with less than 480 res. 712 would be the correct aspect. 474 + pad would give you the correct aspect and svcd frame size.

BTW: These numbers are for the BTwincap driver. Other drivers do 688 not 712. I made a calculator when I did my testing. It uses a tool to read the register values from the BT878a chip after your app/driver sets them. You can then "calculate" what the chip was asked to do.

It's here: http://trevlac.us/BT8x8calc.htm

The spec docs for various chips can be found here: http://forum.doom9.org/showthread.php?s=&threadid=59013 The BT878a doc explains what the calculator is doing.

[Heavensent]

trevlac wrote :

svcd supports more resolution. If it has less like VHS, no point in going higher.

this is not allways true.
i cant explain why, but if you convert a VHS to dvd (at dvd resolution)
you will ALLWAYS get a better picture than if going with Svcd.

i think its the combination of the higher bitrate and the higher resolution.
this is more clearly seen on a big screen tv by the way.
and thats what proves my point.
if you can get BETTER qulity by increasing the resolution + bitrate
(it does come out better, and i think others will back me up here)
than 720X480 is not overkill for converting a vhs and you get the best
qulity you can.

i only mentioned the fact that its more clearly seen on a big screen tv
because THERE is a change in qulity, but with a regular tv you will need to pay much more attention to find it.
but it is still there.

anybody willing to back me up here ?

[lordsmurf]

anybody willing to back me up here ?

No.

Surpassing the source quality does nothing for it.

Feel free to put 30 minutes of 9.8 MB/s 720x480 from VHS source on one DVD. It's your DVD.

But I put 165 or so minutes on one DVD at 352x480 and it's still above the source's quality. Even playing it safe with higher bitrate, I still stick 100 or more on a disc. 352x480 is already above all broadcast, analog cable, some satellite and all low-end tapes (VHS/Beta/8mm).

[proxyx99]

Generally right, since encoding at higher res will not improve the originasl picture. But look at this from a different angle. MP3 encoding and scanning pictures face the same dilemma. This is about the objective fidelity of the copy vs. subjective judgment. Encoding 128 kbit (MP3 equivalent) quality in 320 kbits will always be closer to the original then encoding at 128 bit since, don't forget, you are introducing compression of the original material. More advanced compression scheme, better the copy (not the embedded source signal). In that sense same picture scanned at 300 dpi and 1200 dpi will contain same souce material but compression artifacting aill be lower on 1200 dpi. Thus this material will be closer to thie original at high magnification. For the same pixel of thie original you will have higher count of pixels in your copy. Blowing picture up will reveal less pixelation due to compression scheme. Same for video. Blowing up the same picture on a bigger screen will be of higher quality on 720x480 due, as I said, to less artifacting injected by a coping process.
I always prefer to copy at higher res, scan at higher res then source and encode at least 1 step higher then source. Distinction must be made between what's sufficient for making copy look OK, and trying to achieve higher level of fidelity. I prefer the latter

[SatStorm]

@Heavensent:
A possible explanation why 720 x 576/480 looks better 352 x 576/480 is because some DVD standalone players, simply don't do a good pan & scan job!
Other reason: A bad digital to analogue convertion inside the standalone!

Do a test to see the difference if you can. Burn a test DVD with all the known DVD picture resolutions. Then play this DVD on various standalone players. You'll be suprise to see how anything except 720/704 x 576/480 looks so different from standalone to standalone!

The way I see it, because only home users use resolutions different than the full CCIR 601, the DVD standalone manufactures don't try to make their products do the best possible job for any other resolution. They keep the compatibility because they have to, but only this.

[trevlac]

@Heavensent

I may seem all over the place here, I appologize.

But, I think in theory you are right that 720 VHS is better than 352 VHS. This is for the reason Proxyx99 gave. If I could restate them; lower quality anywhere in the process/system, introduces more error. This error is additive. AKA, a sharp representation of a fuzzy picture is better than a fuzzy representation of a fuzzy picture. (Depth of Modulation)

-------

I also, would never dispute your and proxyx99's real life results. These results could be due to many steps in your process, including the reasons given by proxyx99 or statstorm. I fully believe anyone who wants the 'right' answer and is not willing to test their system, is out of luck.

------

However, as a practical matter, CVD frame size can hold full resolution VHS, even if the resize process to get it there introduces error. So the statement "Use 352 for VHS" is a reasonable rule of thumb.

A CVD frame can NOT hold full resolution NTSC broadcast (4.2MHz). Regardless of how you get it there, it will not fit.

=============

I must say, this has been one of the most informative threads I have read. I especially appreciate the comments from people who have shared their real world results.

Thanks, and I hope c_hernandez32 can tell us what he sees.

[proxyx99]

Yes, I think good information has never hurt anyone. On the other hand this thread shows clearly how little details can hijack the main idea.
Nevertheless, I think that a common sense approach always wins. One should stick to the method that is least complicated and gives good results. Many are adovcating one format over the other. Sit back, times of svcd, vcd etc. are slowly over due to wide proliferation of DVD. Although considering minimal cost of making an svcd it's ideal for sharing family movies on a cd disc. (50 cents?, cheaper then postage).
I'm a strong believer that one shoud adopt the best possible format (like DVD) if it's not cost prohibitive. HD is just around the corner.What's next?

[FulciLives]

Lately for TV shows I've been doing captures at 704x480 which I then resize to 352x480 for MPEG-2 and I use a 2-pass VBR with a MIN of 2000kbps a MAX of 6000kbps and an AVG of 4000kbps

I convert the audio to 2.0 224kbps AC-3 audio and I'm able to fit 3 60 minute TV shows per DVD disc as long as I edit out the commercials.

The quality is most excellent this way.

In fact if I am doing a VHS and it is short enough that I can do a video bitrate of 5000kbps or higher I just do a CBR since Half D1 seems to max out at around 5000kbps

I only use 6000kbps as a MAX on 2-pass mode since I like the idea of my AVG (4000) being exactly in the middle of my MIN and MAX. Chock it up to OCD or some other "odd" psychological problem hehehe

- John "FulciLives" Coleman

[Heavensent]

lordsmurf wrote :

Feel free to put 30 minutes of 9.8 MB/s 720x480 from VHS source on one DVD. It's your DVD.

i didnt say i will do it using 9.8 MB/s.
i can allways use VBR to do it and still fit a movie on one dvdr.
but i understand what you are saying and if it works for you this way
- im happy.
and by the way, i did do it your way and still do it if the movie
has high action scenes, there is less chance of getting microblocks if i use
352x576/480.
i was just saying that its not the only way to do things, and to my eyes
a VHS converted into dvd 720x576 with 3 pass VBR using CCE does look better and sharper.

it still doesnt make it the right choice for everyone, and there is allways the qulity of the source to consider.

[lordsmurf]

some DVD standalone players, simply don't do a good pan & scan job! Other reason: A bad digital to analogue convertion inside the standalone!

This is so true. It is especially evident on PAL->NTSC and NTSC->PAL player conversions. Some players just don't have great MPEG chips.

So original shot source, original published source, recorded format, capture card, software, and finally player can all have outcome onto what the origial actor looked like or the original artist drew the cartoon (in terms of shape and size).