What is the Maximum Capture Bitrate i should use ??

[c627627]

1. So your answer regarding Inverse 3:2 Pulldown and Progressive Source is "it depends?"

Generally speaking, when capturing standard TV signal, we should
• Encode Interlaced,
• UNCHECK: Inverse 3:2 Pulldown

So I'm going to go ahead and continue to change just one setting when the source is an HD TV signal (downgraded through S-Video cable) and that is to Select • Progresive Source instead of • Encode Interlaced.

I'm going to go ahead as a general rule to continue to • UNCHECK: Inverse 3:2 Pulldown for both standard and high definition TV signal.

2. Regarding encoders, since the most recent version of MMC is 9.16, available here
http://ati.amd.com/support/drivers/common/mmc9-16.html
that's where I install everything from except I install a lower version of MMC because it has no time shifting.

I think MMC 9.03 is all right, lordsmurf believes we should go even lower, to MMC 8.7. (That one is not available though, MMC 8.8 is.)

Anyway, ATi Multimedia Center for me does not work unless I install ATi software in this order:
1. ATi Drivers
2. Remote Wonder software
3. Data Access Objects (DAO - MDAC)
4. DVD Decoder
5. Encoder Package
6. ATi Multimedia Center

I may skip 2. and 4. but I install Catalyst 5.12 and not any version after that.

And I install
3. DAO - MDAC from the MMC 9.16 package and
5. Encoder Package from the MMC 9.16 package

[manono]

If you're going to correct something of mine, Nelson37, at least correct something that's wrong.

This removal, "remove 3:2 pulldown", Inverse Telecine, or IVTC, can be accomplished by checking the box in MMC.

So, why is the sample marked as "withpulldown.mpg" still 29.97fps, and still have all the dupe frames. And by the way, what he needs is not a full IVTC, but just a dupe frame removal. An IVTC is a 2-stage process usually. The first is field matching to restore the progressive frames, followed by the dupe frame removal. A full IVTC will just slow the process. It won't necessarily hurt anything, though, but could be why checking the box in this instance didn't work. Or maybe the reason is that he used an S-Video cable. I don't know. Like I said, I don't cap, but I do know a good deal about IVTC.

You also select Progressive source as this is what you have AFTER the IVTC is performed.

That may or may not be true, but doesn't that sound a bit counter-intuitive to you? A Source is before filtering. It seems to me that if you intend to perform an IVTC, since the source is, by definition, interlaced, you'd check Interlaced source. But like I said, I don't cap.

This procedure should NOT be performed on a true interlaced source.

A telecined source is a true interlaced source. You said it yourself earlier:

...shows up as either interlacing ONLY on the fourth and fifth frames...

Now, of course you don't perform it on a source shot on video - shot with interlaced 30fps cameras. But for a 24->30fps movie such as we have here, shot on film, we certainly do restore it to its original 23.976fps progressive state. Maybe using an S-Video cable is what converted it from true telecined video to 29.97fps progressive, but what he capped there is very peculiar. Perhaps it was deinterlaced and made progressive by tossing one field and resizing down from the 1080i source.

For 480 line video: if you find that the pattern in your video is 3 progressive (non-interlaced) frames followed by 2 interlaced frames, then your video has been telecined

http://neuron2.net/LVG/telecining1.html#identifying

Having used Avisynth with at least three different IVTC filters, and a few other programs, I would say the MMC is at least as good as most.

Again, why didn't checking the box work in this case, then? I would not only not say that MMC is "at least as good as most", but that in this case it's pretty damned near useless. And my experience with AviSynth IVTCs on many difficult sources is very different than yours.

[Nelson37]

A true interlaced video is interlaced on EVERY frame.

Telecining is different. It is interlaced on ONLY two of five frames. It is a hybrid. I do not refer to this type of video as Interlaced, IMO it is not. Standard de-interlacing tools do not work correctly. It is a special case. There is Interlaced, there is Progressive, and there is Telecined. Threee different types.

REPEAT - The ONLY case where IVTC should be used is with a TELECINED video. IVTC is indeed a two-stage process, with the dup frame removal being the last step, sometimes a different filter. If MMC seperated the two, then the decimate would be all you need for a vid showing duplicate fourth and fifth frames. BUT, MMC does not seperate these functions. The best choice is not there. So you use what is available.

REPEAT - the source of the video DOES NOT MATTER. Only the TYPE of video.

The choices you can make on a real-time MPG cap with MMC are limited. There are only certain things you can do. You can choose progressive, or not. You can choose the IVTC, or not. You do not have a choice as to which order these operations are run. In fact, I have no idea what order they are run, I have no control over this. If the source is Telecined, and the IVTC is run before encoding, then you are encoding a Progressive video. If the IVTC is run after encoding, then you end up with a Progressive video. The Source Type being referred to is listed under De-Interlacing Options. What exactly this means to ATI is unknown, I do not guess, I examine results.

As for what the examples do or do not show, I did not create them, I have no idea.

It is like the Crop and Pad versus Mask. I see people making guesses with no definite information. The Crop does actually Crop, with certain other settings. With other settings, the result is the same as a Mask. It could be a crop plus a pad, it could be a true mask. How the hell anyone but an ATI programmer could definitively state that it is one or the other escapes me. Since the result is the same, and you have no user control, why anyone would waste 3 seconds debating this makes no sense.

Just recaptured Attack of the Clones, with IVTC. NO dupe frames, NO interlacing in ANY frame. Framerate displayed by multiple progs as 23.976. Note that some will interpret the Pulldown flag and show a 29 framerate. GuiforDVDAuthor, for instance. Vdub and most others correctly report the 23.976 framerate.

I have many dozens of videos captured using this procedure, with good results. Some videos retain the dup frames, some do not. I have not found any particular setting within MMC which seems to affect this reliably.

As for "it depends", since I do not have detailed notes over the last 14 versions of MMC on which I have used this, and since I absolutely recall that a variety of settings on a variety of versions had significant effect on functionality of the IVTC, that is all I can tell you. The only consistency I can guarantee is that it will work, with variable degrees of effectiveness, in every version since 7.6 or so. With 9.03, the framerate got to exactly 23.976, previous versions consistently gave 24.1x. The older ones gave editing problems, the newer ones are much more straightforward to edit.

Just a thought, I very rarely use CBR and seem to recall it had a negative effect on the IVTC.

What I do with each MMC, is run a series of test captures from both SD, HD, and DVD sources, changing settings one at a time, with naming conventions to so indicate, then do frame-by-frame analysis of each test cap with Vdub. The crop buttons will change how they work, MPEG-2 and MPEG-DVD and VCR filetypes change settings, bitrate settings behave differently, video adjustments behave differently.

If you think that making ONE encoding test with ONE file using MMC proves anything at all about what the program can or cannot do, that is pretty much your problem right there.

Also looking for logical or intuitive explanations is largely worthless. It is what it is, it does what it does. Very little of the inner workings can be changed in any way, it is a black box. Shove something in one side, flip a switch, and carefully examine what comes out the other. Then try another switch, and repeat. When you get what you want, stop. If not, keep testing.

Also, if you want to bitch about a "correction", point out EXACTLY what you are referring to, and what it is that indicates my statement is incorrect.

[manono]

Also, if you want to bitch about a "correction", point out EXACTLY what you are referring to, and what it is that indicates my statement is incorrect.

I did. I quoted what you said, and then pointed out how it was wrong.

Telecining is different. It is interlaced on ONLY two of five frames. It is a hybrid. I do not refer to this type of video as Interlaced, IMO it is not. Standard de-interlacing tools do not work correctly. It is a special case. There is Interlaced, there is Progressive, and there is Telecined. Threee different types.

You can make up new definitions all you want. That doesn't make them true. Telecined video is by any definition (except maybe your novel one) interlaced video. It has interlaced frames, and it's encoded as interlaced (i. e., the progressive flag isn't set). Just open one of your telecined caps or a hard telecined DVD in DGIndex and run the Preview. It'll say Interlaced NTSC. It sure as heck won't say Progressive Film or even Progressive NTSC. It is not a new, or different, or special case. It is definitely not a hybrid. A hybrid consists of 2 or more different base framerates. Telecined film has just one - 23.976fps.

The source was originally broadcast at 1080i, wasn't it? Just what do you think the "i" means? By your definition, shouldn't it be called 1080t?

What about a 20fps silent film, "telecined" up to 29.97fps. Is that a 4th type? If 2 of every 3 frames are interlaced, that's not an interlaced source? What about a PAL2NTSC conversion, converted from 25fps to 29.97fps by field blending? Is that a 5th type? If 5 of every 6 frames are interlaced, that's not an interlaced source?

As for what the examples do or do not show, I did not create them, I have no idea.

But weren't you the one that insisted on c627627 using that Inverse 3:2 Pulldown setting:

Yes, absolutely for an HD source.

I'll retire from this one, as I don't cap and can't intelligently help c627627 with his problems. I'll just repeat that c627627 will want to remove the dupe frames, and using AviSynth's Decimate or (better) TDecimate filters is the best way to go about it. I also think he should somehow cap the telecined source as telecined, rather than already deinterlaced (if indeed, that's the case). Maybe that means using a component or DVI or HDMI cable, rather than the S-Video cable. Maybe that means getting a better capture card. What's the point in having a Hi-Def source, if you can't cap it in all its glory?

[Nelson37]

I believe I clearly stated that it does not always work. I said it can work, and it does. I stated that one attempt that did not work is an exceptionally poor proof of anything.

At least two others who have taken the time to experiment and test this feature have also reported success. Several who have called the feature worthless have also demonstrated an amazing lack of accurate information about how the ATI card and its accompanying software actually work.

My caps with IVTC read as true progressive, NTSC film, 23.976 fps, same as most original DVD files. Not all process correctly, not all will edit correctly, not all will author correctly. Some familiarity with the capturing process will inform most people that there a variety of filetypes to be encountered, and a number of error conditions that are beyond control, such as broadcast irregularities, power fluctuations, weather conditions, sunspots, etc.

Capturing a true HD source from Satellite is not currently feasable with the ATI card, nor any other that I am aware of. Capturing an HD feed thru the SD cable is without question the highest quality capture source currently available for an AIW card.

Telecined film does NOT, repeat NOT, have a framerate of 23.976, it it 29 fps. That is what the telecining DOES. The incoming source we are dealing with does not allow for removing the pulldown flag, as would a file. It is what would be called Hard Telecined if it were a file.

We could debate what the actual framerate of a 24fps, true progressive file with pulldown flags embedded actually is. Such a file is not telecined untli it plays. The output device recieves a 29 fps signal, unless you set for progressive, in which case no telecining occurs. The source we are dealing with is 29 fps as received. The software methods usually used for such a source file CAN NOT BE APPLIED. The reason for this is very simple, NONE of them will work in real-time while capturing. There is only ONE available choice to attempt this, the IVTC in the MMC program.

If you believe Interlaced and Telecined is the same thing, fine and dandy. They are processed differently, should be handled differently, come from different framerate sources. One can be returned to its natural, higher quality, lower framerate, original state. The other can not. You could call them Bob and Joe, They Are Different.

Take your HD box, ignore the indicator, and plug in the S-Video cable to something. Your statement that the indicator, or the original broadcast state of the video as received and as sent over the HD cable, has anything to do whatsoever with the output of the S-Video connection suggests you have never tried this. Therefore I agree that you have nothing intelligent to offer.

[c627627]

Wow you guys are duking it out, aren't you?

There's nothing personal about this, so just ignore the other guy's personal comments, so we can all learn something here.


The question about what's the point of having "a Hi-Def source, if you can't cap it in all its glory" is a good point and it may help you both understand people like me, you know, most people who just want to use a blank DVD disc to do what they used to do with VHS tapes, period.

Now you know full well high definition burners cost over a grand a pop, regular folk won't be using that for a while so the "what's the point" question projects your professional understanding and use onto average Joe Sixpacks.


I hooked up my cable box to my AGP video card / capture card ATi 9600 XT and I figured out how to make decent captures and burn them to DVDs. No more VHS for me, which is great, that's where I'm coming from.

Once in a blue moon, I see something that's nice on one of the HD channels that I wouldn't mind taping. All I was asking is: should I change up on any of the capture settings I use to capture Standard TV?


YES. One or two things. So: Leave everything unchanged except:

Select • Progresive Source instead of • Encode Interlaced, right?


The second possible change was whether to leave • UNCHECK: Inverse 3:2 Pulldown setting or not?

First of all, before posting the complicated stuff, you should say something along the lines

a. leave it unchecked
b. change it to checked
c. it depedns on ... so leave it unchecked if ... or change it to checked if ...

Now if you could fill those ... first and then talk about the other stuff that confuses someone who's just looking to fill in those ...

that would be great.


I can only use S-Video or the worse option of the yellow RCA video cable so DVI and HDMI for capturing is right out for me. I am, however interested in hearing about this 'Decimate' process once I figure out the best way to capture an HD signal if I need to (rarely) to burn to DVD instead of to a VHS tape.

[edDV]

...
Telecined film does NOT, repeat NOT, have a framerate of 23.976, it it 29 fps. That is what the telecining DOES. The incoming source we are dealing with does not allow for removing the pulldown flag, as would a file. It is what would be called Hard Telecined if it were a file.

We could debate what the actual framerate of a 24fps, true progressive file with pulldown flags embedded actually is. Such a file is not telecined untli it plays. The output device recieves a 29 fps signal, unless you set for progressive, in which case no telecining occurs. The source we are dealing with is 29 fps as received. The software methods usually used for such a source file CAN NOT BE APPLIED. The reason for this is very simple, NONE of them will work in real-time while capturing. There is only ONE available choice to attempt this, the IVTC in the MMC program.

If you believe Interlaced and Telecined is the same thing, fine and dandy. They are processed differently, should be handled differently, come from different framerate sources. One can be returned to its natural, higher quality, lower framerate, original state. The other can not. You could call them Bob and Joe, They Are Different.

Take your HD box, ignore the indicator, and plug in the S-Video cable to something. Your statement that the indicator, or the original broadcast state of the video as received and as sent over the HD cable, has anything to do whatsoever with the output of the S-Video connection suggests you have never tried this. Therefore I agree that you have nothing intelligent to offer.

I've been monitoring this thread to better understand how the ATI cards capture with inverse telecine. Although I have an AIW-8500 and a Theatrix 550, I've only use these for "utility" captures. But I am all ears if MMC is able to do an adequate IVTC.

So with that as the goal, I can help sort the tech details on live capture sources. First a quick review of what is telecine and how do progressive TV's and DVD players work read this. http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm

Live 480i/1080i does NOT carry a telecine flag nor is their any attempt by program editors or TV stations to maintain a continuous cadence. You have a higher probability of continuous cadence when tuning a movie channel that doesn't insert commercials. The telecine field sequence must be detected by the capture hardware or software before it can perform an IVTC from 29.97 to 23.976 fps. The main issue is how the cadence detector performs when presented with a break in cadence, a non-telecined clip or a field reversal (bad edit). These can cause a multi-frame to multi-second disturbance while the detector resyncs and the jibberish fields created are near impossible to fix later. This is why those serious about capture acquire the 29.97 interlace source and deal with IVTC later, or they just record the telecined 29.97 fields to DVD as 480i so the progressive DVD player or Progressive TV can do the IVTC in hardware.

Now for types of sources, analog NTSC is likely to have been digitized multiple times before reaching the local TV station and some of these steps may have deinterlaced without IVTC thereby destroying the the mixed fields. Same goes for the SD digital playout server at the TV station. Live "network" broadcasts are more likely to have a quality cadence. Cable movie channels are another likely source.

For HDTV, most channels use a mix of pure interlace and telecined 1080i/29.97 as do most cable/sat channels. Only ABC, FOX and ESPN are reliably 720p progressive but 80% of that is converted 480i/1080i. Maybe they got the IVTC right at their end, maybe they didn't.

Cable/Sat HD boxes are another issue. Most output only 1080i/480i or only 720p/480p regardless of what comes in. Therefore if you select 720p or 480p, the box does the deinterlace and may or may not do a good IVTC. For 1080i, you get either a passthrough or you get 480i upconversion or 720p->1080i conversion.

At the S-Video output, you get analog NTSC (analog TV station interlace NTSC or MPeg2 to analog NTSC interlace or 480p to NTSC or downconverted 1080i or 720p->480i->NTSC). Main point, only some of this will have entact telecine. Asking the capture card to IVTC on the fly is asking for trouble unless you have confidence in the source.

The hardware IVTC ("cinema") processor or deinterlacer in the quality progressive DVD player or HDTV are designed to quickly adapt to to anything tossed at it. ATI MMC software probabaly isn't likely to adapt well except maybe from a trusted source.

For these reasons, I like to capture to 480i (ADVC-100 or ATI) to DV-AVI or interlaced MPeg2 without worrying about telcine. I leave the telecine for the progressive DVD player or HDTV to process. Penalty is a 20% larger file.

[edDV]

...
The question about what's the point of having "a Hi-Def source, if you can't cap it in all its glory" is a good point and it may help you both understand people like me, you know, most people who just want to use a blank DVD disc to do what they used to do with VHS tapes, period.

Basic rule is quality in gives quality out. Downscaled 480i from a ATSC tuner/HD cable or Sat box is the highest quality home input available for a 480i/480p DVD. The issues above mainly boil down to 480i or 480p DVD from a 480i source.

I hooked up my cable box to my AGP video card / capture card ATi 9600 XT and I figured out how to make decent captures and burn them to DVDs. No more VHS for me, which is great, that's where I'm coming from.

Once in a blue moon, I see something that's nice on one of the HD channels that I wouldn't mind taping. All I was asking is: should I change up on any of the capture settings I use to capture Standard TV?

Then you need to choose between 480i or attempt 480p for capture and DVD authoring. My choice is stay 480i. If you go for IVTC (for 480p) you may gain a 20% smaller file at the risk of unfixable errors.

YES. One or two things. So: Leave everything unchanged except:

Select • Progresive Source instead of • Encode Interlaced, right?


The second possible change was whether to leave • UNCHECK: Inverse 3:2 Pulldown setting or not?
...

My advice is leave it interlace and live with 20% larger files.

[c627627]

Wait, you're not talking about the 3:2 Pulldown setting, you're talking about the "Progressive Source" setting?

No, that can't be, when I tried capturing HD signals with both "Encode Interlaced" and "Progressive Source" I've got to tell you that "Progressive Source" looked better, so something's wrong there. Can you address that?


Complicating things even further, let me ask you this: I capture from my cable box which is an HD cable box so it outputs 1080i HD signal, right?

But for non HD, it's set to output 480p, you know, the usual.


So it just occurred to me: Does that mean I should really be using the "Progressive Source" setting for all captures, not just when capturing HD channels?

[edDV]

Wait, you're not talking about the 3:2 Pulldown setting, you're talking about the "Progressive Source" setting?

No, that can't be, when I tried capturing HD signals with both "Encode Interlaced" and "Progressive Source" I've got to tell you that "Progressive Source" looked better, so something's wrong there. Can you address that?

...
So it just occurred to me: Does that mean I should really be using the "Progressive Source" setting for all captures, not just when capturing HD channels?

You said you are using the NTSC (S-Video or Composite) outputs? Those are always 480i (interlace).

If you have your box set to 480p and you are capturing from analog component or HDMI, Then the box is deinterlacing and you need to capture progressive. The NTSC outputs are still sending interlace NTSC.

If you are capturing interlace analog NTSC (S-Video or composite), then inverse telecine is a method to convert a pure telecine cadence into 480p/23.976. If the cadence isn't pure, you generate jibberish fields. If you go this way, you need to carefully monitor the resulting 480p/23.976 stream. The problem is, if you find an error, you need to recapture. Get out the listings

If you capture 480i, you can try inverse telecine later and still have the original capture file to use for fixing errors. It is a labor intensive process. Or you can skip inverse telecine and let the progressive DVD player or HDTV process it.

Complicating things even further, let me ask you this: I capture from my cable box which is an HD cable box so it outputs 1080i HD signal, right?

But for non HD, it's set to output 480p, you know, the usual.

I have mine set that way for my CRT HDTV. If the TV was a 1366x768 LCD, I'd have it set for 720p/480p for a medium-low end model (cable box does IVTC/deinterlace) or 1080i/480i if I thought the TV could do better than the cable box (i.e premium processing).

Either way, the NTSC output is always 480i interlace regardless of other settings.

[DereX888]

My simple easy to remember "rules of thumb" that alwas work for me in past many years for DVD-Videos:

if WAV sound - video max = 8500kbps
if compressed sound - video max = 9600kbps

By compressed I mean AC3 or MP2 of 448 or 384 kbps.
If your capture card is really accurate (i.e. many cards DO exceed the bitrates set in their settings while capturing anyways) then you may really calculate exact bitrates according to selected audio bitrates.

[c627627]

All right. Facts:

1. Location: USA
2. Time Warner Cable DVR box set to output 480p standard and 1080p HD signals (480i and 720p are not being output, they've been deselected.)
3. S-Video cable being used.

Is this right or wrong:

If the channel on the box is a regular old SD channel:
• Encode Interlaced
• UNCHECK: Inverse 3:2 Pulldown
• UNCHECK: Closed Group Of Pictures

If the channel on the box is a a High Definition channel:
• Encode Progressive Source
• UNCHECK: Inverse 3:2 Pulldown
• UNCHECK: Closed Group Of Pictures


You see if I have just TWO scenarios, SD or HD channel and I have just TWO options, Encode Interlaced or Progressive Source, what's the problem in telling me which option to select in either of only two scenarios.

Same goes for selecting or deselecting Inverse Pulldown in each of the two scenarios.

The answers may have been posted, but they were drowned in additional information that confused me.


Don't be upset now, just try to answer which of the options I should choose given the facts of what is being used.

[edDV]

You just said you are using the S-Video cable for capture which is always interlace NTSC no matter how you set the YPbPr and HDMI outputs. The interlace NTSC can have telecine for film sources or be normal NTSC (29.97 frames / 59.94 fields per second)

You cannot output progressive on the S-Video output.

If the interlace output contains film telecine and the cadence is correct, it is possible to perform an IVTC. This is risky to do during capture unless you know the source is pure.

[c627627]

OK. All right. You are saying I should use • Encode Interlaced no matter what?

What is the "Progressive Source" option there for then, when the only video inputs are S-Video and the even worse composite yellow RCA.

[c627627]

Tell me what you see:

Progressive Source:
Progressive Source.mpg

vs.

Encode Interlaced:
Encode Interlaced.mpg

[c627627]

And can you help with this one too:

For highest quality short capture from NTSC VHS, I understand I should not capture using full resolution of 720 x 480 and should instead use 352 x 240, is that correct?

Reason being that VHS video is roughly 320 x 240 pixels?


EDIT: This is incorrect: Capturing 720 x 480 is better for VHS as well. Except it's better to use 704 x 480 for capturing TV/VHS signals.

[Use 704 x 480 to capture SD or HD TV, VCR, analog camcorder.
720 x 480 is used by commercial DVD discs and DV format camcorders.]

[Nelson37]

Oh my bleeping Bleep no. Delete the VHS question ASAP, you will have more crap flying at you than you can believe. Do a search, do a test, don't add that can of worms to what you already have. At least make a new post.

For case 2 and 3 below, when destination and playback path can display progressive, the progressive option appears to give best quality. For case 1, rare for movies on the chosen source, Interlaced is best.

Cap some AVI from various sources types on the HD feed. Load these in Vdub. Find a section of horizontal movement. Find the five frame pattern. Depending on the source you will see one of three things:

1. Five frames all showing combing pattern at movement. This is INTERLACED. Encode interlaced, no pulldown, do not mess with it, essentially it cannot be improved for TV playback. Usually not seen on HD channels for anything longer than one hour.

2. From hi-quality and particularly S-Video feed from a hi-def source, you may see a repeating pattern of 3 clean, artifact-free frames followed by two that appear to be interlaced. This has been TELECINED. Unlike interlaced material, this can be improved in quality at the same time bitrate is used more efficiently by checking the 3:2 pulldown box.

If you are not familiar with the absolutely dramatic improvement in quality an IVTC can provide, if true progressive material does not interest you, then do not bother with the 3:2 pulldown removal.

If the source is not a theatrical movie, if it is edited, cut with commercials, forget the IVTC. Ed's comments about discontinuities in the stream causing problems are absolutely correct. Noticeable pauses, jerks, or stutters in the video can result. Compared to software telecines from sources like pre-digital cable premium channels, IMO the MMC IVTC used on the HD feed we are talking about can give superior results. In real-time.

3. Same sources as #2, you may see 3 clean frames, then a fourth, then a fifth that is an exact duplicate of the fourth. This is what your examples show. This apparently falls under the definition of telecined as well, but is not normally used. Had never seen this until HD source became available. Seems to be more common on HBO then HDNET, but I am pretty sure I have seen all three types on each of the two.

As to whether this can be removed by MMC, actually I honestly do not know. I have no way of knowing what the source was before I processed it. The frame duplication was less common, and this result is sometimes seen when I check the box and still get 29fps. Extreme jerkiness shows too many interlaced frames with some duplicates, a result of the bad patterns Ed mentioned. However, both types of failure are quite rare when capping theatrical movies from the HD channel, clear weather and all that.

I would have to do an AVI cap and then an MPG with IVTC, of the same broadcast. I would think that the same movie would be the same every time.

For type number three, the quality and bitrate savings of the IVTC are significantly lessened. Balance this by the fact that the removal of the duplicate frame can be done with absolutely no artifacts or loss of quality, assuming that all-important continuity of pattern.

All I can tell you is that it works, almost all the time. Fairly rarely the finished product will show two or three almost subliminal quirks. Occassionally an obvious slight pause or stutter. Compared to careful observation of the original broadcast, Perhaps the trend would have to be slightly higher than average but certainly not out of the normal range. While allowing for significantly higher bitrate and better PQ.

From experimenting with the IVTC when the frame counter was on, it seems to take a few seconds for it to stabilize, starting a cap when the ratings and Dolby screens are on is ideal. If you are testing by doing 10-second captures during a running movie, you may not see optimum performance.

[c627627]

I don't plan on doing it but if I hooked up my VCR to my capture card, "what capture resolution should I choose" is not worth a 1 second "this x that" answer? Come on man, let's talk about video capturing, I post at overclockers and when a total newb comes to post, I welcome him, assume he doesn't know anything and answer all his noob questions and let the moderators deal with personal attacks, so how about helping me out and answering my VHS question.

As for the rest, thanks, I'll go through the extensive process you described to determine an answer.

You know, these are 30-50 minute documentary and comedy shows I'm capturing, not 2 hour films by the way.


I know a lot more than when I started and I'll know more still in the future. I come back to my old posts and add edits to them with two sentence direct answers, it can't always be done but direct noob questions can have answers with two sentences directly addressing them first before expanding on to things more commonly done by experienced users.

[manono]

Your encodeinterlaced.mpg is finally an example of what you should have been doing all along. You must have changed a setting somewhere to have gotten that. I assume you finally used some sort of interlaced setting, perhaps an Encode Interlaced setting. I was surprised, because if the source was really 1080i, then the card (or your cable box, or something) did a good job of resizing, at the same time not destroying the field structure.

Now, using whatever settings you used to create that one, you can also then apply the Inverse 3:2 Pulldown setting to see if it does a decent job of returning it to 23.976fps progressive. Either results (29.97fps interlaced as in the encodeinterlaced.mpg, or 23.976fps progressive after applying IVTC) will be better than what you were showing us before (29.97fps progressive with the duplicate frames). And, of course, returning it to 23.976fps progressive, if possible, will be the best result.

If the Inverse 3:2 Pulldown works, as Nelson37 believes it should, then so much the better. You said something about this not being a movie, but a docu? If so, then parts of it may be pure 29.97fps interlaced (all frames interlaced). If true, then you don't want to IVTC any of it, but keep the whole thing as 29.97fps interlaced.

Earlier you had said, "Encode Interlaced' for Standard TV source or 'Progressive Source' for HD signal source". That seems to imply that you believed all Hi-Def sources to be progressive. I hope by now you've learned that's not the case. In spite of what Nelson37 seems to think, this encodeinterlaced.mpeg is interlaced - hard telecined and interlaced.

Here it is reencoded with IVTC applied, as 23.976fps progressive, and then with pulldown to output 29.97fps. I also changed those grey bars to black, and redistributed the black to both sides. This is DVD compliant and authored to DVD. So are the other samples you've provided DVD compliant, the encodeinterlaced.mpg being the best. But in my opinion, the grey bars are kind of ugly, and the CBR encoding is a waste of bits. Cappers sometimes have different opinions about such things, but with good VBR encoding, you could easily stick 2 of these 50 minute docus on a DVD5, and not just one with a bitrate of 8000 CBR. Too bad this thing is only 4:3, and not 16:9.

progressive.vob

[edDV]

You know, these are 30-50 minute documentary and comedy shows I'm capturing, not 2 hour films by the way.

Futher to Nelson37's three types of video coming from the S-video from HD cable box

#1 True interlace (no telecine).
This includes all news, sports and everything shot with a TV camera. Even if the original was shot progressive (e.g. 720p/59.94) it will be interlace 29.97 at the S-Video port.

#2 Interlace with telecine.
Film and progressive video shot at 23.976 fps will have the 2:3 telecine sequence. Most likely to have continuous cadence will be movies aired without breaks. Most recent TV series intended for international distribution will have been shot 23.976 fps but most are still edited at 29.97 or 25 fps. Older TV series are peppered with non-telecine 29.97 clips any time video special effects were used. Most "special effects" clips were used to fix the shot and are not obvious.

#3 Duplicated 4th frame.
This isn't common. It results in studdered motion. The reason "telecine" mixes fields for frames 4 and 5 is to smooth the motion. Watch the waterfall in your interlace clip at 1x speed.

Documentaries can be straight interlace (i.e. shot with a normal TV camera) or interlace with telecine if shot 23.976 (i.e. shot on film). Comedy shows could be either. Situation comedies are usually shot on film or 24p HDTV. Concerts and variety comedy are usually shot 29.97 interlace.

I don't plan on doing it but if I hooked up my VCR to my capture card, "what capture resolution should I choose" is not worth a 1 second "this x that" answer? Come on man, let's talk about video capturing,

You asked for "best quality". DVD has limited choice of resolutions. 320x240 isn't one of them.
https://www.videohelp.com/dvd

352x240 single field results in half motion increments and half vertical resolution. 352x480 is adequate but not "best quality". That leaves 704x480 as the "best quality" choice. 720x480 will add 8 black pixels left and right.

[DereX888]

Aren't you guys tired of relpying to same old same old? I am.
I thought we have search on the forums.

[Nelson37]

That's why I recommended he use it.

The debate has gone on for YEARS. Some will say to cap at 352x480 as that is the valid res closest to the res of the video tape. The understanding that almost all video cards have a fixed capture res that is closer to 720x480 than any other DVD valid resolution, and that the card will do an internal resize for ANY DVD-valid resolution, is a complex one. IMO 720x480 is the correct res. There are MULTIPLE PAGES of discussion of this subject, primarily because there is not a "1-second answer."

The result from VHS captures is so inferior to even digital source SD content, and so absolutely butt-ugly compared to captures of HD source feeds, that to me it is not worth the price of a blank. YMMV

On the capping, for documentaries or anything that is NOT a movie, do NOT check the remove pulldown button. I thought I stated this.

For theatrical movies, try it. I like it. You may not.

[Nelson37]

Did not see the new clips before.

The EncodeDeInterlaced.mpg is a perfect example of a standard, Telecined source on which the IVTC can be run. Or it can be recorded as-is and playback normally. The name of the file is confusing as it does NOT appear to be DeInterlaced.

The WithoutPulldown.mpg looks like it was run thru a deinterlace but NOT an IVTC. This is bad, this procedure should not be followed.

[Nelson37]

Just ran a few tests with 9.12, got some answers.

Source movie from HDNM, HD channel, S-Video out to AIW card.

3:2 pulldown with progressive source results in a progressive, 23.976 clip, using zigzag scanning, and an average bitrate approx 2/3 of the standard clip. Framecuts from such a file, as well as moving clips, are as close to DVD quality as an AIW is going to get.

Using Interlaced source with no pulldown results in a 29.970 clip and uses alternate scanning. Obvious telecining artifacts clearly visible. Much larger file. 44K for one minute versus 28K for the IVTC clip.

Setting progressive source and NOT checking 3:2 pulldown results in a telecined video with the interlaced frames somewhat blurred. The progressive frames may or may not show a slight improvement. Zigzag scanning was used.

Zigzag is recommended for progressive, alternate for interlaced. This may account for the perceived improvement from using progressive scanning alone. The blurring effect may be less objectionable on a PC than the original clear interlacing.


Bitrate variation in any ATI VBR file can be quite dramatic. With bitrate set to 11.5 max and 11.0 average, a 24-minute clip had ONE peak at 11.2, two peaks at 9.8, and one at 7.1. Average bitrate for the file was just under 4.0. Quantizer scale almost completely flat, except for most, but not all, of the 4 largest peaks in bitrate. 2 or 3 areas dipped under 2.0 for bitrate.

[edDV]

Is the IVTC adaptive? What happens when it is set to IVTC and a standard interlace program or clip is encountered? Is it smart enough to switch modes?

[Nelson37]

I'm pretty sure it will deinterlace and leave a 29fps clip alone, I basically use it for movies only.

A few tests on HDNET 45-minute documentaries or series do seem to be clearly telecined, I have not examined in detail.

As far as changes in type during the same recording, I have no way of identifying the source after the encode, and again usually use it only for continous movies. I know that TNT, for instance, uses unGodly bastardized patterns that with IVTC on will jerk all over the place. Numerous attempts at software de-interlace of these sources also failed, perhaps not quite so badly.

I know I have had it on for standard interlaced video, and the output was watchable but not ideal. Dr. Phil on SD with commercials, other similar shows.

It takes several seconds to start the IVTC when the movie is already running. An intermediate blending stage seems to happen. Framerate counter would indicate 29 for the first 2-3 seconds in earlier versions. I usually start the cap a few seconds before the fanfare.

Have done 2 or 3 caps of repeated broadcast source with errors, using IVTC on and off. Ran detailed comparison on both files in Vdub, after burning both to DVD and watching on TV. Changes in pattern for just one or two frames could result in a duplicated frame or single interlaced frame, which during playback generally caused a slight stutter, almost subliminal. Sometimes by pure chance the extra frame was earlier in sequence resulting in a very
noticeable stutter. Pretty much the same thing I had seen from DeComb or TeleCide filters in AviSynth. Some of these errors were visible in the original broadcast, though on average they were made more obvious by the IVTC. Some better, some worse.

The HD sources are much cleaner in pattern variation, faster PC seems to improve the performance. Editing, FPS detection, and audio synch varies among programs and has definitely changed with MMC 9.03.

[c627627]

1. For capturing to transfer to DVD to watch on TV, is my choice of 720x480 resolution over 704x480 correct?

2. I was able to figure out that I was correct in selecting "Progressive Source" and NOT "Encode Interlaced" when the source is an HD channel coming through an S-Video cable as you can see, Encode Interlaced has every few frames that look like crap:
https://forum.videohelp.com/images/guides/p1662894/encodingcomparison.jpg

Thanks to manono who told me exactly how to do this: get VirtualDubMod (not VirtualDub which can't load .mpg files) then right click on pic, zoom and use arrow keys to advance the frames one by one to see what they look like.


Of course "Encode Interlaced" should be used when capturing standard TV channels but not High Definition channels when "Progressive Source" should be selected.


So now how about you tell me exactly:
a. how do you tell by examining that De-Interlace should not be used.
b. how do you compare captures with Inverse 3:2 Pulldown CHECKED vs. UNCHECKED?

I could not tell the difference for De-Interlace and Inverse Pulldown. Is it because I would need to view them on standard old TV screen to tell the difference?


I am looking for answers that would use this template (if the question was how do you tell the difference between Encode Interlaced and Progressive Source, the one sentence answer would be):
Get VirtualDubMod (not VirtualDub which can't load .mpg files) then load .mpg and right click on pic, zoom and use arrow keys to advance the frames one by one to see what they look like.

[edDV]

1. For capturing to transfer to DVD to watch on TV, is my choice of 720x480 resolution over 704x480 correct?

Try it and see with your MMC. 704x480 is what is broadcast. Every capture card that I've used for S-Video in gives 8 pixel black stripes left and right at 720x480 like this:

The black stripes won't be visible on a TV due to overscan but will be visible on the computer. If you don't want the black stripes on the sides, capture 704x480.

Of course "Encode Interlaced" should be used when capturing standard TV channels but not High Definition channels when "Progressive Source" should be selected.

Nothing to do with High definition channels vs. SD channels. You select "progressive" and "Inverse 3:2 Pulldown" only when you are confident the source is telecine with continuous cadence. If you are wrong you will get the kind of errors Nelson37 mentioned above and these cannot be corrected later. Non-film programs (sports, talk, variety, news, anything shot with a TV camera) will not have telecine and will be captured incorrectly with those settings. In those cases always capture interlace with "Inverse 3:2 Pulldown" unchecked.

The safer course is to always capture interlace, review the captured file for telecine field sequence (shown below) and then apply IVTC as a second step. If the IVTC is unsuccessful, you still have the original telecine capture.

Here is what a telecine sequence looks like frame by frame over 5 frames. You will see 3 progressive frames then two mix field frames. After IVTC you should have four progressive frames if you get it right.








Note the split fields


Note the split fields

So now how about you tell me exactly:
a. how do you tell by examining that De-Interlace should not be used.
b. how do you compare captures with Inverse 3:2 Pulldown CHECKED vs. UNCHECKED?

I could not tell the difference for De-Interlace and Inverse Pulldown. Is it because I would need to view them on standard old TV screen to tell the difference?

I am looking for answers that would use this template (if the question was how do you tell the difference between Encode Interlaced and Progressive Source, the one sentence answer would be):
Get VirtualDubMod (not VirtualDub which can't load .mpg files) then load .mpg and right click on pic, zoom and use arrow keys to advance the frames one by one to see what they look like.

The discussion here is should you attempt inverse telecine during capture. If you guess correctly that telecine is present and continuous, the resulting file should show entact progressive frames at a 23.976 frame rate. If your source does not have continuous cadence, you will see split field errors in your capture. Virtualdubmod is a good tool for frame by frame examination.

[Nelson37]

There is almost NEVER a reason to de-interlace.

Using IVTC on a true interlace source is similar to de-interlace.

Using Progressive without IVTC seems similar to a blend De-interlace. Not generally recommended. May offer some improvement for the 3 progressive frames, at the cost of blurring the two interlaced frames. This may appear better on the PC, on TV perhaps a minor difference.

On a telecined source, you need to set BOTH IVTC and progressive source for proper operation.

The Interlacing you are seeing and describing as bad should NOT be visible on a television. It is ONLY visible on a PC screen, as these are natively prograssive.

If you are going to select Progressive source, IMO you should select IVTC as well.

You are falling into a very common newbie trap of viewing interlacing on the PC and wanting to get rid of it. Not necessary for TV viewing, and degrades the video. True interlaced video is composed of seperate fields with no matching fields present. This is because each is from a different moment in time. The only way to get rid of it is to throw away half the resolution.

Telecined video is different. The two frames which show interlacing contain fields for which there IS a matching field from the same moment in time. Telecined video can be restored to full-resolution video with NO interlacing artifacts, fully progressive with only four frames where there were five. Fewer frames, frames that are easier to encode, frames with fewer motion artifacts, frames where a better quality scanning method can be utilized.

Let me offer you an example. Get or find a friend with a Progressive scan DVD player and TV. Watch a DVD in both standard and progressive mode. You should see a significant improvement, that is why people pay extra for such equipment. This is what you are trying to achieve with the capture IVTC.

ED, you are suggesting allowing the hardware to do the IVTC. I was under the impression that progressive DVD players simply ignored the pulldown flag and were unable to do an IVTC on hard telecined material, such as the capture source we are discussing. Is this something only available on higher-end units, with Faroudja chips or something similar? Or can ALL progressive scan players do a true IVTC, including my collection of WalMart specials?

[Nelson37]

On the 704x480, I tend to use the 720 and 704 to mean the same thing, and they are not. 704 is actually generally preferrable, but functionally there is not much difference.

Using Vdub as you described is IMO the best testing method.

You may want to read up a bit on fields and frames, interlacing and telecining. The IVTC process is not itself complex, but the streams it runs on may well be, as Ed has described. The process has been described as something of a Holy Grail for encoding. It is the ONLY filter which, if it works, is GUARANTEED to improve quality, NOT make it worse, while at the same time allowing for a smaller file or increased bitrate for the same size, AND increasing encoder efficiency and reducing artifacts. Win, Win, WIN, WIN, WIN.

Other than the ATI MMC, SFAIK there is no other method for doing this in real-time while capturing in MPG format. Add-in captions converted to subtitles, a decent menu, and a printed disk, the result can be mistaken for a Hollywood product.