MPEG2 capture using ATI AIW 9600pro - settings

[videopete]

Hopefully a quick question:

To achieve excellent quality captures is it better to:

A) Capture using MMC 9.02 direct to mpeg2 using GOP structure 1-4-2

OR

B) Capture using MMC 9.02 direct to non-compliant mpeg2 I-frame only using incredibly high bit rate (20MB/s) and then use TMPGENC to re-encode to mpeg2 compliant (1-4-2 gop structure)???

I have done both but cannot tell which one yields better results on the final DVD.

[Nelson37]

There are some on this board who would tell you that method B is "Newby Advice". That is a quote, and they are wrong.

There can be good reasons for doing method B, however it is much more time-consuming. Been there, done that, and same results as you, I can no longer tell the difference. That kinda answers the question, doesn't it?

Since CPU around 1.4 to 1.7 (and above) and MMC 8.1 (and above), I find the real-time MPG caps to DVD-compliant files to be equivalent to re-encode of either a hi-bit, I-frame-only MPG or a Huffy capture. Analysis performed frame-by-frame on two different 19" PC monitors, also on 32" hi-def TV.

Availability of real-time IVTC has removed the last use for re-encoding for me. Have not done a re-encode in well over a year.

[ofbarea]

My captures done at 704x480, seems to be a bit better than my captures done at 720x480.

I have tested this with an AIW 8500DV and with a AIW 900Pro.

You should try it and use whatever resolutions that seems better.

[Nelson37]

704 does seem to offer a marginal improvement, also may be a bit steadier on the IVTC. Close call both ways.

Apparently, 704 and 720 both result in identical playback on standalones. I haven't really tested this extensively but somebody did recently, it's not just the overscan coverage, the player displays both the same. So I figure its that many bits unneeded in the encode.

What version MMC are you on and have you tried the 3:2 pulldown removal?

[ofbarea]

I´m using MMC 9.08. This is the highest version that works stable with both of the AIW cards that I have in my PC´s. I´ma also using Catalyst 6.5 Video drivers.

I have never attempted to use the 3:2 pulldown removal.

[lordsmurf]

There are some on this board who would tell you that method B is "Newby Advice". That is a quote, and they are wrong.

I have done both but cannot tell which one yields better results on the final DVD.

Since TMPGEnc is a poor encoder choice, it would be newbie advice. Quality gains potentially found by such a method would be 100% lost by the encoder, if you plan to use it for high compression. If Procoder, maybe even CCE or MainConcept, was mentioned, it would be okay.

But you have to ask, HOW MUCH benefit is gained? Honestly, not much. If you want to capture then encode, capture an AVI file. Videopete has already experienced this, he cannot see a difference.

I have never attempted to use the 3:2 pulldown removal.

Do not use 3:2 pulldown, that is bad advice. It leaves various temporal artifacts behind, and your source is interlaced. That option is useless to most people 99.99% of the time, it's not needed or suggested.

My captures done at 704x480, seems to be a bit better than my captures done at 720x480

704 does seem to offer a marginal improvement

This is because of the native chipset resolution being closer to 704 than to 720. The 720 requires a size change that incurs aliasing artifacts, something very common for that resolution on most cards, actually.

Apparently, 704 and 720 both result in identical playback on standalones.

Very true, they play back at proper AR, and quality looks fine on them both. It's just that the 704 is a bit better, especially noticed on computer screens.

[videopete]

I thought that it had to be either 352x480 OR 720x480 (NTSC) to be dvd-compliant - my final goal is to produce a dvd-r with my home videos. Would 704x480 work on this final as a final product? I personally don't think it would although I've never attempted it either.

Also, I've heard that TMPGEnc is one of the best encoders around?
Would the ATI mpeg-2 direct capture yield equivalent results? If so I'll just do it this way to save time. Any special settings I should pay particular attention to? Thanks again!!

[lordsmurf]

704x480 is one of many compliant DVD specs. It will be fine, unless you're using some crappy authoring software that does not use the whole DVD spec (TDA is fine).

TMPGEnc is far from best encoder, in terms of raw encoder quality. It's a great program when you need lots of filters (several of which work better than AVISynth or VirtualDub filters, in some circumstances), and when you need to feed it all kinds of source.

But the MPEG-2 files it makes are pretty crappy compared to Procoder or other commercial software. It requires a lot more bitrate to look close to the others.

Yes, the ATI MMC MPEG-2 capture would get equivalent results. Follow the guides at www.digitalfaq.com for ATI MMC tips.

[jman98]

Far be it from me to argue with lordsmurf. While do I think CCE and Procoder and Main Concept are your best choices, TMPGenc has improved the quality of its MPEG-2 encodes by a lot recently. If you've only checked out older versions of TMPGenc, you might be surprised.

I have to give lordsmurf kudos for saying that 3:2 pulldown is useless to most people 99% of the time. I totally agree. I have found it to be way more trouble than it's worth, but those who use it do so with a religious fervor and it's impossible to tell them otherwise.

[Nelson37]

The "newby adivce" quote I am referring to mentioned specifically the hi-bitrate, I frame capture for the purpose of re-encoding. NO SPECIFIC ENCODER was mentioned. This type of capture has at least two specific advantages over AVI capture, namely filesize and avoidance of dropped frames. Whether or not this process involving re-encode is superior to real-time capture with good parameters is a question I think is already answered.

The ATI IVTC, on a clean source such as Digital Cable, or HDTV channel thru S-Video, works at least as well as any software method I have tried, including DeComb and Telecide. The advantages of IVTC are well documented, though I would be happy to explain these if you fail to understand them. The numbers of people who have such a clean source available for capture are certainly greater than 1%. The sources I am referring to are not Interlaced, they are Telecined.

Now, "bad advice" would constitute recommending to someone that they try a procedure, and in particular a procedure that is inherently flawed in some way. No flaw in the procedure, and no recommendation, either. Just a simple question of has it been tried.

[videopete]

The main guide on www.digitalfaq.com concerning ATI Mpeg2 captures only uses 352x... in the examples. Even the movie capture only uses 352x480 ! If this resolution is good for full length movies, why wouldn't it work for Hi8mm home videos?? Why would I need to use 704x480 for Hi8 - it doesn't make sense to me. Can someone help and explain this to me. Thanks in advance.

Also, there is no guide which explains the settings for capturing in 704x480 mode - what bitrate to use (8M - 8.5M seems to yeild decent results). What settings would I use in the ATI MMC 9.06 (which is my version by the way) to yeild the best quality I can get from my Hi8mm home videos?? Can you please include GOP structure as well - any difference between using 1-4-2 and 1-2-2 ?? I personally can't see any but it could be me.

[videopete]

Nelson37 --> what settings exactly are you using? Since you're using ATI capture card and MMC software - I'd like to know how you are running it....would definitely help me...thanks.

[Nelson37]

That 352x480 capture would definitely qualify as Bad Advice. There is almost no situation where this would be recommended, unless quality does not matter to you.

Native res of ATI card much closer to 704x480. Outside of re-encoding, this is your best resolution unless you really need a smaller file at the expense of quality.

Leave the GOP settings as is, see the "What is DVD" section.

Bitrate is determined by length of movie, also by type. ATI filesizes are quite variable even using identical settings. I use an AVG of anywhere from 6.5 to 9.0, usually more like 7.5 to 8.5. Set Max from .5 to 1.0 higher than average. Using a wider range seems to lead to pixelation.

Sound usually at 224, been trying 384 but not sure if any real difference. Sports or talk shows can go down to 128 or even 96 on audio to save bitrate.

Darker movies will give a lower filesize. Brightly lit, hi-action movies will be larger.

If your home videos were shot using a tripod, you should be OK. If not, you may want to go for shorter videos on disk and max out the bitrate. If you have not used a tripod, strongly recommend you get one. Absolutely amazing what this will do for you.

[videopete]

Thanks for your input. Given this, it seems that the guide on capturing Mpeg using ati aiw cards is incorrect as it only mentions 352x480 resolution as a max. even for high quality movies? What I'm hearing here is that this is not enough resolution for a high quality capture - we need to go to 704x480 min to obtain high quality.

[Nelson37]

Yes, that is one among numerous errors you will find there. The site is primarily geared to converting VHS tapes, that resolution suggestion is based on being closest to VHS actual resolution. This is based on something called the Nyquist Theorem, which, while interesting, simply does not apply to this situation.

What Smurf is only lately beginning to admit, after several years, is that matching the ATI capture res to the source res is physically impossible. The card captures at a Fixed Resolution, then performs a Resize to achieve the specified resolution. This is why the Nyquist Theorem does not apply. The Fixed Resolution is closer to 704x480 than any other valid DVD resolution.
I believe it is something like 672x448, but it really does not matter.

Most people who have carefully checked their captures, and wish to preserve best quality, prefer the higher resolution capture.

The next thing you will need to do is to adjust the color, brightness, and tint so that the captured video, When Played Back On The Desired Playback Device, most closely matches the original video. Smurf's "advice" on using the Gamma adjustment is totally useless unless you are using the PC for playback. Gamma adjustment does not affect the captured video.

Frankly, if you want accurate, fact-checked, peer-reviewed information I would avoid that site and use the guides located at videohelp. The only piece of information he ever had that I wanted was never made public, I was told I would have to pay for it, and I have Never seen anyone refer to successful use of this technique, which I believe was only "discovered" by accident and he was never able to duplicate. Nothing on that site I haven't already seen somewhere else.

FYI, he will almost certainly be on here to call me an *******, while offering absolutely no factual information to contradict what I am saying. Perhaps for his third time he may actually come up with something useful, most likely just revert to more juvenile name-calling.

[davideck]

Nelson37 -

The Nyquist Theorem supports your claims. It simply defines a proportional relationship between a given resolution and the minimum samples_per_line required to accurately preserve it. Capturing or resizing down to 352 is factually insufficient for any source beyond 240 horizontal lines of resolution.

[lordsmurf]

FYI, he will almost certainly be on here to call me an *******,

Finally, you say something I won't disagree with.


VIDEOPETE........

These are called "guides". They give you a push in the right direction. While examples may show 352x480, you can equally use 720x480 or 704x480 just fine.

Anyway, 352x480 is more than adequate for most conventional home sources, like VHS, 8mm video, off-air tv broadcasts, satellite and cable. All this rambling about Nyquist theorem (amongst others) sounds nice on paper, but with noisy consumer sources (even "clean" ones like digital satellite and digital cable), and non-pro hardware and software available to us for (anything less than a 5-digit price tag), those theoretical statements are basically pissed away. In practice, the 352x480 will look every bit as clean and clear as your medium-resolution home sources.

Concerning ATI MMC, gamma correction is one of several proc amp functions that works just fine on a system with no errors. The problem, unfortunately, is some people have bugs that prevent everything from working perfectly. When I had Windows ME and Windows 2000, I had no problems. With Windows XP, the gamma control no longer works for me. Oh well, I have a hardware proc amp now anyway, not a bother. I don't know if this is a byproduct of XP, or something else, as I've changed my hardware a few times (RAM, CPU, audio card, etc). Really nothing I cared about.

If you're truly concerned about quality, capture uncompressed AVI at 704x480 and then post-capture encode in a good encoder like Procoder. More money, more time, and probably even more effect than it's worth.

You already compared a few things and saw no differences. Go with your best judgment on what you want to do. Forget about all the crap you're reading here. Guides are here for some help, the forum is here to anser some questions when you're confused, but at the end of the day, if you want to succeed at video, you have to get your hands dirty. Run tests on 5 minutes of footage every way you can think of, and then see which one you like more. Then do that.

[Nelson37]

Whatsa matter, Smurf, can't remember your own bullshit anymore?

It was you who clearly stated that capturing hi-bitrate, I-frame only MPG for re-encoding was "newby advice". Changed your mind, have you?

You have also told multiple poor newbies to "just follow the guides at Digital Faq". Now, they are just general suggestions, not an accurate way to go?

Why you can't get the Gamma control to work is no surprise, though I have never had that problem. However, it has NEVER had any effect on captures, it is ONLY for adjustment of the PC display. Of course, an "expert" such as yourself would never advise someone adjusting the proc amp controls for capture to "learn to use the Gamma control", now would you?

Go pimp your own crap someplace else, play your power games elsewhere. BTW, your story of your "de-modification" doesn't seem to agree with anyone else's.

Home satellite and cable sources often surpass 352x480. This res is not sufficent to capture with good quality. Someone used to 20 yr old VHS might find it acceptable, but many have specifically mentioned testing the two and find 704x480 far superior. Of course, careful testing using A-B methodology and multiple viewers is unnecessary for Smurf, ain't it?

Davideck, you are essentially correct, though there is debate about what the Nyquist theorem actually says in terms of where the oversampling limit lies. However, the Theorem deals with what res a given image Should Be captured at. Once you have captured a given image at a certain resolution (the fixed res of the ATI card) it says Absolutely Nothing about what res that image should be resized to. 704x480 being the closest DVD-compatible number to the fixed res, it therefore represents the least manipulation of the image.

[davideck]

Nelson37 -

I agree that a captured image can be subsequently resized as desired, but Nyquist does imply that resizing (downsampling) a 704 capture to 352 (for example) requires the same low pass filtering that would be necessary to capture at 352 originally. In either case, the resulting image will be bandwidth limited if properly filtered, or may contain alias components if not.

Resizing up does not require additional bandwidth reduction and may therefore provide a better result.

[Scorpion King]

I've tried Smurf's preset examples and have long ago deleted and forgot them as I figure you will do also. While Smurf was creating and running his website the rest of us was capturing video. It's no wonder he now finds himself behing the times and loosing his prestige. His ideas and information is very basic and out of date, to say the least, let alone incomplete as he is now finding out. I sorta feel sorry for Smurf but knowing him like I feel I do, I acutally don't.

Learning to capture video is a lot of fun and can keep you busy for hours on end. I find it more challenging to capture/encode at the lower bitrates that the higher rates. Trying different GOP structures, experimenting with resolutions, encoding programs, etc, can be fun and rewarding and is far better than setting in front of a computer screen reading this junk all the time like some do. Use this information when you feel you need to, but experience is the best teacher. IMHO

Get busy and capture some video. The only rules is to be DVD compliant, that's it....

[edDV]

Wow just got back from Saturday night downtown pub crawl to find the fight is here.

There is alot of good newbie and product specific advice on http://digitalfaq.com/. No reason to nitpick until you can suggest specfic improvements or create a similar guide.

I think the advantage of these forums is to pool advice to the benefit of all. Consumer video has moved rapidly from VHS crap to ~'90 broadcast standard in a short time. HDTV takes the consumer PQ bar to equal or better than ATSC broadcast.

This is good eh?

Later. I'll go sober up now.

[lordsmurf]

There is alot of good newbie and product specific advice on http://digitalfaq.com/. No reason to nitpick until you can suggest specfic improvements or create a similar guide..

That would be too hard edDV. It's so much easier for some people to talk trash and pretend to have knowledge. Meanwhile the rest of us (folks like you and me) earnestly try to help others as best we can, with guides, tests and replying to posts on forums such as this one. "Put up or shut up" is a nice cliche, but sadly people like that do neither.

Resizing up does not require additional bandwidth reduction and may therefore provide a better result.

Davideck, resizing video captures up can add just as much aliasing (sometimes even MORE aliasing) than resizing down. Your theory may say otherwise, but practical application shows otherwise. All resizing, in any direction, can result in aliasing, it's the nature of the beast.

At any rate, some people (not many, just a few nitpickers) get far too caught up in the chipset size and theories. The higher goal is resolution retention of the source, and quite simply, 352x480 is able to maintain the image on most sources. If your capture chipset sucks (BT8x8, for example), then there may be problems. But ATI AIW Theatre chipsets are fine, as are a number of others.

I've tried Smurf's preset examples and have long ago deleted and forgot them as I figure you will do also.

Guides and examples are there for learning. I don't know why people think you have to follow every step, afraid to deviate. That's called "being a newbie". While there are some steps that need not be skipped (do not use 3:2 pulldown in ATI MMC, for example), tweaking the bitrate, GOP size and resolution to your source needs is both suggested and encouraged.

[Nelson37]

Has Smurf EVER detailed a correct procedure for using the ATI proc amp controls? No, his answer is "just buy a $200 external box like I have". This is not needed, it is already there and free. I have detailed this procedure, on this board, several times. His lame response keeps referring to a "bug" in the Gamma control. For the umpteenth time, the Gamma control has absolutely no effect on captured files and is completly useless except for PC playback.

Has he ever stated ANY information about using the different aspect ratio captures available on the ATI card? No. All he does is capture using antiquated sources like VHS. Why does this "ATI Expert" never tell anyone how to use the "non-fixed" aspect ratio preset? He probably has no idea. Usage is this - Set AR to Letterbox. Then change to non-fixed. You will get a perfectly cropped widescreen cap for which you just need to set the AR flag. NO black bars in the capture. May be necessary to use full 720 width, I am still experimenting with this. This is very useful for cap to re-encode to XVID, for instance.

3:2 Pulldown? Apparently he has never done this using software methods on an AVI cap. Otherwise, he would know that the real-time MPG cap using MMC is very comparable. Also, he would know that it is very sensitive to bitrate, source quality, and resolution. Specifically, at his recommended res of 352x480, it does not work very well. Using the 3:2 Pulldown removal on S-video feed from an HD channel results in a PERFECT 23.976 capture with substantial reduction in filesize, allowing for higher specified bitrate. MMC versions prior to 9.x gave a variable framerate of approx 24.0x. I have put this info out, on this board, many times. To just flat out state "never use this" is quite simply bad advice. I guess you should just experiment within Smurf-approved parameters.

Where is his information on the variability of captured file sizes due to MMC's lack of strict adherence to specified bitrate? I have detailed this, on this board, many times. Where is his information on AVG vs MAX gap size and its effect on capture quality? Same answer.

He has on three or four seperate occasions responded to specific technical criticisms of his methods with NO factual response, simply to call me an *******, though usually indirectly. Great comeback, perhaps my 8-yr-old could learn from him for the schoolyard.

For how many years did he continue to put forth the false information that 352 caps are best for VHS because they most closely match source res, apparently unaware of the fixed capture res of the ATI card? Now he argues that it is correct because the ATI card has a good resizer.
Bullshit. Most people who have done the tests seem to agree that 720 caps just simply look better. Guess they're just nitpickers, darn those who prefer quality.

I have checked his site numerous times. Not once have I found a single piece of useful info that was not published somewhere else, and long ago at that. He crowed about how he had been able to get MMC 8.x working on a Rage 128 card, but has never published his method. The excuse was it was just too complicated. You have got to be kidding me. I would bet money he was either flat out lying, or got it to work ONCE and was never able to duplicate the result. Rage3d has published similar instructions many times, on a site not dedicated to capture.

Talk trash? How many times have I called Smurf an *******? Give you a hint, the FIRST one is coming.

Pretend to have knowledge? Give me ONE instance where I have been factually incorrect, not just where my opinion differs from Smurf's. I started capping MPGs with an ATI card he didn't even know existed. I learned about MPG encoding by Individually Testing every single setting in TMPGenc, creating and viewing HUNDREDS of test files.

Now, Smurf, perhaps if you got your head out of the way that finger that can't seem to point directly at an ******* would find one right nearby.

[lordsmurf]

Nelson, you have no idea what I do or what I know, so just shut up already. You assume far too many things, and you're almost always wrong when you do.

[Nelson37]

Same to ya, though I have cited numerous, specific examples. I have asked you to do the same, and you have nothing to offer, except name-calling.

I have stated where you are wrong and specifically described how, and why.

You have called my suggestions "Newby Advice" and then turn right around and recommend the same thing.

IMO you are a know-nothing dumbass interested only in self-promotion, with an axe to grind.

Several times you have personally attacked me and my suggestions with absolutely no alternative whatsoever, useful, workable, or otherwise.

If I am so often wrong, NAME ONE, AND PROVE IT.

[Cobra]

Chill out, everyone. Let's answer the original question with facts, and debate them in a civilised manner. Save anything else for PMs.

Thanks!

Cobra

[davideck]

Davideck, resizing video captures up can add just as much aliasing (sometimes even MORE aliasing) than resizing down. Your theory may say otherwise, but practical application shows otherwise. All resizing, in any direction, can result in aliasing, it's the nature of the beast.

At any rate, some people (not many, just a few nitpickers) get far too caught up in the chipset size and theories. The higher goal is resolution retention of the source, and quite simply, 352x480 is able to maintain the image on most sources.

I agree that the filtering artifacts introduced by resizing can be just as noticeable in either direction. However, Resizing Down also requires additional bandwidth reduction which further reduces the preservable resolution. There is no escaping the fundamental constraints of sampling. Claiming that 352 sampling accurately preserves resolution beyond 240 lines is fundamentally flawed.

It's not my theory. The credit belongs to Nyquist and Shannon. It is one thing to ignore (or insult) those on this forum who have done critical A/B testing and have noticed the shortcomings of 352 sampling in their own practical application. It's another thing to discount fundamental theory.

If others notice something that you don't, that does not imply that it doesn't exist. It just means that you don't see it. Your sources, your setup, your display devices, and/or your eyes must be deceiving you.

[Nelson37]

It is also important to note the degree of resize in each direction. 704x480 is extremely close to the fixed capture res of the ATI card and represents a minimal change. 352x480 represents a reduction in width of almost 50%.

The only reason I have examined the theory is that it was cited, often, as the reason for capping VHS at 352x480. I had already determined that 720(704)x480 looked better, both to my eyes and several others.

Davideck, I was unaware that Nyquist addressed resizing. My reading of it was primarily where it addressed the "correct" amount of oversampling for capture of an image. My rejection of it is based on the fixed nature of ATI capture card resolution. What exact numbers the theorem recommends for capture are therefore irrelevant, since the fixed capture res cannot be changed.

Once we bring resizing into the equation, we must then discuss quality of available resize methods. Almost everyone who resizes has a favorite method, knows of several, and often uses different methods for different material. Very few would accept being locked into a single method, with no selectable parameters, and no opportunity to re-do the process using another method.

I have been rejecting the VHS cap at 352x480 for over two years now. This was at first based on simple observation that 720x480 looked better to me, to several testers I personally observed, and to several posters on this board. Upon learning that the supposed theoretical basis for this fallacy simply did not apply, I am continually astonished at the repeated insistence of many people to maintain support for the 352 VHS cap.

[lordsmurf]

The thing about Nyquist-Shannon is that you're referring to mathematical computations. Whether or not this has any practical application to video and audio is argued both ways. More often than not, sampling arguments within a reasonable scale (of which 352x480 is part of, as well as 704x480 and 720x480) are imagined more than reality.

And more often still, when people perceive quality changes, it's not actually a sampling issue, but rather a statement to the quality of the product.

I've seen several calls for "facts", but apparently nobody here is debating facts, we're debating application and validity of theory, which in and of itself, is not a fact either. We're all using facts, and in fact, the exact same facts. But our analysis (or agreement with somebody else's analysis) is what differs.

[Nelson37]

To me, the facts lead to an inescapable conclusion and indicate that the theory does not apply.

1. ATI (and almost all other) cards capture at a fixed resolution. This cannot be changed.

2. This resolution is in the ballpark of 672x468. Whatever the exact number, it is not DVD-compatible and the closest compatible res is 704(720) x 480.

3. To achieve any other res than the card's fixed res, you must resize. This must occur after the image has already been digitized.

4. 704 x 480 involves a very minimal resize from the fixed capture resolution. 352 x 480 is a major resize, by comparison.

5. Any resize dictated by the specified capture resolution is performed in real-time, with no parameters, by the card's hardware and/or software.

6. Your starting point is not whatever res your source happens to have. Your starting point is the number given in #2.

7. Any mathematics which attempt to indicate best capture res for a source of any given res is completely irrelevant due to #1.

8. All mathematics and examples I have seen indicate it is best to resize as little as possible to reach a DVD-compatible res.

Multiple testers have been checked in proper blind testing config. "What do you think of Video A, what do you think of video B". Two test movies on one disk, 32" HD tube from Composite DVD output. Zero out of 8 to 10 have said lower res cap was better, 3 to 4 had no preference, 3 to 4 had preference for higher res cap, 1 or 2 said higher res cap was significantly better.