VHS conversion problem....video is jerky

[briandugas]

I am having issues with captured VHS. I am following the advice of www.lordsmurf.com, and capturing at 352x480 at 4.0 Mb/s VBR. My playback is jerky in the horizontal field. When things move left to right, they seem to be jerky. Vertical movements are OK. Is this because I'm capturing at full vertical resolution, but not full horizontal resolution? I'm trying to learn how DVD's are squeezed horizontally and unsqueezed on playback. I don't know if this is a factor or not. The video plays fine on my computer. I only notice this when playing on a set-top dvd player. Please help. Thanks.
-Brian

[SatStorm]

Typical mistake:
inverse the field order while you encoding to mpeg 2.
For most analogue captures, it is Field A
For most DV it is Field B.
TMPGenc has it as default to Field B first. Change that to Field A first and see the results.

[fmctm1sw]

What SatStorm said and I'll add that telecined material does the same thing for me when I play it back. Is it a movie? Have you tried IVTC?

[briandugas]

Gosh, thanks for the quick replies. Very generous. I haven't run the video throgh any programs. I captured the VHS to my computer in the 342x480 MPEG2 format, and went straight to TMPGEnc DVD author with it. I'm guessing I need to use the regular TMPGEnc to take care of the field order? I'm using an ATI all-in-wonder 128 pro 32 mb ram, which is relatively old, so I can only run MMC 7.1 I believe. There is no option for Fields. As for IVTC, I'm not sure what that is you are referring to. Is that Inverse Tecline? I've only heard of that. Any good pointers on some good readings abotu whatever IVTC is? My only real custom settings on my capture beyond what I have set, is the P and B frames. I'm capturing at 4P and 2B. I set the motion estimate quality to 100. Don't know if that's my problem. Thanks.
-Brian

[fmctm1sw]

AVIsynth is the best way to determine field order (IMHO). Look up some guides on Inverse Telecine (IVTC) or check out www.lukesvideo.com He has some good info about interlacing, telecine, and other stuff there...

Edit: And I'll assume you meant 352x480 not 342....

[briandugas]

Yes, you are correct to assum 352x480, I goofed. I'll check out that info you posted. I'll report back if I can't get it worked out. Thanks for the help.

[bottle-necked]

Gosh, thanks for the quick replies. Very generous. I haven't run the video throgh any programs. I captured the VHS to my computer in the 342x480 MPEG2 format, and went straight to TMPGEnc DVD author with it. I'm guessing I need to use the regular TMPGEnc to take care of the field order? I'm using an ATI all-in-wonder 128 pro 32 mb ram, which is relatively old, so I can only run MMC 7.1 I believe. There is no option for Fields. As for IVTC, I'm not sure what that is you are referring to. Is that Inverse Tecline? I've only heard of that. Any good pointers on some good readings abotu whatever IVTC is? My only real custom settings on my capture beyond what I have set, is the P and B frames. I'm capturing at 4P and 2B. I set the motion estimate quality to 100. Don't know if that's my problem. Thanks.
-Brian

Reading along in this post I was thinking you were capturing with a Ulead product until I saw the words "ATI all-in-wonder 128 pro 32" then I know what your problem is. I had one (my first ATI card) and I had this same problem and figured it out. If I can only remember the proper settings. You may need to experiment a little but I'm right in theory. ok

In the MMC setup for your capture template there is are 3 check boxes on one of the screens at the bottom. I can't remember that version of MMC exactly but one is like 3:2 pulldown, Deinterlace, and the third was ...uh... (something) I can't remember. If you check both of the two bottom boxes you get like field order a, if you only check 1 you get interlaced, if you don't check either you get field order B (or something like that). I know it sounds wierd (it was wierd) but that's the way mine was. Just experiment with those 2 bottom check boxes and you will figure it out. That was the answer for mine.

I checked the ATI site and the newest version of MMC for that 128 pro I think is 7.2. lol

One last suggestion for that 128. Download and try the trial version of Ulead VideoStudio v7. You will probably need to set it to capture field "B" but it will capture interlaced video and do a better job for you than MMC 7.2.

Hope this helps.
Good luck.

[briandugas]

I had a question on Field A / Field B. Do I need control over my ATI All-in-wonder to make it capture Field A first? or is it something that is taken care of after in TMPGEnc? I was looking around in TMPGEnc and found the drop down box for Field order, but I don't have an option in my ATI MMC control area. I can see how my horizontal jerkyness problem may be a result of the wrong field first. But I did some reading on inverse telecine, and I can see how that may be my problem too. My question on that subject is: If I do a reverse telecine, doesn't that put the movie back to a 24fps like it was originally created? What good will a 24fps source do me?
Thanks.

-Brian

[FOO]

You should fix the field order first . IVTC is a whole other
can of worms and cannot be done on some video .

[briandugas]

OK. My program doesn't tell me which field order it is capturing in. Do I assume it's backwards since I have this jerkey motion? Are there programs which allow to capture in the proper field order so that no conversion is needed? Thanks.
-Brian

[FOO]

It should capture to MPEG2 correctly , otherwise what's the
point in capturing to MPEG2 if you have to reverse it ?
Mebbe you are dropping frames.


Restream will flip the field order of a MPEG2 Elementary stream
which means you demux , flip it and remux .
Try that and see if it fixes the problem

[briandugas]

Well, I tried frameserving with AVIsynth, no luck.
I tried swapping fields with Restream, no luck.
I tried hanging garlic around my neck, no luck.

I downloaded Ulead Visual Studio, and I was able to set the capture field to Field B, and finially, I got some results. No more jerky left-to-right motion. I know everywhere I read said that analoge sources should go with Field A, but for some reason, Field B is working for me, so I'll go with it. Thanks for all the suggestions.
-Brian

[FOO]

Ulead has it backwards . Field B is top field first

[briandugas]

Well, then that would mean that my ATI MMC software is capturing Field B first, which was causing my problems. Glad I got it worked out though.

[SatStorm]

And then people blame me, why I dislike the (older) ATI cards....

[Bstansbury]

FOO, someone has this wrong. If you look into the Settings>Advanced Tab, you will see they say the Field order: "Top field first (field A).

Soooo who has it right. When I transfer DV via Ulead VideoStudio and save it in DV format (.avi), it comes in as Field A. Here is the funny news,.... when I process this through TMPGEnc, even though I set it for field A, after it comes out in SVCD mpeg format, it says it is field B. Somewhere there is an error here????

I have a separate question. From what I can find out, the experts say if you really want the best quality mpeg compression, you have to get rid of the interlace artifacts before you compress in MPEG. On this web site there is a complete discussion on de-interlacing and 4+ ways to get it done. However, it does not talk about which programs are good at doing this. VirtualDup has a "Deinterlace MAP filter" and there are other filters like "Deinterlace-smooth, etc. My question is, who has good experience on using deinterlace filters or software on Digital Video from Camcorders. If you read the article "What is Delintelacing" on this Web site it mentions that DV is not the same as interlace analog TV. There is a time base difference. Will the standard filters work on DV data.

Thanks in advance for you help

[FOO]

All I know is that if I import video into Ulead software ( MF2 )
It better say field B or it ends up backwards when playing on TV

DV does come in bottom first

[lordsmurf]

You need to be using ATI MMC 7.7 or higher.

The field order thing is a flaw of older software, especially older ATI MMC's and some ULEAD products.

Welcome to video. And all it's quirks.

[Bstansbury]

I just got some more information from Jerry Jones, who is the WEB site technical support for Ulead's WEB forum. He points out that Field selection is "Device Dependant" and that the standard for Firewire DV input is "Field A"

FOO, when you say "Import", is this done via file transfer, DV firewire, S-video, ???. What ever it is it must require Field B order.

I did not mean to change the orginal focus of this post, therefore I've added a new post on "Deinterlacing" for DV .avi in VirtualDub.

Lordsmurf; You have a lot of expertice. Maybe you could give me a suggestion on how to deinterlace an .avi file (Type-1). VirtualDub can't handle the audio in a Type-1 DV file.

Bud

[briandugas]

[deleted to start new topic]

[Bstansbury]

Before this thread completely dies,... It would really help to clear things up.

In a post near the top, SatStorm said

Typical mistake:
inverse the field order while you encoding to mpeg 2.
For most analogue captures, it is Field A
For most DV it is Field B.
TMPGenc has it as default to Field B first. Change that to Field A first and see the results.

Later on FOO indicates;

Ulead has it backwards . Field B is top field first

and;

All I know is that if I import video into Ulead software ( MF2 )
It better say field B or it ends up backwards when playing on TV

DV does come in bottom first

_+_+_+_+_+_+_

All of this is confusing because there 4 things mentioned;

Field A
Field B
Top Field first
Bottom Field first

In TMPEGen, which I have heard no one say is wrong, in the settings panel and under Advanced, there are two options;

Top Field first (Field A)
Bottom Field first (Field b)

It seems someone is wrong??? Foo, if TMPGEN is correct, there is no such thing as "Field B is Top field first"

SatStorm,... Your point was that analog should be Field A, and DV should be Field B. This seems OK as when I tried to push DV through TMPgen in Field A, it reversed it. It did this twice on me and I made sure I set the fields correctly.

This means the info I got from the Ulead web site was wrong and that DV in not Field A. Funny thing is, as soon as I turn my DV camcorder on and load anything into Videostudio,... it is field A. Lordsmurf says this is wrong. Could be since this is VideoStudio ver 6.02 which I got back in 2002, and updated in 2003.

I guess the only way to know for sure, since everyone has diffenent equipment, software, etc. , is to try it. When I played the MPEG out of TMPGen (which was Field b) on my DVD, there was no jerky video.

Hope someone can help with a answer.

[FOO]

It gets worse than that

there are
top field - bottom field
upper field - lower field
first field - second field

odd field - even field
field 0 - field 1
field A - field B

.. and probably some more

Odd/even sucks because nobody knows whether we start at 0 or 1
A/B sucks period

I think some AVI formats don't have provisions for marking fields
which means there is NO way for a program to find out without
looking , which is hard. Note TMPGenc likes to say
"judging field order" when it starts.

Sometimes you even have to worry about which field pair go together
--- you can flip field order 2 ways - swap every pair -- or just skip the first

The MF2 I have , definitely thinks fileld B is top field

[SatStorm]

There are also broken drivers, which the field order is a matter of luck!

If the drivers are OK, then it is Top field first for analogue captures and Bottom field first for DV transfers.

But if the drivers are not okey or the hardware follows some manufacture rules (which happens and they are different), things turn really complicated and only the personal testing reveals the right setting.

That's why I said "inverse the field order". When this problem occurs, the inverse field order solve the problems.

Let me give you an example on this:
I have a hauppauge win TV primio FM for analogue capture. If I use the vfm drivers (the one shiped with the card), I never know which field goes top, without testing. If I use the universal tweaked btwincap 8xx(x) wdm drivers, it is always top field A (to field first). So in this example, is a typical driver issue...

From what I know, earlier Unlead products (the one I tested, I don't follow the programs of this manufacture any more..) used their own built-in drivers for their applications. Maybe those are determining field order different that other programs, which are based on the universal use installed drivers

In this case, material captured with the same hardware on the same PC, have to encode different to mpeg 2, depending the program you used to capture...

General Speaking, if you use wdm drivers and virtualvcr to capture, only rarely it is bottom field first (field B).

[Bstansbury]

Thanks everyone for the input. The answer is you have to test your own system, hardware, and software configuration, including the entire process down to the disk and on to the TV to get it right. Stopping short of that won't tell you if you got it right.

I've also confused a lot of the "Jagged edges" with the "Interlace" or not discussion, which can really drive one into a tail spin. Now we can mulitiply things by"N" because of what the difference is in the way Digital Camcorders Interlace the .avi file, and the way that Analog signals Interlace video. Then we multiply them by "N*2" because of the way that MPEG encoders treat interlaced fields.

Hope that drives everyone crazy.

[Bstansbury]

Since my last post on this thread, I've received a lot of additional information. The most important fact is that when I use TMPGen is works on my data and the results are correct on my TV.

The bad news is that just about the entire Movie, and Video industry has had the "Field Order" question wrong at one time or another. Why???

Even AVID had the Field order solution wrong and admitted it on their WEB site,...after they fixed some of their software. AVID is just about the most reqpected Advanced NLE software/hardware company in business.

The problem can be somerized as follows;

1. TV signals are interlace. One field is on Top and one is on the Bottom.
2. Someone started calling them Field A, and Field B a long time ago.

3. The TV industry has settled on "Bottom Field First". For some reason
they did not want to sync or capture the top field first.

4. TMPGen sets .AVI to "Botttom Field First", ... and they label it "B"

5. Ulead VideoStudio also wants .AVI to be "Bottom Field First", but they call this Field A.

6. Pretty much all the high end professional sites that I have been to don't agree on anything except TV interlace signals should be "Bottom Field first.

7. Last point is that,... what every it is on your system don't get the input and output reversed,... and once you get it working don't change it.

It drove me crazy for a week because VideoStudio would input my DV .avi and say it was field A. Then no matter what I tried to set it to in TMPGEn it came out as Field B (bottom field first). It has always worked so they can't be inverted.

Hope this post will help someone still searching for answers.

[LordVader]

I was trying to capture a movie from a VHS tape recently, and i have that jerky movement.

I have like 5 years capturing from Directv, and i've encoded my most recent MPEG-2 files using "Field A" first. Cause is analog of course.

But i tought that i was the same process for VHS tapes, and then i found this problem. I still don't know how to fix it, but all know is that is not working with any of the fields first.

I don't want to waste a DVD blank media, so if you're telling that this thing is going to be fixed using IVTC , i will do it. :P thanks.

[vhelp]

The thing about IVTC is it has its limitations.. or does it 8)

If you final encoded MPEG is STILL giving you trouble..

Really, you MUST parse through your source .avi file for any sudden glitches
in any frames. But, the methods describing this varies because you have
to account for things like,

* How long did you capture THAT particular source (2hr tv movie for ie) for
* Capture cards limitations (ie, does it choke after a certain lengh of capping)
* Did your source have glitch alrady in it
* Did the Station that airs/broadcast such source, did they taint/jimmy the
...source. They do this w/ many movies/tv shows. There is a thread elsewheres
...where I described how (when they alter the speed of source for air-space)
...that this messes w/ the IVTC
* Did that station/broadcaster use a badly done source, and just aired it
...as is.
* Condidtion of the source (ie, if vhs tape, things like age, worn, or many
...uses threw same vcr or worse, other vcr models/brands) causes unstable
...frames. Not even an TBC can fix this. They will sometimes throw off the
...field (in effort to fix frame)
* Dropped frames too. These act like missing fields or bad field orders etc.
* Much more. . .

But, if you are getting jittering (I'm assuming studdering/jerks) in your final
encoded MPEG (specially after an IVTC, of which you believe is a perfect IVTC)
any one of the above could be the reason for the issues that some of you are
still having.

Its true that an IVTC process cannot perfectly de-Telecine a source, (not)
unless your source was broken to begin with. If your source was untouched,
and w/out blemish, even the simpliest of IVTC would produce perfect 23.976/29.970
encodes.

In doing my SW ANH vhs tape, I found far too many glitches that kepts throwing
off my IVTC. I couldn't figure it out (though I had my theories, which I proved
to myself, that they WERE in fact true) why MY ivtc method (a so called perfect
method) would fail in various areas. Then, after realizing my theories, and
putting them to the test, I concluded that there were a number of factors that
I threw into the pile of crap, as causes/excuses to throw off the IVTC.

So, what were they, you ask ?? ?? ??

Simple. As I described above in my bullets, in my VHS sources, they were issues
with respect to the condition of the tape. My SW tape was view many times by
many SW fans. It is a safe bet that many viewers had viewed the vhs tape in
different VCRs and many were more than likely, unkept condition (ie, dirty
heads, missallied heads, leaking recording heads etc) So, it was a no wonder
why my IVTC was constantly broken during the encoding phase. I could see it
inside TMPG's output window. Then, I had it in me to just do a few simple
tests, in vdub's time-line window, and watch in various sections of the video
for any signs of Interlace (aka, broken fields that falled up my IVTC process)
So, What I would do is:
* Count all the bad areas
* Parse as a separate IVTC
* Frameserve all sections (segments) into TMPG for a perfect fix.

This is undoubltly (and in my own brief way of putting it) what you will have
to resort to if you want better encoded sources that require IVTC.

Now, there only source that an IVTC cannot fix are those that have had theirs
fields deleted (for time space reasons or another) These will most probably
not work no matter what you do.

Now, basically, I've given you ALL, the ultimate answer to ALL YOUR IVTC
wows. Now that the secret is out.. Now that I've opended your eyes
to the trueth,
what will YOU ALL do in your next encounter project ??

-vhelp

[Bstansbury]

Loardvader,...

What software or adapter did you use to capture. What format did you capture,.. .avi, MPEG,??? What did you use to encode to MPEG.? Was your field order set to Bottom field first,... or just Field A. (Depends on the software.

Just about anything you can capture is going to be in "Bottom Field First" order,... and if you use TMPGen to encode, it will detect that order and I believe it will force it to be "Bottom Field First". It doesn't matter that it calls this "Field B",... it is still bottom field first. That was my point in my post above,.... VideoStudio makes it bottom field first but says that is "Field A". TMPEGn makes it "Bottom Field First" and calls it "Field B".

IF you still have your source you can load it to VirtualDub or TMPGEn in preview and step to a scene you know was jerky. There you can use single frame step and go forward over several frames. If you see the object jump from one frame to the next and it looks like the motion is going backwards,... that is jerky because of bad field ordering.

I've had it drummed into me that any source that came from TV, either in composite, S-video, or RF,..is interlaced, and should always stay interlaced all the way through capture, editing, and down to re-encode to MPEG and burn to DVD. AFter trying to argue about it for weeks, I did a lot of detailed testing and it works,... even if I don't believe it,... or understand why it works. That's next weeks class,... MPEG encoding for Dummies.

[LordVader]

I have lot of time capturing, probably since 1997, i still have my old Video Highway Xtreme 95, and i've used this bt848 card with amazing results, since then.

In my old times, i always did my captures in 320x240 with my vhs tapes, but across the time i found a way to capture in 640x480, i didn't knew in that time, what was that horizontal scanlines in the image. But all i knew is that i can get rid of it, using an option of de-interlacing.

By that time, i didn't knew how important was the vertical resolution, but since i discover how digital media really works, i've being trying to get the most perfect material near to the source.

Now, in the last 2 years, i've being doing my captures in 720x480, with the famous M-JPEG codec, and i tell you, i don't have any problems at all encoding Interlaced Videos, using my DirecTV decoder as a video Source.

I haven't tried to make a VHS capture since a year ago. But recently i did a test with some tapes, and i did the same process i'm doing with my other captures, and i found that is not showing properly in my TV screen.

My captures are made with virtualdub, and I use tmpgenc, like most guys here, and encode with "top A" first.

I'm not to friendly of the de-interlacing filter, since i know that i have to encode them interlaced to preserve resolution.

I tell you, i have made tests with TV shows, cartoons, movies, and the problem is still there. I also did a test with a FRAME proximity lockers in a very high LEVEL just to convince my self, that it was just a VHS problem, and didn't work either.

Here's my big doubt: why it doesn't happend when you put your vhs tape in a S-VHS using the s-video output ?. That's pretty interesting.