60p AVCHD to NTSC MPEG-2 DVD Bug

johnmeyer wrote on 10/30/2012, 4:00 PM
When I render 60p AVCHD video from my Sony CX700 to an NTSC MPEG-2 file (for playback on a DVD), Vegas randomly duplicates frames. This is 100% repeatable, and I have created a simple test case that anyone can use to reproduce the problem.

This is a devastating bug because there is no way around it.

I briefly returned to the forum a few weeks ago to help pinpoint the nested interlaced video bug. I am back one last time to report a much more troubling and very repeatable bug that potentially affects anyone who creates DVDs from their AVCHD footage.

I have confirmed this bug using Vegas 10.0e Build 737 32-bit running under Windows XP Pro. I do have Vegas 11 64-bit running under Windows 7 64-bit, but haven't tried to duplicate the problem on that version.

Here is a link to a sample clip, and a Vegas 10.0e VEG file that you can use to duplicate the bug:

MPEG-2 Render Bug Test Clip and VEG file

The AVCHD clip is directly from my Sony CX700 camcorder without any alteration or change whatsoever. Other clips also exhibited the problem.

All you have to do is load the project, and then render using the "DVD Architect NTSC Widescreen video stream" template. The only changes to that template is that I used 7,600,000 constant bitrate, and "best" for the quality. Here is a snapshot of my Render As template: [edit] Later versions of Vegas default to Upper Field first; I used Lower Field First.



If you put the rendered MPEG-2 onto the Vegas timeline, and match project properties to that clip, and then walk through one frame at a time, you should see many duplicated frames (i.e., no movement from one frame to the next). In my test, I get duplicated frames at frames 96 & 97; 108 & 109; 111&112; 126 & 127; 141 & 142, and probably many more.

I have owned every version of Vegas from 4 through 8, and also own 10.x, and have rendered many thousands of MPEG-2 files. I am quite certain this bug has never been there before.

Here's one more bit of information: I copied the events from the Vegas 10.0e timeline and was able to paste them into Vegas 8.0c. Much to my surprise (and relief) that worked, and I was quickly able to recreate all my work in 8.0c. Here's the important thing:

The result rendered perfectly to an MPEG-2 file, using the same settings shown above.

So, if anyone else can duplicate this, perhaps Sony will take a look at it. They certainly are not going to look at my technical support submission since it is clear that everyone in Sony tech support has permanently gone home. For instance, even though Sony acknowledged the nested interlaced bug here in this forum, I still have not received a reply from my tech support ticket which I submitted over six weeks ago.

Sony really should admit that they no longer provide any support whatsoever for this product, and drop the price down to the $40 level that zero support implies. It is unconscionable that they charge hundreds of dollars and then provide no support whatsoever. I certainly am not going to spend even the "early-bird" upgrade charge to purchase Vegas 12.

Comments

Kimberly wrote on 10/30/2012, 4:38 PM
@johnmeyer:

Thanks for the heads up. I have some spare time on my hands these days so I'll try to duplicate this in VP 12. Stay tuned.

Regards,

Kimberly
musicvid10 wrote on 10/30/2012, 5:56 PM
Does running your TS through TSMuxer, TSDoctor, or Quickstream Fix in VideoRedo TVSuite first help?
I've read several hints here that VP 12 is more touchy when it comes to these files.
flyingski wrote on 10/30/2012, 6:46 PM
John,
Your bug is easily reproduced with your file on my machine with 10e using a CBR render. However, when I changed to a 2-pass VBR render using 7,800,00 max, 7,000,000 avg and 2,000,000 min it rendered perfectly. I'd say it's a bug in the CBR template. I seldom use CBR so never noticed the bug. Have you tried it with HDV video or is this something unique to AVCHD? I'm going to drop an AVCHD clip of my own in and see what happens. I'll post back shortly.
flyingski wrote on 10/30/2012, 7:06 PM
I used one of my own AVCHD clips and the duplicated frames appeared in exactly the same place as in your clip. For what it's worth, my clip was shot with a Canon M-500 using the 30p setting. I did a CBR render on a piece of HDV video also shot with a Canon 30p setting and it showed no frame duplicates so that seems to indicate the problem is in how the CBR template handles AVCHD video. You definitely found a bug. Thanks John.
johnmeyer wrote on 10/30/2012, 7:08 PM
Does running your TS through TSMuxer, TSDoctor, or Quickstream Fix in VideoRedo TVSuite first help?I purposely did not do that because I didn't want to create a "non-standard" file. I do own VideoRedo, so I could try that. Thanks for the suggestion.


Your bug is easily reproduced with your file on my machine with 10e using a CBR render. However, when I changed to a 2-pass VBR render ... it rendered perfectly. ... Have you tried it with HDV video or is this something unique to AVCHD?Interesting.

No, I haven't tried to find workarounds -- either by re-muxing as musicvid suggests, or by changing encoding. I also haven't tried to see if other video has the same problem. I was just happy to be able to reduce it to a simple test case and submit it to the black hole that is Sony Creative Software technical support.

Also, for this particular video I really wanted to keep the encoding rate as high as possible in order to maximize quality, so CBR seemed like the right choice.

Thank you VERY MUCH for confirming the bug.

It is possible that this only shows up with 60p material, or that it only shows up with certain flavors of AVCHD. I don't know. What I do know is that this is a very standard workflow and that it uses video directly from one of Sony's recent camcorders (I bought this nine months ago).

Since people have become accustomed to streaming video that "glitches" (like most video on Vimeo, for example) perhaps they haven't been aware of this bug. I shudder to think of all the glitched DVDs that may have been produced over the past few years, especially if this bug is still in Vegas 11 & 12.

We used to classify bugs into three levels, with level 1 being the most important. This is clearly a level one bug because it means that you cannot produce a DVD with this program using video from this camera (and I suspect other similar Sony camcorders), at least not using the CBR setting. In addition, the VBR workaround may or may not work all the time (who knows??). Like so many such bugs, once discovered, it is hard to trust the program to do that same thing without screwing up. For instance, because of the long-standing Vegas bug where blank frames get created during rendering, I actually have created a program that searches out and finds all blank frames in rendered video. What a pain to have to run this every time I render using these newer versions of Vegas. Thank heaven for Vegas 8.0c and 7.0d!! Also, thank goodness that there has been nothing essential added in any release after these versions. Except for Elastique, there has not been one feature added since version 8 that I really need. The only reason I initially used Vegas 10 for this project is that I thought the timeline performance might be faster (and it is, sort of).

Sigh ...

flyingski wrote on 10/30/2012, 7:19 PM
I edited my previous post while you were writing... I can't duplicate the problem with HDV video but certainly can with AVCHD from a Canon camera so it's not just a Sony camera problem.
john_dennis wrote on 10/30/2012, 7:40 PM
John,

I ran the test with my laptop using Vegas Pro 11-701. Results and render template here.

Though I matched your template faithfully, the VP11 template was initially Upper Field First. I had to change it to match your settings.

I also rendered the project VBR 2 Pass UFF with VP11-701. The only spot I saw anything stay the same for two frames was around 396. Files here.

Edit: Missed 306-307 earlier.

If I have time tonight, I will try Vegas Pro 12 on a different machine.
johnmeyer wrote on 10/30/2012, 8:22 PM
The VP11-JAD Render Test.mpg file definitely contains duplicate frames (126 & 127, for instance).

The VBR 7-6 UFF.mpg shows duplication at 306 & 307. I have a program which automatically scans and finds duplicates (similar, but different from the program that finds blank frames). I could run that and find them all.

It is interesting that your setup produced the duplicates at different locations. That makes me wonder whether the VBR workaround actually stops the problems from happening, or just makes the duplicates happen less often or at different locations.

Thanks for doing the tests!! It shows that some version of the problem is contained in version 11 as well as version 10.
Kimberly wrote on 10/30/2012, 8:42 PM
@johnmeyer:

I was able to reproduce the error in VP11 and VP12 with your 60p footage, but not with 24p footage from my CX760V.

I'll shoot some 60p on my camera and see what that does. In the meantime, here is my 24p footage in case anyone wants to see if they can confirm or deny my findings:

https://dl.dropbox.com/u/41612696/012.MTS
johnmeyer wrote on 10/30/2012, 9:34 PM
No problem with your 24p footage, but I only did cursory tests.

Also, it now looks like this bug is in all versions from at least Vegas 10 (I don't have 9) all the way through Vegas 12. It was NOT in Vegas 8 (see my initial post).

john_dennis wrote on 10/30/2012, 10:04 PM
Vegas 9.0e CBR 7,600,000 bps Lower Field First here.


Vegas 12-367 CBR 7,600,000 bps Lower Field First here.

Stringer wrote on 10/31/2012, 12:12 AM
I was able to reproduce the problem with your footage in VP 11, but not with my 60p footage from Pan TM700.
NickHope wrote on 10/31/2012, 3:45 AM
John, I duplicated your findings using your footage in V10.0e 32-bit. Duplicate frames in the CBR render. OK in the VBR render.

Edit: Well that's confusing. Kimberly reproduces the error in V12 but John Dennis doesn't.
john_dennis wrote on 10/31/2012, 5:25 AM
"Edit: Well that's confusing. Kimberly reproduces the error in V12 but John Dennis doesn't."

Nick,

When I opened the VP12 file, visually, I was able to see at least one duplicate at 306-307.

Wonder if the 90 frames between 306 and 396 is meaningful? Wonder if some counter overflows or some buffer fills up at 306 frames? One thing is for sure. I'll never know.
john_dennis wrote on 10/31/2012, 6:16 AM
I opened the previously rendered video, VP12 7-6 CBR LFF in a Vegas Pro 9.0e project. With the display set to Best/Full, I visually found suspects at 126-127, 306-307, 348-349, 351-352, 378-379 and 381-382. No sophisticated software for analysis here.
NickHope wrote on 10/31/2012, 6:22 AM
Oh I see that now John. I wonder if Kimberley sees the duplicate frames in the same place.

Glad I render my MPEG2 externally.
musicvid10 wrote on 10/31/2012, 9:20 AM
I tried your footage in 8.0c Vista 32 -- mind you, no GPU and an older version of Mainconcept.

I copied your render template to the letter.

Unable to duplicate your issue; no dupes or drops anywhere, and the MPG looks great on the timeline and in a player.

So it's a bug somewhere.

john_dennis wrote on 10/31/2012, 12:21 PM
Someone, everyone, check my work, but I think the behavior began with Vegas Pro 10. Someone else independently verify that the pattern didn't exit in my VP 9 render above. I won't be near a machine today.

johnmeyer wrote on 10/31/2012, 12:42 PM
Thanks again to everyone.

So I think this is the summary of all the tests:

1. Problem with my footage confirmed in Vegas 10, 11, & 12.

2. Problem does happen with footage from some other cameras, but not others.

3. Problem does not seem to happen with 24p footage.

4. Problem does not happen in Vegas 8. However, no one has yet confirmed whether Vegas 9 has the problem.

And finally, no response from Sony tech support, although it's only been 24 hours. For those who've been around for a few years, that actually used to be enough time with a critical bug, especially one discussed here in the forum. I don't expect I'll hear anything from them ever again. They have gone home for good, and those of you here (I'm only visiting one last time) are most definitely on your own.

The only two things that keep me using (older versions of) Vegas is that I know every inch of it and have done just about everything you can think of with it, and more important, I can write scripts that save hours of time. If anyone knows of an editing program that includes scripting, I am gone for good. I wouldn't touch Vegas 11 if you paid me, and I sure as heck am not going to pay money for Vegas 12 to a company that will no longer support me, especially given all that I have done for them (my deshaking scripts for Vegas, for instance, to say nothing of my thousands of posts here in this forum).
Andy_L wrote on 11/1/2012, 6:40 PM
Has anyone noticed how there's no VP12 sticky at the top of this forum? What on Earth is going on around here??
videoITguy wrote on 11/1/2012, 10:18 PM
Andy L, We believe it is the policy in the current forum not to update the sticky for VegasPro12 until the first update (revision) of the file excecutable occurs. So current versions do not get stickies.

However, this was a battle bitterly fought in the companion Video Movie Studio forum and SCS finally relented and acknowledged that Video Studio 12 had been release for quite sometime without being announced! Wooaaa!

Also note that DVDArchitectPro sticky has been stuck for years on version 5.0 build when in fact Version 5.2 is the current build and was the version prior....way off!
Andy_L wrote on 11/1/2012, 10:59 PM
Thanks for the info -- I'll take that as evidence that SCS hasn't been shuttered entirely. Just the support division, I guess. And the people who care whether or not Vegas MVP's are giving up...
PeterDuke wrote on 11/2/2012, 3:48 AM
"Also note that DVDArchitectPro sticky has been stuck for years on version 5.0 build when in fact Version 5.2 is the current build and was the version prior....way off! "

I think that might refer to DVDA Studio.

There ought to be separate stickies for the two versions of DVDA.
johnmeyer wrote on 11/2/2012, 5:14 PM
I actually got a response from tech support today. That is great progress compared to the bug I submitted a month ago which for which I just received a report one hour before this one. However, while the response time is great, the actual content is useless. I guess you could read this in different ways, but my reading is that they are saying they have no idea as to whether the problem is fixed or not, and that it is up to me to download the Vegas 12 trial and do my own testing.

The fact that I received the reply for my one-month old ticket, and this much more recent one, almost at the same time makes it appear that they are just "sweeping" all the unanswered questions off the table without having actually verified or tested them all.

Perhaps I am wrong.

But, what the heck, as a few people have pointed out in these forums, who makes DVDs anymore? If each DVD contains a few dozen duplicated frames, it won't make much difference, right?


Here is the exact text of the reply:

Hi John,