Hi,

I'm a software developer and I'm working on code to upsample the chroma in mpeg2 video. I was wondering if anybody here knows if the Fx1 in Cineframe 30 mode writes out interlaced or progressive mpeg2. The obvious choice would be to write out progressive, but I know that some cameras in "progressive" mode still break the frame into two interlaced fields and store them as interlaced so that software expecting interlacing will decode them correctly.

I've read before that the Fx1 stores CF30 as interlaced, by my tests show the opposite (see examples below: first is progressive, second is interlaced).

Thanks,
Landrew

P.s. Note that this is not a deinterlacing issue, but rather an issue of properly matching up corresponding luma and chroma samples in a frame. More on the subject can be found here.

http://www.hometheaterhifi.com/volu...bug-4-2001.html

Attached Images

  

I think you already have your answer at dvinfo.

http://www.dvinfo.net/conf/showthread.php?t=120453

All Sony HDV2 spec cams write to tape as interlaced, 60i. Well, if not all, certainly the FX1 does.

Interesting, because the pictures above seem to quite clearly contradict that answer. They were generated in a fairly straight forward way using AviSynth to decode CF30 footage both as interlaced and progressive. I also wrote my own upsampling code starting from the raw YV12 and tried it both ways and came up with the same answer.

What is the source of your answer, if you don't mind my asking? Have you ever heard of anyone actually testing it?

Thanks in advance.

If you spend some time searching the boards back at dvinfo, as well as sony hdv info dot com, you'll find numerous discussions of this from the many professionals who use these cameras daily, as well as links to spec sheets and other documentation from Sony and others...

Sorry I can't post any more specific links, but I do know this has been discussed many times in the three years I've been reading these forums.

If I come across anything I'll post it here or at dvinfo.

I would not trust to what most of these professionals say, as all they know is how to point the lens and press the Record button. This is what they are supposed to -- shoot video. Many of them don't know the inner workings of a camera and they don't have to. If your tests clearly shows that chroma is encoded as progressive, I would go with these results as they are 100% factual, not some bogus info from the specs. I've seen so many times how specs did not describe the product, heck I am working with a device that works in totally opposite way compared to that is described in the specs.

If you do you software for end users, just let them specify the parameters. I would have high-level parameters, like "Camera model: FX1", here you can use your experimental data. Then low-level parameters for advanced users, that override your defaults for the cameras, like picture type or chroma encoding. It would also be nice if you could try both types of processing and deduce programmatically, which one is proper. For example, you can read pixels from every other line, compare them and if there are too many similarities in pixels with every other line, then maybe it is interlaced. Good software should require as little input from a user as possible.

Have you read how Adam Wilt descibes it? http://www.adamwilt.com/HDV/cineframe.html

If it is that important to you, he might answer an e-mail.




Rich

__________________
FireWire Survey

You are probably just viewing both fields at the same time. That would explain why cineframe has better chroma.