hello all,

Well, I just received my DVX100 and after playing with it for a couple of hours, I have some easy questions.

1.
First of all, I know that I will need some special software to edit in 24p, so right now I am only using 30p. Now, before, with my TRV950 video(60i) I would do the deinterlacing in AE by importing two clips, selecting upper field first on one and lower field first on the other(with detect motion quality on both), then put one on top of the other and set the opacity on the top one to 50%.

Do I still have to do this in progressive mode on the DVX? My guess is that since everything is still encoded at 60i the answer is yes, but just wanted to make sure. I tried it, and the de-interlaced version gets rid of some jaggies that are seen in the raw footage. How do you guys do it? All i have is Premiere and AE.

2.
Is there ANY advantage to recording in letterboxed mode? I ask because, even though I can mask off the top and bottom of a 4:3 image in post-production, perhaps the fact that the DVX doesn't have to compress(as in encoding to DV) those lines causes the algorithm to produce slightly better quality on the 16:9 video...like it has the same amount of space to compress less data, so the resulting quality is better than if it had to encode a couple of lines more?

So far I am very impressed. The amount of manual control parameters alone is worth the upgrade from my TRV950. On the other hand, even though the video is a little sharper on the DVX, I am impressed also with how good the TRV950's quality is for a consumer camera.

Thanks,
Juan

1. You need to set your project to progressive, not interlaced. Your old way sounds insane.

2. The DVX actually processes the entire image in 16x9 mode, it just ads the 16x9 matte afterwards, and the small overhead it takes to do this off sets any performance advantages of not.
If the DVX processed the 16x9 at the CCD level your exposure readings would change.


I prefer to shoot in regular but mask of the LCD, that way i cna adjust up and down in post if a subject goes out of frame.

Thanks for the reply,

I tried loading some 30p footage in AE and setting fields to off in footage interpretation, and it seems to work extremely well.

However, as far as shooting on the camera's 16x9, what I meant was as far as the DV algorithm works. For example, if you compress a file full of '0''s with just about any algorithm, you'll get a smaller file than if the original file is full of random information.

Likewise, my question was if having the DVX100 encode the image with parts of it in plain black(16:9 cropped) versus a larger chunk of random video data(4:3) could result in ~slightly~ better quality in the video due to the extra space available not needed to compress plain black.

I think I found a page that answers this question and indeed says that using the on-board 16x9 results in slightly better resolution than just cropping a 4:3 image in post:

http://members.macconnect.com/users/b/ben/widescreen/

Cheers,
Juan

On that resolution quote I'd want to see hard numbers. People are easily fooled by perception. Check to see if that page you referenced says that the footage shot in "onboard" 16x9 LOOKED better or actually TESTED better.

I used to be into audio about 10 years ago and what cured me was when reviewers started pushing case dampers for stereo components. There are companies that make gelled weights for damping chassis vibrations in components. My personal final straw was when a particular guy from Stereophile or Absolute Sound made this claim about loosening the screws on his amp's case (to improve the sound). Think I'm kidding? I'm quoting facts.

Anyway, I had a trv900 and the dvx blows it WAY out of the water. It isn't even close. As you use it more and more and truly LEARN the camera you'll continue to be amazed.

I edit in Premiere 6.upgrade and I dump my 30p right into the normal timeline... parameters say 29.97 frame rate. The video comes out STUNNING... no lie... STUNNING... in comparison to the trv900?... there isn't any.

You can tell that the timeline is truly working with progressive scans as easily as scrubbing the clips and watching on an analog out... i.e. tv monitor... every image is clean and flicker free. It doesn't matter where you stop the output is clean.

I have a realtime card with realtime analog out running into a monitor on my edit station... so if it were me I wouldn't bother with the deinterlacing process that you (and I) used to do with the Sony. Personally you might think it looks better but check for certain 'cause you may just be getting your screws loose.

I dont buy it for a few reasons, but I am going to hafta do my own tests to confirm.

Briefly though, DV is compressed with Discrete Cosine Transformation in frequency groups of 8 pixels square, the compression is not variable on the entire image, so although the black crop marks are reduced in size, the algorithim for the rest of the frame is still independantly compressed with the 5:1 pixel blocks as normal.

Beyond all the weebee geebee technical talk, you need to still figure out if the camera adds the 16x9 overlay after the compression stage or before it.

Good point.

From the tests the person did in the website it seems like indeed there is a small improvement, but if it is also the case for the DVX100, the improvement might be due to something else, not necessarily the compression.

I would think that the black overlays are added before compression, as it would make sense to perform all operations on the image data captured from the CCD and then just have a hardware section dedicated to encoding to DV and then recording....but i might be wrong.

yes, it would make sense as the DSP functions are added before compression, and also in standby mode (pause) you would think the black bars would disapear if they were only added at the compression stage.. but, depends on if it treats it as a timecode out similarity or not.

That page would suggest that it interpolates the image larger (zoom) does the algotithims on it then shrinks it down to the cropped size.. sounds good, but if you are digitally zooming an image then compressing it down I dont think you are gaining anything.

Im goona go off now and do a DVX test on this.

<<<-- Originally posted by Jarred Land :
Briefly though, DV is compressed with Discrete Cosine Transformation in frequency groups of 8 pixels square, the compression is not variable on the entire image
-->>>
The compression ratio is variable per individual 8x8 block. This is a major improvement over MJPG. It's also the reason that the black-bars=better-compression theory should work.

<<---so although the black crop marks are reduced in size, the algorithim for the rest of the frame is still independantly compressed with the 5:1 pixel blocks as normal. -->>>

The fact that the black bars are reduced in size, means that there are more bits left over to allocate to the rest of the frame. There is an overall "bit budget" per frame, but those bits are allocated to macroblocks on an as-needed basis. Because extremely few bits are needed in the black bars, that means more bits are left over for the actual image data, and therefore it should mean better image quality (because less compression).


<<---you need to still figure out if the camera adds the 16x9 overlay after the compression stage or before it. -->>>

Unquestionably, undoubtedly, absolutely it adds it before compression. Compression is the very last thing done to the image before being written out to tape. ALL DSP, color shift, gain, or digital effects are processed on the image before the compression is applied - this is an undisputable fact.

yes barry.. you are right on the DSP and everything like I mentioned. I just remember someone at Nab telling me something about adding the bars after compression as an overlay and not as compressed data to maintain accuracy (straight lines), as it was just hearsay and I can't remember the details so I just threw it out, thanks for the clarification.

Now on the block per block... I was under the impression that the compression in each block is the only variable aspect of the compression scheme, and there was no give a penny or take a penny between blocks. You are talking about macro blocks, and you lost me :(

At any rate, thanks for the technical clarification Barry.

Barry, the "compression ratio" adaption is only possible for each macroblock (16x16 or 8x32 pixels). If the coding algorithm is "smart enough" and correctly adapt the Q-tables, the bit budget will be mainly allocated to the effective picture area(16:9) which will result in higher resolution and less artifacts (compression noise).

forgotten to add...
"will result in higher resolution and less artifacts (compression noise)"...in the top and bottom side of the picture where the optimized macroblocks reside!

Since you all seem to know a lot i figured I would ask another question i'm kinda confused with...

I got the DVFilm Maker software and on this webpage under brief instructions it says that 'thin' mode must be used since it has the best detail:

http://www.dvfilm.com/maker/dvx100.htm

However, i've done a bunch of tests with different scenes (mostly 30p) and it seems that "thick" mode generates better sharpness than thin mode? Has anyone else noticed this?

Thanks,
Juan

Thick mode results in a "fatter"(coarser) vertical edge enhancement which at some viewing distances, and depending on the image content and display/system intrinsic resolution, results in a subjectively sharper images. It also reduces some artifacts (interline flicker..)