DV.com's Texas shootout questions?

[Jack Walker]

Originally Posted by Tim Dashwood

Yes the Th13x3.5 lens is much sharper than the stock 16x5.5, especially edge to edge. It hardly breathes and doesn't seem to vignette at longer focal lengths. The CA is lessened considerably but still exists. I posted TIFF frame grabs from both lenses a few months ago that clearly demonstrate the differences. I also posted a real-world "rack focus" breathing comparison. I'll try to track them down.

What is the approximate "wideness" of the 13x3.5 lens as compared to the 16x5.5 lens.

If I were to put an adapter on the stock 16x5.5 lens, what would I need to get the same "wideness" as the 13x3.5 lens? Would it be .8? .7? .65? or what?
Thanks!

[Tim Dashwood]

Originally Posted by Jack Walker

What is the approximate "wideness" of the 13x3.5 lens as compared to the 16x5.5 lens.

If I were to put an adapter on the stock 16x5.5 lens, what would I need to get the same "wideness" as the 13x3.5 lens? Would it be .8? .7? .65? or what?
Thanks!

3.5/5.5 = .63

Attached is a measured comparison of Field of View.

[David Ziegelheim]

Originally Posted by Brian Luce

It's funny but the think that struck me about it was how much better the JVC's stills looked as compared to the hvx. In my book that was a pretty resounding endorsement. And let me tell you, some of the fanboys weren't too thrilled about it.

The HVX pictures seemed to have a much softer focus. In the models by the lake shots, the HD100 has the bush behind next to the models in sharp focus, the HVX out of focus. Throwing a couple of shots into Photoshop, the same points have a B 10-20 points lower in HSB on the HVX than the HD100, as if either one was overexposed or the other was underexposed. IMHO.

Originally Posted by Ken Hodson

Just curious why there has been such little discussion on the shootout in regards to the HD100. The six frame "freeze" issue! Why hasn't this been refuted? I have never heard of it befor, yet the DV report now leaves it as gospel fact. Whats up? Also the low scores for motion compression. I'm shocked that a 6GOP structure would be so outperformed by a 15GOP system. I am a long time HD10 user and have never encountered these issues, so I figured the HD100 would only be a step up in perfection. Did they have a gimped cam? Why isn't the JVC community disclaiming these points? Maybe after fighting off SS for so long everyone is tired of defending technical problems.
So bottom line. Good review or bad?

I'm curious. Has non-movement of small artifacts been reported in other places? How would the algorithm decide not to move a leaf, but to move a larger object?

My own questions: was audio recording tested? I thought that was one of the bigger issues with the HD100 vs the HVX200.

In the stills, the HVX200 images frequently had rough edges where the HD100 had sharp edges. The motion shot is a good example. In Jan's presentations, it seems to be just the opposite, with the 4:2:2 recording playing a role. Am I misunderstanding something here?

Thanks,

David

[Chris Hurd]

Originally Posted by David Ziegelheim

was audio recording tested?

During the planning stages of the Texas HD Shootout, everyone involved agreed that there was no practical way to do any serious audio testing given the severe time constraints under which we operated. Our four days were packed to the gills already, and in fact there were a number of video tests that had to be taken out due to the limits of the schedule. Sorry about that,

[Tim Dashwood]

So I just went back and re-read Adam's article. It seems to me that this "6-frame sticky" has been blown out of proportion by some who have obviously read between the lines. (The paragraph in question appears under the photo for "Scenes #60-#67 - Location Work.") It might be easiest to search the article for the word "sticky."

The impression I got from the way everyone was talking about this on dvinfo over the past 24 hours was that the "6-frame stick" was a 6 frame "freeze frame," similar to a Sony 15-GOP dropout, which would be very noticable and obviously a problem if it ever happened.

Upon re-reading I now realize that the specific test the Texas shootout guys conducted for this was to shoot gently moving foilage to push the HDV codec to the extreme. Adam talks about how "some elements" of the image would freeze slightly, then jump slightly instead of moving continuously. This isn't new news for anyone familiar with MPEG compression and I think that in the future we should all re-read the article before jumping to conclusions (myself included.)

I have not seen the results of the Texas Shootout tests with my own eyes but I have done my own similar tests over the past year.
Specifically, when I tested the effects of the motion smooth filter last August I also photographed gently moving foliage.

Here's a sample m2t that has been accessable on my server for almost a year. This was shot with the original HD100 with v1.12 firmware back in August 2005.

In my opinion this is a very complicated scene for any Mpeg encoder and I think it handled it well.

[Stephen L. Noe]

Originally Posted by Jack Walker

What is the approximate "wideness" of the 13x3.5 lens as compared to the 16x5.5 lens.

If I were to put an adapter on the stock 16x5.5 lens, what would I need to get the same "wideness" as the 13x3.5 lens? Would it be .8? .7? .65? or what?
Thanks!

16x = 48 degrees
16x & WAA = 58 degrees
13x = 70 degrees

The 13x is definately the eng style lens to have. The question came up about adding the WAA to the 13x lens. What would happen? 80 degrees?

[Tim Dashwood]

Originally Posted by Stephen L. Noe

16x = 48 degrees
16x & WAA = 58 degrees
13x = 70 degrees

The 13x is definately the eng style lens to have. The question came up about adding the WAA to the 13x lens. What would happen? 80 degrees?

The WAA for the 16x5.5 won't fit on the 13x3.5 lens.

However, there are two wide adapters from Fuji designed for it.
The WCV-85C will give you 0.8X for effective 2.8mm.
The WAT-85C will give you 0.7X for effective 2.45mm.

All the accessories are on page 16 of the lens manual.

[Stephen L. Noe]

We did discuss this "sticky" last year but few took notice.

If you want an example of "sticky" then take a look at this previous thread. You'll find an m2t I shot last year that displays the effect. It's not a big deal because when the raw m2t is re-encoded or uncompressed (whichever the case may be) the floating dissappears. This floating only exist in the transport stream in my experience.

any other comments are welcome.

[Ken Hodson]

Originally Posted by Stephen L. Noe

It's not a big deal because when the raw m2t is re-encoded or uncompressed (whichever the case may be) the floating dissappears. This floating only exist in the transport stream in my experience.

Very interesting. But how can this be? When the data is encoded to HDV all artifacts should be permanent, and therefor transfer over to what ever codec and should not show a reversal. In my opinion the only way the problem can disapear is if it really isn't hard-coded into the HDV stream, but is a problem with the decoding of the stream, not the actual data in the stream.

[Stephen L. Noe]

Maybe it's the downrez of putting it on BetaSP or DVD but it's not apparent. What can trigger it is an extremely slight wobble of the camera. Normal to Massive movements of the camera would never show this effect. It's the slightest camera bump that could possibly make the effect reveal itself and even then you'd have to look pretty hard and know what you're looking for.

Most people would never see this and the fact that they mention it in the article makes me wonder why? I want to see the m2t's they captured. The article makes it seem like the effect is blatantly obvious when in fact nobody mentions it. Now, what was blatantly obvious was SSE. Thank God almighty that fire's been put out.

[Greg Boston]

Adam Wilt can see things that most of us can't. A comment he made while we were watching the playback was, "you never want to go to see a movie with me, I'll ruin it for you"

He has an extremely critical eye just as some folks can hear stuff in audio that the average person would never notice.

-gb-

[Ken Hodson]

If anyone has some "sticky" footage, simply converting it to Cineform codec (AspectHD, ConnectHD or PPro 1.5 with 1.51 update) and let us know if the effect is still there. If it is then we will know it is hard coded and not just a decoder problem.
In the same line of thinking, is there anyone who can provide any example of "sticky" video for us to observe? Anything?

[Stephen L. Noe]

Cropped 640x480 and slowed down to show the effect of motion estimation error.

Les Dit discovered it in my footage last year and commented on it in this thread (click).

[Ken Hodson]

Thanks Stephen. Subtle, but very obvious. Have you been able to get the "sticky" effect when you didn't use the motion smoothing function? I noticed that Steve Mullen guessed the problem right off the bat. "sicky" must be a well known side effect of using motion smoothing.

[Steve Mullen]

Originally Posted by Tim Dashwood

Upon re-reading I now realize that the specific test the Texas shootout guys conducted for this was to shoot gently moving foilage to push the HDV codec to the extreme. Adam talks about how "some elements" of the image would freeze slightly, then jump slightly instead of moving continuously. This isn't new news for anyone familiar with MPEG compression and I think that in the future we should all re-read the article before jumping to conclusions (myself included.)

Tim's 100% correct. With MPEG-2 math is used that breaks a picture into "objects." The the motion of the objects -- hundreds if not thousands of them -- are computed. The tracking information is compressed and recorded. It really is like a Transporter. The actual image vanishes into numbers. The the objects are recreated. Many, many assumptions must be made by the math and many, many measures are taken.

With MPEG-2 there are NO rules that must be followed during the encode. Every codec is designed to meet the esthetic judgements of it's designers. And, every decision involves tradeoffs.

The only rule is the the encoded result must be able to be decoded.

So when we say Sony's codec doesn't handle motion well because it's GOP is 15-frames long -- we are assuming this. That may be the reason, but it's equally possible Sony R&D decided other image aspects were more important.

IF there is such a thing as the so-called 6 frame freeze, it is possible this is not an error, but a side-effect of a decision that JVC made to optimize something else.

Sort of like a drug that cures cancer, but kills 1 out of every million people.

There can be no perfect compression system. Or, to put it another way, every system can be broken. (Already folks have found out how to make perfect pixel copies of HD DVDs!)