How PsF video from the V1 is different than "p" or "F" video

[Steve Mullen]

Originally Posted by Piotr Wozniacki

What has to be done is establish a common and unambiguous terminology for phenomena such as line flicker vs marching ants, mosquito noise, interlacive stair-stepping etc.

Your point may be important. Line flicker is an exect 25Hz alternation of a single line. Is that what you are seeing? Or, are you seeing "noise" above horizontal lines?

Adam Wilt's PRE-review notes that there is a bit of "aliasing" above horizontal lines. Clearly this can't be "stair-stepping" aliasing since it's not a diagonal. So it's seen as "noise."


--------

There may also chroma filtering in the decoder to help reduce ICE -- Interlace Chroma Error. This can appear as a serration along a horizontal edge between contrasting colors. This may well be what some called "ants." So the hardware/software decoder can play a role in what folks see.

Adam also notes the V1's V. rez may be 800-TVL with static images. That's very good!

[Piotr Wozniacki]

Originally Posted by Steve Mullen

Your point may be important. Line flicker is an exect 25Hz alternation of a single line. Is that what you are seeing? Or, are you seeing "noise" above horizontal lines?

Steve, it IS *very* important that we talk and think the same thing. By "line flicker" I mean an alternation, not noise; however it is not exactly 25 Hz - much less than that. I have posted a screenshot from VLC playing a short clip in 25p (raw m2t file), you can find it here:

http://rapidshare.com/files/17431478/V1E25p.png.html

When you look at how bright the car body looks (even though it was actually very dirty:), just imagine it's contrasty edges (such as the horizontal thing above the registration plate) alternating with uneven, but rather low frequency - this is what I call "line flicker". As you see from the grab, there's no noise whatsoever around it! But again - the flicker was only present when I played the clip back from the camera hooked via component; in the captured file - whatever input I use to feed it to the monitor - the flicker is gone.

[Steve Mullen]

Originally Posted by Chris Hurd

The vertical resolution loss is only about 10%. It is not line doubling.

Incorrect. At a New York press event in November 2006 among some thirty-odd invited journalists, Canon USA presented a paper describing exactly how Frame mode works. Hope this helps,

Is that white paper at their site? If not -- it doesn't "help."

Line doubling is a general term for doubling the number of lines. Obviously to get frame with no motion artifacts ALL the lines that wind-up in a frame must be capture at a single point in time. Since the CCDs are, according to Canon, running in interlace -- only a single field can be used. Since a field has only 540-lines per color -- SOME system must double that to 1080-lines per color. Canon can say its NOT "line-doubling" all they want, but at the end of the process the number of lines has been doubled. Moreover, it doesn't matter what anyone calls the system -- it's not a 24p camcorder. And, that's shown by the rez measures.

Adam Wilt et all measured the Canon. It is 540-TVL in 24F and 700-TVL in 60i. That is not a 10% decrease. He also has measured the V1 at about 800-TVL. That's is much higher than the Canon. These are public numbers that anyone can verify by reading the Texas Shootoff series in DV.

Sorry for the OT.

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Steve, it IS *very* important that we talk and think the same thing. By "line flicker" I mean an alternation, not noise; however it is not exactly 25 Hz - much less than that. Alternating with uneven, but rather low frequency - this is what I call "line flicker".

Ahhhhhh -- so it's not "line flicker" as in alternating fields at 25Hz.

There is another term "line twitter" which is almost always said to be the same as "line flicker." But, for years I've been wondering if it really was. What you are seeing seems much better described by the word twitter.

And, it's not new. The VX100's 24p mode had a LINE THICKENESS setting to prevent crap from appearing when progressive was displayed. In the B model it got 3 levels. It was a low-pass or gaussian filter that slightly decreased V. rez.

IF I'm right, the V1 passes the full V. rez to your monitor, but your graphics card has a "filter" that very slightly cuts V. rez. It may be a real filter -- or only the result of less that perfect components.

[Piotr Wozniacki]

Originally Posted by Steve Mullen

IF I'm right, the V1 passes the full V. rez to your monitor, but your graphics card has a "filter" that very slightly cuts V. rez. It may be a real filter -- or only the result of less that perfect components.

This is probable, as I obviously don't get as bright and punchy a picture from the computer (either via DVI or component), as it is from the camera - but hey, it would mean the V1E DOES have a problem, after all! However, this is NOT resolved by turning down the sharpness, which would turn the whole picture into mush.

PS. Steve, do you think that hooking up the V1E to properly deinterlacing HDTV (like the new Sony Bravias) would eliminate, or increase the line twitter phenomenon that we have just defined?

[Douglas Spotted Eagle]

Originally Posted by Steve Mullen

Is that white paper at their site? If not -- it doesn't "help."

Canon did present it; regardless of whether the paper is in the public domain, it *was* presented to journalists and industry professionals.

"Twitter" and "Flicker" are not interchangeable words. Poynter is the only person I've seen to deliver a comprehensive explanation of the difference, but they are indeed, different. One is an interlacing artifact and the other is contrast-driven in conjunction with interlacing artifacts.
Most high end camcorders have twitter filters that reduce the vertical detail information that would typically cause twitter in an interlaced image. It's not hard to test.

[Douglas Spotted Eagle]

Guys, this one is about off the rails; let's stick to discussion of the Sony camcorders, and specifically the progressive and interlaced characteristics of the V1. This thread is not a discussion of Sony vs Canon vs anyone else.


Thank you,
The Management :-)

[Steve Mullen]

Originally Posted by Piotr Wozniacki

This is probable, as I obviously don't get as bright and punchy a picture from the computer (either via DVI or component), as it is from the camera - but hey, it would mean the V1E DOES have a problem, after all! However, this is NOT resolved by turning down the sharpness, which would turn the whole picture into mush.

PS. Steve, do you think that hooking up the V1E to properly deinterlacing HDTV (like the new Sony Bravias) would eliminate, or increase the line twitter phenomenon that we have just defined?

1) We want the camcorder to output the max possible rez. It would be a "problem" if it didn't.

2) I see no twitter with my Sony via HDMi from the V1 -- so it does seem to be monitor dependent. Have you turned down monitor Sharpness? DSE confirms twitter is different than flicker.

3) Record an edited 25p back to the V1 and see what the video is like. It may be that after one decode and one encode -- twitter is will be gone.

I ran a whole series of 60i, 30p, 24p tests at all Sharpness levels today. Will view tonight.

But here's the neat thing. With Peaking ON, you can see the detail be removed as you dial down or see it be added as you dial up. Using this field test -- I could see no difference between I and P.

[Piotr Wozniacki]

Originally Posted by Steve Mullen

1) We want the camcorder to output the max possible rez. It would be a "problem" if it didn't.

Agreed; IF it's a monitor problem, then I could only welcome this as a sign that the V1 is indeed pumping much higher rez (and dynamic range) than other cameras in "prog" mode.

Originally Posted by Steve Mullen

2) I see no twitter with my Sony via HDMi from the V1 -- so it does seem to be monitor dependent. Have you turned down monitor Sharpness? DSE confirms twitter is different than flicker.

This makes me even more anticipating my tests with the V1 conected via HDMI to one of the new Bravia HDTVs.

Originally Posted by Steve Mullen

3) Record an edited 25p back to the V1 and see what the video is like. It may be that after one decode and one encode -- twitter is will be gone.

Will do just that.

Originally Posted by Steve Mullen

I ran a whole series of 60i, 30p, 24p tests at all Sharpness levels today. Will view tonight.

But here's the neat thing. With Peaking ON, you can see the detail be removed as you dial down or see it be added as you dial up. Using this field test -- I could see no difference between I and P.

Do not quite follow what you mean, sorry - could you reword it for me? Thanks!

[Steve Mullen]

Originally Posted by Piotr Wozniacki

This makes me even more anticipating my tests with the V1 conected via HDMI to one of the new Bravia HDTVs.

Do quite follow what you mena, sorry - could you reword it for me? Thanks!

Be sure to drop Contrast to about 70% and Brightness to about 50%. Turn sharpness down to about 20%. Turn off every other "special" function like DNR and Gamma Enhance. Color at 40 to 50. And, Color Temp to WARM.

Since the peaking output follows the Sharpness control -- you can see the Sharpness alter the peaking display. Higher Sharpness causes more detail to become clearer -- hence more yellow detail.

[Steve Mullen]

I viewed my Sharpness tests tonight. So far it's only V1 via HDMI to monitor.

I-mode: at "7" -- I could see very little EE. (Remember, my HDTV is adding almost no EE.) A small amount of EE is necessary to get clarity. Lowering below "5" reduced fine detail. At zero, video was very soft -- but not mush. The control is not aggressive. However, fine detail became even better as I increased above "7." The control really is a DETAIL control.

30p and 24p: VERY slightly less sharpness on vertical lines/edges than 60i/30p. Aliasing in the form of "dancing dots" on horizontal lines/edges. Dancing dots on VERY thin diagonals.

No line-twitter.

[Piotr Wozniacki]

Originally Posted by Steve Mullen

Since the peaking output follows the Sharpness control -- you can see the Sharpness alter the peaking display. Higher Sharpness causes more detail to become clearer -- hence more yellow detail.

Oh, that. Sure, but it's a very rough indicator, as it also depends on the LCD own contrast and sharpness.

[Piotr Wozniacki]

OK, so I have uploaded an MPEG-2 clip for you to check and comment. I'd like to upload raw .m2t files, but there's a limit of 100 MB so it's difficult to shot exactly what you want, and use just below the file length limit. This means some basic editing (trimming etc) is needed; in the old DV days I used a great shareware utility that could trim files (avi or mpg) without re-compressing. It was quick and didn't change anything in the original video; does an utility like this exists for HDV m2t files? Please let me kow!

In the meantime, I've trimmed and merged in Premiere 2.0 two clips - one is 25p and the other is 50i, they show roughly the same scene - you have branches in the sky etc. Of course, due to the season of the year we're having, the sunshine is not as bright as needed for a perfect HDV - but you will get and idea of differences.

Here, the progressive clip played back from the camera via component showed terrible line twitter. The same clip - captured and played back from the computer on the same monitor - doesn't show even a sign of it, as I mentioned before. Please let me know if you can see it - take a close look at the horizontal board edges in the shed's wall. Thanks!

And here's the link: http://rapidshare.com/files/17560397/test2.m2t.html

Please share your opinion on the video (not cameraman - it's poor) quality; whether you can find flaws (the line twitter, oil paint - anything)!

PS. Apart from line twitter, which hopefully is just a matter of display device - after a closer inspection I can see a liitle more noise in the progressive (both were shot at 0dB, sharpness 7 and 1/50th; exposure and focus were auto). This noise *might* have been the reason for some DNR to kick in, causing oil paint effect in the previous firmaware version (just IMHO). I can now safely say that the very same scene looks softer (less overal rez), but also less noisy with the A1 (also at default sharpness, but with -3dB gain, IF it's comparable at all).

PS.PS. here is a clip from Canon; 25F various picture profiles:
http://rapidshare.com/files/1760054...nce_01.m2v.html

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Apart from line twitter, which hopefully is just a matter of display device - after a closer inspection I can see a liitle more noise in the progressive (both were shot at 0dB, sharpness 7 and 1/50th; exposure and focus were auto).

Thank you. Using VLC to play via HDMI to my hdtv I see three things.

1) on P: aliasing on thin horizontal lines that appears as "dancing dot noise.

2) on P: diagonal stripes through slanted lines -- what I believe is called "barber pole."

From an article on ultra-sound:

"If the [low] I/Q Nyquist sampling rate is maintained all the way through the detector, axial spatial aliasing would result, which is visualized as a distracting "dancing dot" or "barber pole" artifact in the image whenever there is axial motion from frame to frame."

In short these are aliasing artifacts. It matches Adam Wilt's pre-review of the V1. Aliasing seems to be increased by movement (axial motion).

3) on P: no line twitter UNTIL I switched VLC from deinterlace OFF to BOB, DISCARD, LINEAR, and X. OFF, BLEND, and MEAN show no line twitter.

Line twitter is also caused by vertical movement over very thin lines.

None of this I see on I.

Anything that filters vertically seems to fix the problem. What do you see when you export back to HDV?


I'll burn your file to HD DVD when I get the chance.

UPDATE:

30p and 24p: Aliasing in the form of "dancing dots" (looks like noise) on horizontal lines/edges and on VERY thin diagonals.

Actual resolution seems the same on tree limbs and wood grain textures in all modes.


Can you send me a 3 sample and a 5 sample to add to the HD DVD.

[Piotr Wozniacki]

Steve, I'm now *[EDITED]* sure the twitter is display dependent in the sense of whether or not it's trying to deinterlace; if it is not - the twitter is gone (the side effect of this conclusion is that my LCD component input *IS* in fact deinterlacing - hence twitter in from-camera progressive, but no hotizontal motion stair-stepping, or combing, in interlaced).

This leaves me with another concern that I would like to be assessed before I re-consider the V1E: what can be done with what I simpy called "noise", and you more precisely descripted as "dancing noise". It can be very distracting; do you believe it can be eliminated by reducing sharpness? I'm asking about the theory behind it; practical tests I will do myself later because right now it's snowing here so no right comparison to yesterday's clips is possible.

Thanks!

EDIT: I played the clip with VLC and like Steve says: the Bob deinterlacer re-introduces heavy horizontal lines twitter, which is present when playing back from camera. This means it definitely (and not just probably) is a display-device related problem.