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

[Piotr Wozniacki]

Steve, here are my observations:

- when connected LIVE though component to my 1920x1200 monitor, once the progressive is engaged the contrasty lines flicker (these same lines are rock-solid in the interlaced mode)

- the captured progressive video doesn't flicker, either before (as m2t) or after rendering to 1080/25p in Edius, Premiere, or Procoder2 (as mpeg-2).

I'm far from drawing any conclusions yet (have to record under different conditions, in both interlaced and progressive), but I can see NO reason whatsoever to reduce sharpness from default - that's for sure.

Why the above facts do not stick to your theory, I don't know.

[Piotr Wozniacki]

I was given an opportunity to check the fixed V1E, having all the great experience with the Canon A1. As I mentioned before, it's a little too early for me to draw final conclusion and pass a judgement, but a couple of things I can say for sure about the progressive mode:

- no less resolution in 25p than in 50i
- no oil paint effect so far
- line flicker only visible when monitoring live, or playing back, from the camera hooked up through component to a HD LCD; after capture and even without rendering to 1080/25 MPEG-2 (ie playing back from a computer file) no line flicker or marching ants
- no reason to reduce sharpness in progressive!

All this leads to conclusion that the nimb of the V1 still not delivering in progressive mode (even after the fix) has probably a lot to do with the way we're actually watching the progressive material. I'd really appreciate somebody proposing the best workflow (sorry Steve - while what you're saying about it is logical, it contradicts some factual findings of mine: e.g. line flicker, and when it is visible). 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. Only having done this can we:

- compare the picture quality between the I and P-modes (or the V1 to other cameras)
- elaborate a proven workflow (separately for the main delivery means)

[Bob Grant]

Let's just be clear about one thing, for over 50 years in this country and a bit longer in other region 50 countries, 25PsF has been broadcast, originally is SD and now in HD. In case anyone doesn't get how this was done, it's done in a telecine. So 25PsF is hardly a new challenge in PAL land. The question is what is the V1E/P doing that is different to a telecine.

I suspect the answer lies in that a telecine traditionally is scanning each film frame as two fields. Yes those fields can be merged to produce the original frame, almost.

The V1E/P is scanning one frame and splitting that into two fields. At first glance this should yield exactly and precisely the same result, you can merge those two fields to get back the original frame, no almost about it.

I suspect the difference and hence perhaps some of these problems lie in how line averaging is used in interlaced scanning, not only to improve signal to noise but also to reduce line twitter. Line averaging will reduce the vertical resolution by 30% but gain you 6dB in S/N relative to scanning the sensor progressively. Attempting to emulate this by reducing sharpness has a problem, you're reducing both the vertical and horizontal resolution, oops.

Vegas does offer the ability to emulate line averaging by using guassian blur in the vertical direction only, I think it's one of the few NLEs with this FX out of the box.

All this aside though it still leave one obvious question, why is this only a problem with the V1E/P. OK, I can explain why 25PsF from the V1E/P might be different from 25PsF out of a telecine. Doesn't explain why Canon's cameras in 25F don't have this problem it seems. And NO, the fact that they're recording 25p is irrelevant. Almost for certain that 25p is still going to end up as 25PsF as there's almost no way to deliver true 25p, only 25PsF exists in HDV and SD DVDs. The only possible explaination is that the 25F from the Canon has significantly lower vertical res than the 25PsF from the V1E/P, perhaps Canon are doing line averaging before munging the fields into frames?

[Tony Tremble]

Originally Posted by Bob Grant

The only possible explaination is that the 25F from the Canon has significantly lower vertical res than the 25PsF from the V1E/P, perhaps Canon are doing line averaging before munging the fields into frames?

Having owned the V1e and now have the Canon XH-A1 does not have a significantly lower vertical res to the V1E.

By the time you've reduced the sharpness to stop crawling edges and line twitter the Xh-A1 image is far superior in terms of resolution.

There are some instances where there is some stair stepping of bright diagonal edges almost exactly the same as the V1E produces. This is the only thing you notice when comparing 25F with 50i.

It is claimed that Canon clock the green CCD a field out of phase to the red and blue CCDs then the DSP get to work constructing a progressive frame. Think of it as temporal pixel shifting. It works well and by compressing the frame as progressive it is more efficient for the MPEG encoder and as such the Canon image is without the ringing round thin contrasty lines like tree branches against a clear sky.

When comparing similar shots from the V1 to the XH-A1 you find that the XH-A1 has bucket loads of resolution without any hint of EE. If I wasn't so busy I'd get some comparative clips together although I expect I'd only be duplicating what Noel is doing.

Cheers
TT

[Steve Mullen]

Originally Posted by Piotr Wozniacki

But a couple of things I can say for sure about the progressive mode:

- no less resolution in 25p than in 50i
- no oil paint effect so far
- line flicker only visible when monitoring live, or playing back, from the camera hooked up through component to a HD LCD; after capture and even without rendering to 1080/25 MPEG-2 (ie playing back from a computer file) no line flicker or marching ants
- no reason to reduce sharpness in progressive!

All this leads to conclusion that the nimb of the V1 still not delivering in progressive mode (even after the fix) has probably a lot to do with the way we're actually watching the progressive material. I'd really appreciate somebody proposing the best workflow (sorry Steve - while what you're saying about it is logical, it contradicts some factual findings of mine: e.g. line flicker, and when it is visible). 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. Only having done this can we:

- compare the picture quality between the I and P-modes (or the V1 to other cameras)
- elaborate a proven workflow (separately for the main delivery means)

1) When you say so difference in resolution -- are you using equal Sharpness for I and P? And, what value(s)?

2) I agree with you that IF you see you line flicker -- it will be "viewing from the camera." You are seeing PsF video WITHOUT anything being done to it. And, the monitor INPUT is obviously in "video" input mode -- not "PC" mode.

3) Playing back a file opens the question of the player software and nature of the computer output. Playing back a file also opens the question of which INPUT type is being used with the monitor.

Bottom-line -- it appears that some see flicker while you don't (and I don't) when they play captured video.

My "workflow" was only to force there to be NO fields in the video for those who DID see flicker in captured video. Forcing the video file to have only frames removes the possibility.

However, Bob is correct. Any progressive video converted to PsF should have flicker. In fact, the V1 in I mode should have flicker!

A) The reality is that 1080i broadcasts do have flicker -- which is one of the reasons ABC, ESPN, FOX, DOD, and NASA don't use it. I suspect the reason we don't see much with movies is because a guassian filter is used during telecine.

B) Since there appears to be no sensitivity difference between I and P -- it seems the V1 is ALWAYS reducing vertical resolution by 30%. Therefore, there should be as much flicker in I as there is in P. But, for you -- and others -- this does not seem to be true. WHY?

You raise an important point -- I'm glad you said it -- that it "probably a lot to do with the way we're actually watching the progressive material."

If we back way up and stop assuming the artifacts are IN the V1, we solve a lot of problems. We can stop wondering why there are reported issues (flicker and ants) in P mode, but not in I mode. We can stop wondering why some see flicker and some don't.

Instead, we can look at why some displays have issues with PsF verses I HD video. For example, we know pure progressive stills will flicker on an interlace monitor but not on a progressive monitor. We don't blame the computer because we understand WHY the display is showing the flicker.

I've often wondered, given the near total lack of HDTV in Region 50 -- what kind of monitors are being used. How many folks have a real HDTV at home? Or, are they using computer monitors. If over 50% of real HDTVs fail to deinterlace correctly -- I'll bet none of the computer monitors do. Likewise, I'll bet few of the 16:9 EDTV's sold in Region 50 do -- because they simply show 540-line fields.

In short -- it may well be that V1 video "artifacts" are artifacts from the deinterlacer used in player applications and monitors. This explains why there are so many conflicting claims about the V1 from Region 50.

(We don't get flicker reports from Canon shooters because it uses only half the CCD's vertical resolution in 24F mode. By definition, it is "smearing" the image vertically because with motion video it is taking 1 field and line-doubling it to 1080-lines. That's why Canon can't use the term "24p" or "24PsF." It is not a progressive camera.)


The good thing is that your report makes clear that other than line flicker -- the fixed V1E/P have NO problems. And, if the flicker is a display artifact, then you guys have the same quality V1s as we do.

[Piotr Wozniacki]

Steve, it's too early to say it's flawless - even Sony admits it is. I have to test it further. One thing I can confirm however - there is NO need to reduce resolution in progressive (to answer your question, I've used default for both I and P - is it 7?)

Let's not forget those alarming screengrabs of the red shed's wall - they were clearly "oil-painted" in progressive, and it was the NTSC version. So far, I haven't noticed anything like it.

Speaking of monitors: yes, when fed directly from the camera, mine is using a "video" (as opposite to "PC") input (Component). I can use this input also when playing back a captured (raw m2t, or rendered to 1080/25p MPEG-2) file - through a separate HDTV component output of my graphics card. There is no flicker there, just like there's no flicker when playing back using the PC input (DVI-D).

But it cannot be true that my monitor doesn't do anything when fed through component. If it didn't, I would see the interlace artifacts (stairstepping) with I material; I don't! Of course, when a file is palyed back from the computer, and the material is interlaced, I can see stairstepping...

This only adds to the complexity of the problem, because - why clearly some process (deiterlacing? which method?) is used by the component input, I don't know much about it. Whatever the process, I'll never agree it reduces the vertical resolution by half - not at all, the 1080i video is clearly full resolution, but with no interlace artefacts.

If it applies some treatment to I material, it probably does so to P material as well, because it cannot "know" it is PsF and not I (some software can; eg. Sony Vegas correctly reads the progressive flags and reports the captured m2t's as progressive). I predicted this potential problem a couple of months ago, when first reports on the 25p mode on the V1E surfaced - I wrote the monitors people use are probably trying to "deinterlace the deinterlaced"...

Last but not least, I cannot leave uncommented your claim about the Canon 24F showing only half the vertical resolution - it clearly is NOT the case. Objectively, it has been measured to resolve just some 10-12% less horizontal lines than the I-mode; subjectively - on the same monitor, I didn't notice ANY resolution drop in the F-mode.

[Piotr Wozniacki]

Well, first sign of weakness spotted: single-color surfaces in shadow areas of the 25p picture DO have more noise than in interlaced. Is it when some NR used to kick-in, creating the infamous "oil paint effect"? Would it also appear with the firmware version I'm testing now? To be seen...

[Tony Tremble]

Originally Posted by Steve Mullen

(We don't get flicker reports from Canon shooters because it has only half the vertical resolution in 24F mode. By definition, it is "smearing" the image vertically because it is taking 1 field and line-doubling it to 1080-lines. That's why Canon can't use the term "24p" or "24PsF." It is not a progressive camera. It is a 24fps, one field, line-doubled, interlaced camcorder.)

You made the claim. Now prove it.

Show us some clips that prove your case. Are you a credible source of information or not? Let the public decide by posting examples that demonstrate your claims.

Thanks

TT

[Steve Mullen]

Originally Posted by Piotr Wozniacki

I've used default for both I and P - is it 7?)

Let's not forget those alarming screengrabs of the red shed's wall -- they were clearly "oil-painted" in progressive, and it was the NTSC version. So far, I haven't noticed anything like it.

But it cannot be true that my monitor doesn't do anything when fed through component. Whatever the process, I'll never agree it reduces the vertical resolution by half - not at all, the 1080i video is clearly full resolution, but with no interlace artefacts.

Last but not least, I cannot leave uncommented your claim about the Canon 24F showing only half the vertical resolution - it clearly is NOT the case. Objectively, it has been measured to resolve just some 10-12% less horizontal lines than the I-mode; subjectively - on the same monitor, I didn't notice ANY resolution drop in the F-mode.

RE: The Canon is using a very nice line doubler to get the best it can from a single field. Static measures allow a line doubler to do a good job. They can't do the same with motion. The measured dynamic rez shows this: 24F is only 540-lines (one field) while 60i is 700-lines. That's a huge difference.

RE: Yes, your monitor must deinterlace. Everyone who has a "failing" HDTV claims they would notice. They haven't read my reporting on deinterlacers. bob presents 540-lines per field -- which is obviously not half of a field. You do get 1080-line per frame.

It's not that an HDTV fails to deliver 540-lines per field. It fails because it doesn't do better. A motion sensitive deinterlacer can present a much BETTER pix. It can do 1080 PER FIELD for static pixels and about 800 PER FIELD for dynamic pixels. Obviously, you can't see HOW MUCH BETTER your pix could be until you compare two monitors side by side. My Sony passes the test -- so obviously I can't see how much poor it could look.

Anyone working with 1080i must have an HDTV that passes the test. But, I doubt such tests are even being run it Region 50. And, I doubt many of the buyers of HDTVs anywhere have any idea about this issue. They think resolution and size are the important issues. You can download an MPEG-2 file and test your own monitor.

RE: I never saw any issue with the red shed. I played my 24p transferred to HD DVD tonight, and saw only perfect video. I think you can forget worrying about "oil paint" -- there hasn't been a report from anyone in a month. Obviously, it was a bad version of firmware that went into the first batch of cameras.

I finally did spot a bit of what might be line flicker at night with +9dB gain, but no more than I see on any video-based 1080 broadcast.

I also watched 720p60 I shot with a JVC GY-HD250. With 720p, I have a 100% progressive path to my Sony 720p display. Just as with 720p60 broadcasts -- no flicker. The only way to have artifact free HD is by shooting 720p60. It also removes the motion judder artifact of 24p and 25p. :)

I've also set my HDTV Sharpness to 16 of 100 which removes 95% of EE from the TV. I'll be getting Video Essentials on HD DVD next week. Finally a way of testing an HDTV. Have you calibrated your monitor?

The default for CINE is 5 and VIDEO is 7.

[Greg Boston]

Originally Posted by Steve Mullen

(We don't get flicker reports from Canon shooters because it has only half the vertical resolution in 24F mode. By definition, it is "smearing" the image vertically because it is taking 1 field and line-doubling it to 1080-lines. That's why Canon can't use the term "24p" or "24PsF." It is not a progressive camera. It is a 24fps, one field, line-doubled, interlaced camcorder.)

That's simply incorrect. Canon doubles the ccd clock rate to 48hz and pulls 24 frames out of those two fields with some intelligent blending by the DIGIC DV II processor. There is a about a 10 to 25 percent drop in vertical resolution.

It's also believed that they disable the row-pair summation mode of the interlaced ccds to maintain higher vertical resolution.

The only reason Canon calls it F instead of P is because the CCD's themselves are native interlaced. Call it 'truth in advertising'. There are also licensing issues with 24P recording in video cameras because a patent was granted that IMHO should never have been issued. So Canon is able to produce an output virtually identical to 24P without paying the licensing fee. They don't simply line double one field and call it 'progressive'.

-gb-

[Steve Mullen]

Originally Posted by Greg Boston

That's simply incorrect. Canon doubles the ccd clock rate to 48hz and pulls 24 frames out of those two fields with some intelligent blending by the DIGIC DV II processor. There is a about a 10 to 25 percent drop in vertical resolution.

It's also believed that they disable the row-pair summation mode of the interlaced ccds to maintain higher vertical resolution.

The normal clocking rate for 60i interlace is 60Hz so they SLOW IT DOWN to 48Hz for 24F.

Canon refuses to explain HOW it gets 1080-line frames. There are several ways it could be done: vertical green-shift based interpolation is raised as a possibility. A 2D FIR filter is another option. Highly filtered "weave" could be used. Highly filtered "bob" could be used. A motion sensitive system could be used. And, yes a motion sensitive system could "blend" both fields, but only for static video. Once there is motion, it would have to switch out of blend. "Blending" is simply a form of line-doubling.

Bottom-line, measured dynamic rez shows this: 24F is only 540-TVL (one field) while 60i is 700-TVL. The 540-line measure tells us that upon motion -- only a single field is used.

By the way -- a 25% drop is rez is hardly "virtually identical to 24P." And, no matter HOW it's done -- line-doubling/interpolation/etc. is not equal to progressive video because resolution varies based upon motion. That is the KEY to understanding why 24p is so much better a system.

There IS a slight 1dB drop in sensitivity with 24F, but not the 6dB expected or 3dB given the slower shutter-speed. And, I've not seen any reports of a 3dB decrease in S/N ratio. So there is no evidence for turning-off row-pair summation.

The data does not support your claims. Canon can't call it p because it ain't. What a bizarre effort to explain away Canon's design. In any case -- discussing HOW the Canon works is way OT in this thread.

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Well, first sign of weakness spotted: single-color surfaces in shadow areas of the 25p picture DO have more noise than in interlaced. Is it when some NR used to kick-in, creating the infamous "oil paint effect"? Would it also appear with the firmware version I'm testing now? To be seen...

DNR of any kind must reduce vertical information in order to reduce noise. The problem is that DNR can be in the camera, in the encoder, or in your MONITOR. (Have you disabled every special mode in your monitor?)

Alas, one can't escape the point you raised -- the monitor is part of the system. It's why the AVS Forum is so critical to use BEFORE buying an HDTV. To solve this -- I only recommend 2007 46-inch Sony Bravias. It's a full spectrum LCD.

[Greg Boston]

Originally Posted by Steve Mullen

The normal clocking rate for 60i interlace is 60Hz so they SLOW IT DOWN to 48Hz for 24F.

No kidding? Really? When I said double the scan rate to 48hz, I was referring to DOUBLE THE INTENDED FRAME RATE. As opposed to the XL2 which has native progressive CCD's and actually clocks them at 24hz.

Originally Posted by steve mullen

In any case -- discussing HOW the Canon works is way OT in this thread.

I agree, which is why you should have left that remark out of your earlier post. You were the one who brought Canon into the discussion. In fact, I could edit it out for you if you like.

-gb-

[Chris Hurd]

Canon... uses only half the CCD's vertical resolution in 24F mode.

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

Canon refuses to explain HOW it gets 1080-line frames.

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,

[Piotr Wozniacki]

Originally Posted by Steve Mullen

Alas, one can't escape the point you raised -- the monitor is part of the system. It's why the AVS Forum is so critical to use BEFORE buying an HDTV. To solve this -- I only recommend 2007 46-inch Sony Bravias. It's a full spectrum LCD.

Steve, while of course I'm admitting my knowledge is not as profound as yours - some of your claims simply do not reflect the reality I'm vitnessing here (with both 25p and 50i of the V1E, but also with the 25F of the Canon A1). But in order to give them justice, tomorrow I'm taking my V1E to the Sony dealer here; if it doesn't show line flicker at one of the 2007 Bravias, I'll admit you are right on that.

The fact is: 25p introduces a lot of flicker when fed directly (live or from tape) to the 1920x1200, calibrated LCD through component.

The same material - after having been captured into a raw m2t file (or further rendered into 1080/25p with an NLE) - doesn't show ANY flicker. Regardless on whether it's played back with a software player and fed to the same monitor via DVI-D, or using component output of my graphics card and component input of the same monitor, which showed heavy flicker direct from camera! This fact means that - as opposed to what Prime Support's advise has been for the V1E users - there really is NO reason to turn down the sharpness (unless perhaps one is going to only watch his recordings from the camera, without editing - but how many of us do?).

Whether the "true" 25p of the Sony V1E is at least as good as the "fake" 25F from the Canon A1, is still to be established, though.