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

[Tom Roper]

Originally Posted by Piotr Wozniacki

However, the ultimate test will be for me that of trying how a Bravia HDTV can interpret a progressive material, put on a BlueRay player.

Therefore, I'm in an urgent need to put some HDV 1080/25p file (around 2 GB) on a regular DVD (don't have HD or BlueRay burner) in such a way that it be readable by a BlueRay player. Is it possible? Which authoring application will allow me to create the structure on my HDD, so that I could just copy it to a DVD as data? I've heard this is possible with Ulead VideoStudio, but only with HD DVD.

At this time, you cannot play it on BluRay. HD-DVD doesn't help either since there is no 24p output. HD-DVD does have 60p output but unfortunately there is an internal conversion from progressive to interlace before going to 60p.

[Steve Mullen]

Originally Posted by Bob Grant

Sorry but I cannot find any post from Simon saying he no longer sees the problem. His last comment on CML (that I don't have access to) seems to be that the excessive coring issue has been reduced in the V1E post the fix.

Simon sent me an email saying the problem was fixed. Since he found the problem, I'd put his post-fix evaluation ahead of anyone elses. And the UK review had no mention of oil paint. Plus, I believe my eyes and equipment since I've got both V1U and V1E samples.

If I blow a single frame up 4X in PS, I can see that solid color areas in P mode are less filled with noise than in I mode.
The absence of noise shows coring in action. Of course, an absence of noise is also a good thing.

Moreover, we have no idea if there is a camera to camera variation. I lived thru the JVC SSE problem. There were huge unit to unit variations. So maybe we should call a truce.

[Bob Grant]

Let me ask a question, simply because there's something that I really don't understand.

I certainly agree based on what I've seen that "sharpness" is very likely the key to avoiding this problem. I don't really understand what Sharpness is in this context. I undertand resolution, MTF, and what edge enhancement attempts to do.
Is the Sharpness setting in the camera reducing resolution i.e. by reducing the setting are we doing irrepairable harm to the image or are we merely reducing the amount of artificial edge enhancement.

Where I'm coming from is you were saying the only question is is "3" acceptable. In my view if "3" is only reducing the amount of artificial EE why wouldn't it be acceptable. My understanding is it's quite easy (render times aside) to add edge enchancement in post. Doing something in post is way more preferable to having something happen in camera that cannot be undone in post.

If you're right I'd have to ask why 3, why not 0 or less. If we can add EE in post surely the cleanest possible recorded image is the way to go.

[Piotr Wozniacki]

Bob, IMHO lower sharpness reduces bandwidth, thus killing fine detail - you won't get them back in post!

But I still have the gut feeling the solution to what constitutes the main problem - line twitter - must be simple and obvious. If a software player, feeding a progressive LCD monitor through DVI, doesn't show line twitter, it implies it is the method of de-interlacing (bobbing) on HDMI or Component inputs that produces it, not the camera.

[Piotr Wozniacki]

Originally Posted by Tony Tremble

I did create a custom profile approximating the look of the V1 but have since reverted back to less colour and edge enhancement. I guess I prefer the flatter less edgy look that the Canon gives. All the edges are sharp but just not over enhanced. That is my preference. But is it yours? Judging by your posts on this forum you do like the aesthetic of the Sony image. And there is certainly nothing wrong with that.

Tony, could you please tell us what settings on the A1 did you use to aproximate the V1 look? How much did you have to crank up the sharpness, and was the picture still quiet (I'm not talking about the "grain" inherent to the A1, but the aliasing artefacts that spoil the V1 progressive picture)?

Thanks!

[Piotr Wozniacki]

Originally Posted by Tom Roper

At this time, you cannot play it on BluRay. HD-DVD doesn't help either since there is no 24p output. HD-DVD does have 60p output but unfortunately there is an internal conversion from progressive to interlace before going to 60p.

I've googled for the Blu-Ray disk format and found this white paper:
http://www.blu-raydisc.com/assets/do...2955-13403.pdf

What strikes me, there is no 25p mode in the specification! (page 17).

What the heck, I mean how is 25p going to be delivered, and if through 50i, what is the point of shooting in 25p?!!

[Steve Mullen]

Originally Posted by Bob Grant

If you're right I'd have to ask why 3, why not 0 or less. If we can add EE in post surely the cleanest possible recorded image is the way to go.

The BBC reports on the A1 and Z1 actually have graphs of frequency response and edge response. At NORM, the frequency response (detail) is very good AND EE is low. They say it is the best balance. And, unlike the pro camers they and you are used to, there is no separate H and V control. I assumed the V1 was the same -- until I ran tests from 0 to 15.

Amazingly, there was no big increase of EE even at 12. This explained why Sony could set Portrait at 15. In my tests the EE seemed relatively constant from 3 to 12. What did change was detail. So it seems in the middle range, Sharpness really means detail. And, Piotr tests seem to confirm that EE isn't a big problem.

I've been sent Rez charts of V1 and it appears the V1 does an honest static 800x700+ resolution. But, there is a good bit of aliasing on the V axis. Aliasing is not new to Sony. Every review by Adam Wilt points to Sony getting max detail even if it means aliasing. Canon chooses the opposite tack.

I'm not defending Sony in the least. It's that with the U the aliasing is acceptable -- perhaps because I'm used to aliasing with Sony DV camcorders -- up to 8 with P. My review of the V1U will definitely point out that even at 5 -- with P there is a bit of aliasing.

In the 25p Piotr sent me, at 3 the image is free of line-twitter and aliasing, just like the V1U at 5. Cutting the frequency response naturally removes line-twitter and aliasing. The problems is that while lowering the frequency response is good in the vertical axis -- it is not needed on the horizontal. And that is the problem. The pix is too soft!

Can Piotr live with 3 softness OR find 4 an acceptable compromise OR find a way to V. filter in post. FCP has a FLICKER FILTER:

Reduces flicker caused by interlacing in still frames that have thin
vertical lines, such as title pages with small text. Three settings are
available: minimal, medium, and max. These settings allow you to
selectively trade off between the amount of flicker and the amount
of vertical softness in the resulting video image.

Seems like this real-time filter is just what you guys need to use with Sharpness set to 5 to 9.

[Mikko Lopponen]

Originally Posted by Piotr Wozniacki

I've googled for the Blu-Ray disk format and found this white paper:
http://www.blu-raydisc.com/assets/do...2955-13403.pdf

What strikes me, there is no 25p mode in the specification! (page 17).

What the heck, I mean how is 25p going to be delivered, and if through 50i, what is the point of shooting in 25p?!!

We watch 25p material daily here in EU. You can send 25p easily in 50i worlds.

[Tony Tremble]

Originally Posted by Piotr Wozniacki

Tony, could you please tell us what settings on the A1 did you use to aproximate the V1 look? How much did you have to crank up the sharpness, and was the picture still quiet (I'm not talking about the "grain" inherent to the A1, but the aliasing artefacts that spoil the V1 progressive picture)?

Thanks!

I'll try and look for my notes when I get home and create the settings again. It was not rocket science though. I had a FCP showing a frame of V1 footage on one monitor and the live input from the XH-A1 into another monitor. The two monitors were colour matched to make the comparison reasonably accurate but not absolute.

The Canon sharpness control is much more simplistic than the Sony V1's. The Canon employs simple edge enhancement while the sony employs a routine more akin to an unsharp mask operator so the looks are not identical.

The any edge enhancement/sharpness setting can increase the visibility of noise depending on the threshold it is set to. I have the Canon set to -2 sharpness and even when toying with +ve values did not notice increased visibility of noise.

I boosted the colour and use the steeper gamma curve along with quite a lot of fiddling with the matrix controls to mimic the colour reproduction and balance of the V1. The standard XH-A1 look is not great it has to be said.

Anyway a month of fiddling and my preset of choice has evolved into something I am extremely pleased with. As far away from the Sony progressive look as possible.

TT

[Piotr Wozniacki]

But the fact is that BLEND eliminates line twitter completely - even when I send the video from VLC through component, and even at 12 sharpness setting! The image loses some resolution, but is still sharper than Canon's and almost equally quiet in terms of aliasing.

Now, I'd appreciated it very much if somebody told be how to implement the same effect in Vegas. Which filter should I be using, so that the H-rez is intact? I could then re-encode some of my test shots, print them back to tape and ruch to the Sony dealer to check it on a Bravia HDTV again.. But I'm running short of time!

[Piotr Wozniacki]

Originally Posted by Tony Tremble

I'll try and look for my notes when I get home and create the settings again. It was not rocket science though. I had a FCP showing a frame of V1 footage on one monitor and the live input from the XH-A1 into another monitor. The two monitors were colour matched to make the comparison reasonably accurate but not absolute.

The Canon sharpness control is much more simplistic than the Sony V1's. The Canon employs simple edge enhancement while the sony employs a routine more akin to an unsharp mask operator so the looks are not identical.

The any edge enhancement/sharpness setting can increase the visibility of noise depending on the threshold it is set to. I have the Canon set to -2 sharpness and even when toying with +ve values did not notice increased visibility of noise.

I boosted the colour and use the steeper gamma curve along with quite a lot of fiddling with the matrix controls to mimic the colour reproduction and balance of the V1. The standard XH-A1 look is not great it has to be said.

Anyway a month of fiddling and my preset of choice has evolved into something I am extremely pleased with. As far away from the Sony progressive look as possible.

TT

Tony, thanks for the info. You're confirming what I remember as my impression when fiddling with the sharpness settings on the A1 - it adds more EE instead of increasing fine detailness. As to the colour though - well, I played with Console so you don't need to convince me that virtually any look can be squeezed out of the A1...

[Bob Grant]

Originally Posted by Piotr Wozniacki

Now, I'd appreciated it very much if somebody told be how to implement the same effect in Vegas. Which filter should I be using, so that the H-rez is intact? I could then re-encode some of my test shots, print them back to tape and ruch to the Sony dealer to check it on a Bravia HDTV again.. But I'm running short of time!

I've used Gaussian Blur on 1000s of high res stills for 16:9 SD DVDs. That FX lets you adjust H & V sharpness, very handy tool.
Back around V5 days it did have a problem, it seemed to calc the blur at project res not source res, so even 0.001 was too much. For that reason I used to run HD projects and then downres. SInce V6 the problem seems to be fixed.

[Bob Grant]

Originally Posted by Steve Mullen

The BBC reports on the A1 and Z1 actually have graphs of frequency response and edge response. At NORM, the frequency response (detail) is very good AND EE is low. They say it is the best balance. And, unlike the pro camers they and you are used to, there is no separate H and V control. I assumed the V1 was the same -- until I ran tests from 0 to 15.

Amazingly, there was no big increase of EE even at 12. This explained why Sony could set Portrait at 15. In my tests the EE seemed relatively constant from 3 to 12. What did change was detail. So it seems in the middle range, Sharpness really means detail. And, Piotr tests seem to confirm that EE isn't a big problem.

I've been sent Rez charts of V1 and it appears the V1 does an honest static 800x700+ resolution. But, there is a good bit of aliasing on the V axis. Aliasing is not new to Sony. Every review by Adam Wilt points to Sony getting max detail even if it means aliasing. Canon chooses the opposite tack.

I'm not defending Sony in the least. It's that with the U the aliasing is acceptable -- perhaps because I'm used to aliasing with Sony DV camcorders -- up to 8 with P. My review of the V1U will definitely point out that even at 5 -- with P there is a bit of aliasing.

In the 25p Piotr sent me, at 3 the image is free of line-twitter and aliasing, just like the V1U at 5. Cutting the frequency response naturally removes line-twitter and aliasing. The problems is that while lowering the frequency response is good in the vertical axis -- it is not needed on the horizontal. And that is the problem. The pix is too soft!

Can Piotr live with 3 softness OR find 4 an acceptable compromise OR find a way to V. filter in post. FCP has a FLICKER FILTER:

Reduces flicker caused by interlacing in still frames that have thin
vertical lines, such as title pages with small text. Three settings are
available: minimal, medium, and max. These settings allow you to
selectively trade off between the amount of flicker and the amount
of vertical softness in the resulting video image.

Seems like this real-time filter is just what you guys need to use with Sharpness set to 5 to 9.

Well Vegas has it's Reduce Interlace Flicker event property switch, to be honest I avoid it now, makes the images way too soft for my liking, the GB FX as I mentioned above is far more selective.

Your observations re sharpness are interesting. What I've seen directly from the camera in they way of noise down towards the blacks looks really odd, no chroma just black chunks. I thought that could have been from EE applied to noise but then elsewhere in the frames is normal noise, very strange.

Problem seems to be that whatever it is when it hits the encoder it goes into overload and that's what's generating that smeared look perhaps. Prior to encoding it doesn't look so bad I think but I'd need to run some more detailed test to be 100% certain.

What's really confusing now is learning that the camera is probably sending 720i (or is it p) to the monitor out it's HDMI port, yish, if this wasn't confusing enough.

Now here's another thing.
Aliasing occurs when the frequency of a signal reaches half the sample clock rate, the Nyquist frequency. One reason stills are always problematic with video in NLEs is the 'cost' of implementing a brick wall filter. Vegas tends to let things go a bit too close, it keeps resolution really high but you can get some real aliasing headaches. Now for DV the sample clock always runs at 13.5MHz in both PAL and NTSC but what about HD, DV can keep the clock the same despite the frame rate as the frame resolution keeps everything the same but this doesn't work for HD, I think.
Could it be that the R50 versions of the camera runs the sample clock at a higher frequency than the R60 variant. Only problem with this theory is that should produce the opposite result, the R60 variant would have more aliasing.

[Piotr Wozniacki]

OK guys - I encoded my test clips in Vegas, using minimum possible ammount of vertical Gaussian Blur (0.001). While it removes aliasing completely (even with original sharpness 12), I like the result of the BLEND de-interlacing option of VLC better. When used with the original video, it removed 90% aliasing while not reducing perceived resolution.

I guess we have done what I can call a quantum leap in understanding what's going on with the 25p video from the V1E. Thank you, guys.

The question remains open, though: what can be done to make the HDMI/Component inputs of our HD monitors and TVs stop deinterlacing the 25pfs video, or at least blend it instead of bobbing?

PS: Somebody is cutting away sections of our posts. Oh well:(

[Bob Grant]

What you've done is the same as I do with high res still for SD, when I say high res I'm talking about images from 10M pixel DSLRs. Thing is how this problem can only happen when you downscale, you need a frequency higher than the sample clock in the source to bring it on. I could probably dig up a post from one of the Sony engineers about SinC functions etc, etc.

Just to further expand that point, the other way HD stills are usually handled is to batch convert them in PS to target res, when you drop that onto a FCP or Vegas T/L there's no problem with line twitter.

So this opens the question as to just where this is happening, given that we're dealing with HD and a camera with a vertical res of say 800 something is fishy. The camera would have to have better V res than 1000 to bring this on I think. Of course if your display device is less than 1080 then this explains it all very nicely.

But even then, working with my HD stills in SD I could never, ever see any aliasing problems using Vegas's internal preview monitors as they are always field merged (aka weave). That's all Vegas can do as the refresh rate of the LCD displays used on a PC is typically 60 or 75 Hz. Could it be that some of these HD LCD and plasma TVs are really displaying interlaced, I can't think of any reason why they couldn't be refreshing the display at the video's frame rate. If so then the problem isn't de-interlacing the de-interlaced, it's that it's NOT de-interlacing. It's displaying the two fields from the 25PsF, 40mS apart.