EX1/EX3 line twitter and shimmering

Alister Chapman · November 19th, 2009, 01:07 AM

I would place money on it being monitoring. There are very few consumer TV's that do a good job of EX footage. Try turning down any detail or sharpness settings on the TV.

Vincent Oliver · November 19th, 2009, 01:55 AM

Just to add my two pence worth.

I also had twitter problems which were very noticeable when shooting handheld. Slight movement on the camera caused any fine detail to shimmer. Using a tripod cured a lot of the problems.

I am still working on acheiving a better result, but I can live with what I have now.

Bob Grant · November 19th, 2009, 07:00 AM

I've only had one bad line twitter problem with footage from my EX1. I shot 720p50 and the small hall had dark brick walls but the bricks had a very fine white line running though them.
The walls shimmered something fierce. The problem was wrangled easily enough using Vegas's Gaussian Blur applied in the vertical direction only. This filter has been my goto for years, well before the EX1 was even thought of. Vegas seems to have much the same flicker reduction filter as Ppro and it just too much blur to be usefull. I can't find the same quality of GB filtering in Ppro CS3. I'm pretty certain though that FCP has a GB filter that has both vertical and horizontal blur control. As I'm not a FCP user I cannot comment on how effective this is.
The line twitter / aliasing issue is nothing new, I and others went through the same issue with the V1P shooting 25p, the V1U produced slightly lower res and seemed not so prone to provoking this issue. Ultimately this is not the fault of the camera, I see a lot of the same problem making it's way into OTA broadcast material.
Shooting hand held would make the issue more noticeable, more movement of fine detail vertically is going to make the problem more noticeable. Giving up on shooting hand held seems a backwards way to wrangle this issue. Once you recognise what the cause is it's then much easier to deal with it and I've yet to find any well reasoned argument to suggest the problem lies in anything other than vertical resolution, too much of it with interlaced video is a bad thing.
Even when shooting 25p or 30p and authoring a 25p or 30p BD disk if the player feeds interlaced to the display you're going to be in trouble. I have never noticed line twitter on a true progressive display.
I could be wrong here but I'm pretty certain that the camera outputs interlaced over its component output as well.

Piotr Wozniacki · November 19th, 2009, 07:23 AM

Yeah, Bob and myself went through the 25p-mode line twitter problem with the V1E(P) camera 2 years ago, and discussed it in detail on the relevant forum of DVInfo.

What I could add now is this: the PAL version of the (otherwise great) V1 camera did really suffer from this problem to a much, MUCH greater extent than the EX cameras do, even though its vertical resolution has been quite a bit lower. Bob even proposed a very insightful explanation of this phenomenon (to sum it up, the frequency filtering being too weak in 1080/25p mode when compared to the NTSC version due to it being dependent on the V1 SD PAL being 576p vs. 480p of its NTSC version; Bob please correct me if I summarized it wrong). That would make sense, as the EX do not do SD...

Frankly speaking, ever since I started using my EX1 I never *ever* have come across this problem (neither in the native 1080/25p material, nor its BD renders - heck, not even after down-rezzing to PAL DVD, which sounds to create a real disaster to so many people).

But that just me; perhaps I have taken the V1E lesson seriously enough, and know how to avoid the problem (display device settings being the most important factor)...

Piotr

Bob Grant · November 19th, 2009, 02:18 PM

Your analysis of the V1 situation is the same as mine was.
In fact at the time I produced a test pattern of alternating lines of black and white. You cannot get an image of higher resolution than this in 1080p. Our good friend Spot took that test pattern around CES to see what the HDTVs on offer made of it. Something like 80% of them failed to display the test pattern correctly.

This seems a very logical approach, take the camera out of the equation. Use the standard ISO res chart or a zone plate test pattern. Drop that into your NLE and render it out to whatever and view it on various display devices. Ideally take the test pattern into your NLE at higher than target resolution and do a small vertical pan in the NLE.

I don't have access to a lot of HDTVs and BD players to run this kind of testing myself. The limited tests I have done do show problems with most consummer displays. It's only displays that do genuine pixel to pixel that seem to cope OK.

On the other hand the BBC's tests of the EX cameras show the camera it be very artifact free. If I recall correctly the same paper also tested the Z5 and found it to have some issues. It would seem reasonable to say the EX produces the highest resolution clean image of any camera at anything like the same price point. Like Piotr I have yet to have an issue producing clean 50i SD from my EX1 however I am always shooting HQ 50i. The times I have had an issue that I had to wrangle I was shooting 720p50. The problems people are having seem to stem from shooting progressive.

It is simple video science. When shooting interlaced the camera using line pair averaging which reduces vertical resolution to around 70% of full raster. Note carefully, it does this to the full raster image. If you want to shoot progressive and still deliver images that'll survive any form of interlacing then you must emulate this in your post workflow. Trying to wrangle the issue after downscaling and you will be in trouble. It's not that easy with most NLEs to determine the order in which things are done as mostly it doesn't matter. Apply color correction before or after downscaling matters little. The same does not apply to controlling issues with aliasing or line twitter.

When I have had to wrangle these issue I used a HD project, apply vertical GB at full raster and then downscale. Remarkably enough after downscaling I may apply an Unsharpen Mask FX to emulate the 'look' of SD video. The order in which this is done is critical. Add edge detail before rather than after downscaling and a jittery mess is the outcome.

November 19th, 2009, 01:07 AM	#16
Alister Chapman Inner Circle Join Date: Dec 2005 Location: Bracknell, Berkshire, UK Posts: 4,957	I would place money on it being monitoring. There are very few consumer TV's that do a good job of EX footage. Try turning down any detail or sharpness settings on the TV. __________________ Alister Chapman, Film-Maker/Stormchaser http://www.xdcam-user.com/alisters-blog/ My XDCAM site and blog. http://www.hurricane-rig.com

November 19th, 2009, 01:55 AM	#17
Vincent Oliver Trustee Join Date: Sep 2004 Location: Chislehurst, London Posts: 1,724	Just to add my two pence worth. I also had twitter problems which were very noticeable when shooting handheld. Slight movement on the camera caused any fine detail to shimmer. Using a tripod cured a lot of the problems. I am still working on acheiving a better result, but I can live with what I have now. __________________ Eyes are a deaf man’s ears. Ears are a blind man’s eyes

November 19th, 2009, 07:23 AM	#19
Piotr Wozniacki Inner Circle Join Date: Aug 2006 Location: Poland Posts: 4,086	Yeah, Bob and myself went through the 25p-mode line twitter problem with the V1E(P) camera 2 years ago, and discussed it in detail on the relevant forum of DVInfo. What I could add now is this: the PAL version of the (otherwise great) V1 camera did really suffer from this problem to a much, MUCH greater extent than the EX cameras do, even though its vertical resolution has been quite a bit lower. Bob even proposed a very insightful explanation of this phenomenon (to sum it up, the frequency filtering being too weak in 1080/25p mode when compared to the NTSC version due to it being dependent on the V1 SD PAL being 576p vs. 480p of its NTSC version; Bob please correct me if I summarized it wrong). That would make sense, as the EX do not do SD... Frankly speaking, ever since I started using my EX1 I never ever have come across this problem (neither in the native 1080/25p material, nor its BD renders - heck, not even after down-rezzing to PAL DVD, which sounds to create a real disaster to so many people). But that just me; perhaps I have taken the V1E lesson seriously enough, and know how to avoid the problem (display device settings being the most important factor)... Piotr __________________ Sony PXW-FS7 \| DaVinci Resolve Studio; Magix Vegas Pro; i7-5960X CPU; 64 GB RAM; 2x GTX 1080 8GB GPU; Decklink 4K Extreme 12G; 4x 3TB WD Black in RAID 0; 1TB M.2 NVMe cache drive

November 19th, 2009, 07:00 AM	#18
Bob Grant Trustee Join Date: Nov 2005 Location: Sydney Australia Posts: 1,570	I've only had one bad line twitter problem with footage from my EX1. I shot 720p50 and the small hall had dark brick walls but the bricks had a very fine white line running though them. The walls shimmered something fierce. The problem was wrangled easily enough using Vegas's Gaussian Blur applied in the vertical direction only. This filter has been my goto for years, well before the EX1 was even thought of. Vegas seems to have much the same flicker reduction filter as Ppro and it just too much blur to be usefull. I can't find the same quality of GB filtering in Ppro CS3. I'm pretty certain though that FCP has a GB filter that has both vertical and horizontal blur control. As I'm not a FCP user I cannot comment on how effective this is. The line twitter / aliasing issue is nothing new, I and others went through the same issue with the V1P shooting 25p, the V1U produced slightly lower res and seemed not so prone to provoking this issue. Ultimately this is not the fault of the camera, I see a lot of the same problem making it's way into OTA broadcast material. Shooting hand held would make the issue more noticeable, more movement of fine detail vertically is going to make the problem more noticeable. Giving up on shooting hand held seems a backwards way to wrangle this issue. Once you recognise what the cause is it's then much easier to deal with it and I've yet to find any well reasoned argument to suggest the problem lies in anything other than vertical resolution, too much of it with interlaced video is a bad thing. Even when shooting 25p or 30p and authoring a 25p or 30p BD disk if the player feeds interlaced to the display you're going to be in trouble. I have never noticed line twitter on a true progressive display. I could be wrong here but I'm pretty certain that the camera outputs interlaced over its component output as well.

November 19th, 2009, 02:18 PM	#20
Bob Grant Trustee Join Date: Nov 2005 Location: Sydney Australia Posts: 1,570	Your analysis of the V1 situation is the same as mine was. In fact at the time I produced a test pattern of alternating lines of black and white. You cannot get an image of higher resolution than this in 1080p. Our good friend Spot took that test pattern around CES to see what the HDTVs on offer made of it. Something like 80% of them failed to display the test pattern correctly. This seems a very logical approach, take the camera out of the equation. Use the standard ISO res chart or a zone plate test pattern. Drop that into your NLE and render it out to whatever and view it on various display devices. Ideally take the test pattern into your NLE at higher than target resolution and do a small vertical pan in the NLE. I don't have access to a lot of HDTVs and BD players to run this kind of testing myself. The limited tests I have done do show problems with most consummer displays. It's only displays that do genuine pixel to pixel that seem to cope OK. On the other hand the BBC's tests of the EX cameras show the camera it be very artifact free. If I recall correctly the same paper also tested the Z5 and found it to have some issues. It would seem reasonable to say the EX produces the highest resolution clean image of any camera at anything like the same price point. Like Piotr I have yet to have an issue producing clean 50i SD from my EX1 however I am always shooting HQ 50i. The times I have had an issue that I had to wrangle I was shooting 720p50. The problems people are having seem to stem from shooting progressive. It is simple video science. When shooting interlaced the camera using line pair averaging which reduces vertical resolution to around 70% of full raster. Note carefully, it does this to the full raster image. If you want to shoot progressive and still deliver images that'll survive any form of interlacing then you must emulate this in your post workflow. Trying to wrangle the issue after downscaling and you will be in trouble. It's not that easy with most NLEs to determine the order in which things are done as mostly it doesn't matter. Apply color correction before or after downscaling matters little. The same does not apply to controlling issues with aliasing or line twitter. When I have had to wrangle these issue I used a HD project, apply vertical GB at full raster and then downscale. Remarkably enough after downscaling I may apply an Unsharpen Mask FX to emulate the 'look' of SD video. The order in which this is done is critical. Add edge detail before rather than after downscaling and a jittery mess is the outcome.