New Sample Footage

[Ken Ross]

Well, regarding firmware C, I can say that my FX7 does not have a consistent level of noise. If one goes up to my 50" plasma and looks at large expanses of sky, you can see a bit of grain. However, this is totally unnoticeable at anything approaching normal viewing distances.

For most scenes you simply can't see noise in reasonable lighting. I still think the biggest issue (and it's not really a 'big' issue) is the occasional EE under certain conditions (tree limbs and wires against a sky).

However, I think your overall theory has some merit.

[Bob Grant]

Steve,

I think you're right, well I'd suspect that the rock and the hard place is the noise from the sensors and what the mpeg-2 encoder can cope with.

I'd add that I've seen similar smearing or smudding or banding or whatever it's being called on other Sony CMOS HDV cameras, the HVR A1P has caught me out when I made a boo boo and had too much gain dialled in.

Can I suggest hooking the camera up via component into a big monitor and looking at what you see there under low light straight from the camera head, i.e. not from tape, just live.

[Steve Mullen]

Originally Posted by Ken Ross

Well, regarding firmware C, I can say that my FX7 does not have a consistent level of noise. If one goes up to my 50" plasma and looks at large expanses of sky, you can see a bit of grain. However, this is totally unnoticeable at anything approaching normal viewing distances.

Great news! That's means the reviewer was wrong, which is great because for those who only want 60i -- the FX7 is a much cheaper camcorder.

I'll alter my posting.

[Steve Mullen]

Originally Posted by Bob Grant

Steve,
I think you're right, well I'd suspect that the rock and the hard place is the noise from the sensors and what the mpeg-2 encoder can cope with.

Can I suggest hooking the camera up via component into a big monitor and looking at what you see there under low light straight from the camera head, i.e. not from tape, just live.

I just got an email from someone who claims noise and paint is not an issue on his CRT monitor, but it is on his LCD. I selected a Sony A10 LCD HD monitor that correctly deinterlaces video. However, the majority of the tested 2005 and 2006 fail the deinterlacing test.

UPDATE
See:

http://digitalcontentproducer.com/hd...work_01082007/

If DNR has removed detail in an area -- and if the monitor uses "bob" deinterlacing -- the paint effect will be increased. Since the majority of HDTVs and softer de-interlacers use "bob" -- I'm not surprised some are seeing a more of a problem than others.

Since I specifically bought an HDTV that correctly deinterlaces -- we need to factor that into the equation. Not only do I have a prototype -- my monitor is one of those that pass the deinterlacing test. My quality descriptions ONLY come from watching on my HDTV monitor. I never use a computer to judge quality!

[Ken Ross]

I had once mentioned that the viewing device is VERY critical in assessing the quality of the video. If a display introduces noise and artifacts as SOME displays do, those people will unjustly accuse the acquisition device and not the display.

[Mikko Lopponen]

Originally Posted by Steve Mullen

If DNR has removed detail in an area -- and if the monitor uses "bob" deinterlacing -- the paint effect will be increased.

No. The paint effect isn't noise, its areas that lack detail. Bob will not "increase" those areas in anyway. What happens when you take a 1080p still and resize it to 540? It will lose details EVERYWHERE, it won't increase the paint effect. It will actually decrease it as all the details get closer. If you don't believe then just try fooling around with photoshop and deinterlacing some of those paint examples.

LCD's will always have more noise. The blend example is wrong, but I posted a response to the other thread.

[Thomas Smet]

I think the progressive noise on the V1U is because of the water color effect and this is why:

It is due to dancing or chattering pixels. The water color effect is very very subtle on the V1U to the point where most people seem to not notice it. This of course is a good thing because it works since most of the people do not notice it. It is bad however because a water color effect that is very small will dance around from frame to frame and cause chattering that is so small it looks like noise.

This makes sense why as still images some of us notice a water color effect but when in motion very few notice it. When the video is playing all the dancing water color pixels look like noise.

On top of that of course the level of the water color filter is different for different sections based on who knows what. which can show some areas having more noise then other areas.

I checked this with the video clip of the church which I played with the VLC player in full screen on a 1920x1200 LCD monitor (letterboxed of course to fit in 1920x1080). When I paused the video I noticed the water color effect in the church and the clouds. When the video played I didn't notice the effect as much but the effect was replaced with dancing pixels or noise. The effect is sort of like when you have a bad bluescreen shoot and the edge of the key has jumping chattering edges and you end up using a matte choker to get rid of it.

This is my theory at least and it makes sense as to why SONY might suggest turning the sharpness down. If reduced sharpness could get rid of the tiny amount of water color filtering then there would be no filter effect to cause the chattering pixels.

As far as I can tell deinterlacing has nothing to do with this and the only reason why some deinterlacing methods seem to reduce the noise is because overall the image is softened slightly.

As for a CRT based HDTV, well CRT is a different beast then a digital display and CRT's are usually better at smoothing out details and noise whereas a digital display will show you every pixel. I think the advantge to the V1U on the CRT has to do more with the fact that it is analog rather than an interlaced device.

It sure would be nice if somebody with a V1U could shoot some stuff at the different sharpness settings to see if this works. I would rent one but nobody near me rents the V1U yet.

[Brett Sherman]

Originally Posted by Bob Grant

Steve
Can I suggest hooking the camera up via component into a big monitor and looking at what you see there under low light straight from the camera head, i.e. not from tape, just live.

That's what I first did to identify the problem. I hooked up the camera directly to my Sony LMD-232 and flipped between progressive and interlaced. So it's obviously not the HDV encoding that is causing the problem. The ambient light was rather low, so I had gain up to +6db. But, I was looking at the paint effect mostly, not noise. The paint effect was definitely present. So much so, that you could see it in the little 3.5" LCD also.

[Brett Sherman]

Originally Posted by Todd Giglio

So... if there is anyone who would like me to test the camera under specific situations (and I don't mean just good lighting, I mean they way that you think the camera shows these flaws) let me know. Give me a shot list and I'll see what I can do.
Todd

Architectural details seem to reveal the paint effect the most. Roofs, Building Facades, etc. Probably because they have straight lines with a consistent color that changes slightly because of lighting angle highlights or texture.

[Ken Ross]

Originally Posted by Thomas Smet

As for a CRT based HDTV, well CRT is a different beast then a digital display and CRT's are usually better at smoothing out details and noise whereas a digital display will show you every pixel. I think the advantge to the V1U on the CRT has to do more with the fact that it is analog rather than an interlaced device.

Not sure I agree there Tom. I've got a Sony Bravia LCD and a Fujitsu plasma, both fixed pixel devices. The LCD seems to introduce noise of its own and the Fujitsu tends to show a much purer picture. This can be seen with the same sources. So I would never use an LCD as a critical viewing device.

[Stephen van Vuuren]

Originally Posted by Brett Sherman

That's what I first did to identify the problem. I hooked up the camera directly to my Sony LMD-232 and flipped between progressive and interlaced. So it's obviously not the HDV encoding that is causing the problem. The ambient light was rather low, so I had gain up to +6db. But, I was looking at the paint effect mostly, not noise. The paint effect was definitely present. So much so, that you could see it in the little 3.5" LCD also.

I looked at your footage in Vegas (could not get my AE to read it) using color channel seperation, scopes etc. and the problem is clearly visible in the signal. It's unlike any HDV issue I've seen (no variation across GOPs etc) and not caused by display.

It's there in the I & P signal, so I would fully except it on direct camera output which would run through the DSP/sensor processing. I still think it's related to the NR falsely diaganosed noise and smoothing out detail.

[Michael Chenoweth]

I attended an HD shootout at the local reseller here in SLC today. Very, very impressed with the performance of the V1U against the other cams which included the SDX900, HVX200, XLH1 / A1, JVC 250 and Sony F350.

Lots of charts and low light. I must say that I'm very impressed with how the V1U performed. If I can get permission, I'll post some of their results which were all live to monitor.

Bob Grant posted some footage I shot yesterday.. and some shot awhile back. The trick was to be able to tell what was shot V1U / Z1U - it takes a bit more finessing, much like the high end cams. When calibrated correctly though, it out performed or equally performed with those cameras above aside from sensitivity, which is to blame on smaller chips but still didn't look bad.

So over the course of the past few days, I've come to love the camera much more. Still tweaking but for the most part, content now.

As for any anomalies, there may be some under different conditions. I do believe that for the most part, they're correctable. then again I could be wrong...

mike

[Ken Ross]

Originally Posted by Stephen van Vuuren

It's there in the I & P signal, so I would fully except it on direct camera output which would run through the DSP/sensor processing. I still think it's related to the NR falsely diaganosed noise and smoothing out detail.

But people aren't complaining about it the I signal.

[Steve Mullen]

Originally Posted by Mikko Lopponen

No. The paint effect isn't noise, its areas that lack detail.

LCD's will always have more noise. The blend example is wrong, but I posted a response to the other thread.

Blend is the correct deinterlace method for static video. It provides the maximum vertical resolution. You need to learn more about deinterlacing.

Of course, the water color effect isn't noise -- were the effect to REALLY EXIST, it would be the RESULT of noise REDUCTION -- specifically "coring" and line-based DNR. Coring and line-based DNR reduce noise by removing chroma detail. Coring makes any area with minor chroma variations -- become a uniform color -- hence a watercolor artifact. Line-based chroma DNR smears color detail vertically -- hence a watercolor artifact.

----------------

And, while Ken is right that LCDs introduce noise VS plasma -- this generalization is very risky. Walmat sells under $1000 plasmas that will have far more noise than a $4000 LCD/Plasma. Particularly suspect are the EDTV plasmas -- the type sold in Region 50. These are optimized for widescreen PAL DVD.

The only generalization that is valid: flat-panels generally have more noise than CRT HDTVs. Even more important, CRT HDTVs require NO deinterlacing. That eliminates a whole set of issues.

But the real issue is that these devices differ significantly by brand and model. Which means the very different reports of noise on, for example, the FX7 are not because the camcorder is different, but because the viewing devices are different. See Ken's post and camcorderinfo's review. Totally opposite on the claims of noise.

Bottom-line, when you add display variability into the different levels of Sharpness and Gain -- reports of noise simply can't be trusted.

That doesn't mean there isn't noise, but its very hard to prove given we have no objective way of measuring it. So I remain open on the question. I will say that on a good display, it doesn't show-up.

[Steve Mullen]

Originally Posted by Brett Sherman

Architectural details seem to reveal the paint effect the most.

Brett, I've downloaded all your clips and played them via HDMI to my 50-inch Sony LCD. Your video looks wonderful and has only three artifacts:

1) on the static INT>PROG sample -- at the switch point the video becomes slightly softer. This is consistent with my experience. There is no doubt that P-mode is softer than I-mode.

2) on the static INT>PROG sample -- at the switch point the branches get noisey. In particular, the horizontal branches. A display should switch to "blend" on static video. When I switch VLC to blend, the noise is eliminated.

What if there was some movement in the frame? Some displays would switch to "bob" and the noise would return. But, a good deinterlacer will use blend on the startic areas while switching to bob -- or something better ONLY for the moving AREA. (This is how my Sony works.) Even better deinterlacers will switch to bob -- or something better ONLY for the moving PIXELS.

What if the branches were moving? If the deinterlacer uses "bob" the line doubling will show the noise, but it will be hard to see since the branches are moving. If the deinterlacer uses something better than "bob" -- noise may not be increased.

Bottom-line -- I suspect the V1U may have some inherent noise, but the amount you see depends on the display.

3) The three clips where you used 9 obviously have significant EE. And that raises the question, with the EE so boosted, it's possible that this may be introducing other problems. Frankly, I think these should be removed from your site because they were shot incorrectly. Yet they are being used for many claims. And, that raises yet another question -- for the INT>PROG sample -- did you use 5? Because if you didn't, then this too is a bad sample.

-----------------

The good news is that for all your video samples -- and stills -- there is NO watercolor effect. Not on the church. Not on the shed. It's no where to be seen.

In fact in all the V1U samples I've downloaded, there is no watercolor artifact. Not one! I'm convinced that there is no watercolor artifact. There's certainly nothing in hours of 24p India footage that looks like a watercolor artifact.


I think it is time to stop repeating the claim of a watercolor artifact in posts with no reference to any evidence. Repeating a false claim has a way of making it seem true.

I will continue to pounce on any claim of watercolor artifacts where the poster provides no link to a CORRECTLY SHOT VIDEO (not still) clip and no description of "where" and "what" the poster is seeing.

I've got 4 months experience with the V1U -- and I've never seen anything except softer P-mode video. Not even any noise.