NOTICE FROM ADMIN
 

Note from Dvi forum administrator Chris Hurd: This entire thread is an unfortunate occurence which resulted from my own inability to keep an eye on all things at all times. There is some serious misinformation here, and not only is the discussion now locked, but I'm considering deleting the entire mess. Part of my goal for this forum was to create a discussion area for reliable, accurate information. What follows is neither: the claim that the PV-DV953 is "not a true 3-CCD camera" because it uses an established, proven 20-year-old technology called Pixel Shift is absolutely, patently false. Of course the PV-DV953 is indeed a true 3-CCD camcorder, with the CCD block arranged in the standard arrangement of a prism directing light to individual image sensors based on wavelength. Pixel Shift technology involves a precise alignment of the green CCD which is offset from the red and blue CCD's by one-half pixel in both vertical and horizontal axes, with the offset happening either electronically or physically or both. The following claim that Pixel Shift results in inferior color reproduction due to higher sampling of the green CCD is also false... the green CCD also provides luma (brightness) information, and not "double green" as the post below suggests. Pixel Shift is not marketing hype by any stretch of the imagination, rather it is a very real and very effective and very much established Panasonic technology. For those who need empiric evidence in order to disbelieve this most recent conspiracy theory, simply walk into to any retail store which has a Canon GL2 on display with power. Compare the superb color reproduction of the GL2 with any single-chip camcorder, and you'll see for yourself how the GL2's Pixel Shift does NOT affect color reproduction whatsoever. As I said, I'm not sure if I'm going to let the misinformation which follows stand up for much longer, but for now here it is, anyway.

Chris Hurd
San Marcos, TX

Original thread:


Panasonic PV-DV953:

It is not a true 3-CCD camera; the way the CCD's are placed creates the same image as a 1-CCD camera with 3x the resolution of one of the chips. Just because the marketing people claim that this is the same 3-CCD system as on their pro cameras does not mean that it is so.

This is one of Panasonic descriptions: “Three CCDs, each with 800,000 pixels, combine with Panasonic Quad-Density Pixel Distribution Technology to deliver true 3-megapixel still images. The result is full-fledged digital camera performance that greatly expands your expressive possibilities.“


1 CCD vs. 3 CCD:

In a true 3 CCD system the CCD's are aligned precisely so that pixels of each color component is being read at the same time. That is what makes 3 CCD camera color better, although on DV the color is quite compressed (4:1:1) so the improvement with 3 CCD's is not as great as on 4:2:2 compression scheme.

What Panasonic did was they shifted the 3 CCD's so that it creates illusion of one CCD with 3x the pixel count. This is great for stills but for video it fully defeats the 3-CCD advantage.

The video camera color properties are equal to a 1 CCD system.

1 CCD uses individual filter on each pixel; 3 CCD cameras use a prism and 1 filter for each CCD. If 1 800 K pixel CCD only was used, of the same size, as one of the 3 CCDs, we would have the same light sensitivity. The 3 CCD system does not make the camera more light sensitive since the prison needs to split the color into 3 components.

There are always potential problems with the prism quality, alignment, etc., so for a camera in this price range one CCD may possibly be an advantageous.


Comparison to Sony PDX10:

Sony does not record letterboxed signal on PD150, and maybe on all their cameras; I am not sure. They process the signal before recording so they use the whole frame for recording. This means less compression than e.g. letterbox mode on DVX.

This camera can't be compared to PD10. PD10 has 500-line resolution viewfinder -- it is B/W so the pixel count does not have to be as high as with a color one for the same resolution. PD10 LCD has something like 800 lines (as far as I remember) horizontal resolution. They achieve it by special pixel arrangement, so the pixel count again does not have to be as high as on normal pixel arrangement.

DVX has about around 300-line resolution on LCD and viewfinder, as far as I can tell. This little thing will at most have the same display resolution -- my guess. Check the DVX and this thing pixel count for the viewfinder and LCD. If it is color and there is no note of special pixel arrangement to boost resolution, and the pixel count is similar as on the DVX, the resolution will be similar too.

Should I go on?

Would I buy it? If you want a camera to shoot your kids outdoors and take still pictures, fine. Pro application? Definitely not! 15 Lux rating? I would have a serious issue with that.

Its an interesting claim panasonic makes that its the same ccd block in the 953 as that used in their pro cams, I did'nt realise how bad their pro cams are.

Question: What do you mean? Then even the XL1s is a 1CCD camcorder or something, since it too uses the similar technology(called Pixel Shift)?????

XL1s is an older design and the pixel count is low so to aciheve the full luma (B/W) resolution, the pixel shift is used. And yes, it makes it into a 1 CCD camera. There is no pixel shift on CineAlta, Varicam, or even the least expensive pro camera -- PDX10, which I incorrectly referred to as PD10 in my post.

Canon used the pixel shift to get a good luma (B/W) resolution when they came out with their XL1; they had to; they do not manufacture CCDs and probanbly had no alternatives. Panasonic's use of this chip arrangement serves a different purpose -- 3 CCDs sell and the shift makes better stills. But both camera's color performance should be considered equal to 1 CCD cameras.

this sounds like its getting in to the argument of mine is bigger than yours...
You know, it saddens me to think that people arent appretiative of the fact that this camera DOES take good video.

Low light is all good, but in all honesty, who here is shooting in zero light? Why would you WANT to shoot in zero light should be the question..

Dont get me wrong there are times where lights jsut cannot be used, and sure teh MX is a little weak in that department, but i have seen afew cameras which dont even go this far...
On top of that, bad lighting can be fixed in post if needed...
IMHO Its one thing not to use lights, its another thing to use lights effectively, and most clients dont mind compromising if it means theyre getting a better end product.

either way, for this price, it does what it says.. in effect 3ccd means 3 seperate colour streams...
whether or not the canon or the pana dvx and pro cams do this, the fact of the matter is, its a camera and it films decent footage..

if people dont like it, dont buy it.. simple...

If people are wary, go out and test it in a dark room. Most video suppliers have store rooms so play around with ANY cam before you buy it...

Im a newbie here, but ive been on boards such as this one for years... I also run my own tech forum...
one thing ive noticed about this forum in particular however, is that many people forget to realise their luck in the fact that they are doing something they love, and that is creating...

I just find it disheartening that and feel that there are many more things to worry about than lines on an LCD viewfinder or pixel shifting technology which is used by practically every manufacturer out there...

By the way, pixel shifting technology is used to to apply a movement process to the framebuffer.. basically creating a buffer of the image PRIOR to recording to tape (or taking a still), which in turn optimises teh picture quality...
Electronic image stabilisers work in a similar fashion.

This "movement" could be an actual moving image, or a still buffer or as some say a "Hyper Interlation" (cant remember where i heard it, but it describes teh process)
This is why the MX takes decent stills.
The cam apparently "pixel shifts" the CCDs to give a still resolution of 2048 x 1496.

Some use larger CCD's (like Sony) To achieve a similar result, but in turn, this boosts prices...

Basically pixel shifting makes up for the smaller CCD.
Pixel shifting doesnt cost much to develop compared to the Hadware equivalent, and as CCD's get smaller, we'll be seeing alot more processing like Pixel Shifting being used to fill "the gap"...

I dont see the point in complaining about pixel shifting, considering that its been used for High Definition recording and will be seen alot more in the future as High Definition becomes readily available and CCD's decrease in size to cut production costs...
I mean we're already seeing it, the HD cams are runnign single CCD's but have a boosted "Pixel Shift" processing routine to make up for it...

technology will always change, and there will always be different ways of doing things..

Deciding on what to buy in accordance with Specs alone is silly IMO
Every cam has its weakness and strength, and as far as advertising and promo blurbes go, any company that says their machine does something, must in fact do it, else its a breach of Fair trading practices.
Like the music industry which i have been a part of for almost 10 years, many people complain about what synth or sampler can do what.

But in the end, the Camera is just a tool and YOU are the artist.. Not the other way around. People seem to forget that.
I know film makers who cant afford a second hand cam, and from seeing that, i just think we should all be greatful for what we have... and if we can afford any better, even better.

Thats just a comment in general and not an attack on anyone, so i dont want that to be misconstrued in any way...

thats my 2 cent ramble :)

Not aligning the pixels of the 3 CCDs defeats the purpose of 3 CCDs to deliver better color.

You need to get the highest quality camera you can. Photographer can pick up $1000 still camera and use it to compete with the top pros. Video or movie maker must pay $100.000 or more so he can do the same. That's why all the fuss about getting the best tool (camera). You can't paint a decent picture if your colors are crappy and you only have one large paintbrush, unless you are going for the Blair Witch type art.

If I wanted perfect video, I'd use a motion film camera and shoot in 30P. Then I'd have it transferred to interlaced video. But of course I cannot afford this, few can, and I have no need to do this when many of today's consumer/prosumer cams capture great footage. Instead of worrying about perfect color, I suggest shooting for beefier audio and implimenting smarter skills such as lighting, interesting angles, dolly and jib arm techniques. But all this is nothing if you don't have THE STORY. Sorry.

Mr George did you switch your camps? It looks like you're employed by Sony or their marketing people this week.

I never worked for JVC but was hired by a PR firm that worked with a company that was going to do modifications to this camcorder. Those plans ended when they found out of the poor low light sensitivity of the JVC. So they cancel having the camera air shipped from Japan. Then, after everything was cancelled came even worse news of the washed out colors and of a planned blu-ray HD DVD camcorder by a competitor.

Joseph,

What? Your post (same as on DV.com) just rambles on and on to what appears to be at best an estimated guess. Its ok to spin your gears but just make sure they have teeth. We all know that the 3MP stills on the DV953 are bogus but what hard evidence do you have that the video portion of the DV953 is equivalent to a single chipper?????

Check out the brochure tag from my DV953. Sure looks like a typical configuration to me.


http://image1ex.villagephotos.com/2536345.jpg




I'm cool with the rants but lets keep it professional.

Let me explain it differently. On a single chip camera each pixel is at a different location. The way the system works, you need to read all colors at the same time. It is different than with still digital cameras. That's why pro cameras have 3 chips and the pixels of all the chips align so at any time you read the pixels of all 3 colors. By shifting the pixels you basically create an imperfect 1 CCD camera since you can never control the pixel offset as precisely as you can on a 1-chip CCD. It is actually a disadvantage to use 3 chips in this type of setup. Now tell me how are you going to guarantee precise alignment of the three CCDs, each having nearly 1,000,000 pixels and each having diagonal size 1/6". How can you guarantee the prism's absolute accuracy? You simply can't.

So the problem is 2-fold. 1. The 3 CCD arrangement does not improve color over 1 CCD setup. The (3) 800K pixel CCDs arrangement creates less than perfect 2.4 meg CCD for stills.

In regards to posting the same thing at DV.com. The guy there got the camera and did excellent testing and comparison to an older Digital 8, and the old camera, which couple years ago cost a lot less than the new Panasonic costs now, won without any doubt.

The guy at DV.com retersted the camera after reading the manual and said that the image is amazing.

What does it mean? Excellent picture on 1 CCD camera. On a 4:1:1 compressed color system, which DV is, it does not make much difference if 1 or 3 CCD's are used, especially with a megapixel camera.

I'm still not buying into this theory of yours.

The GL2 has the same horizontal and vertical pixel shift as the DV953 and you don't hear anyone rallying about it not being a true 3CCD cam.

What cams have you compared the DV953 up against? I have footage from Canon Optura Pi, JVC GR-DVL9000U, Sony VX2000, Sony DCR-TRV120, and Panasonic MX1000. The DV953 produces more accurate color than all of these cams save for the VX2000. I will note that the MX1000 is in a very close third place. I'm comparing these on an ISF calibrated HD Wega.

http://www.canondv.com/gl2/f_pixel_shift.html

Somewhere I though I read that pixel shifting helped to give more of a film grain effect, as opposed to having them all aligned in the same rows and columns.

>The guy at DV.com retersted the camera after reading the manual and said that the image is amazing.<

But wait, the story continues. Now he wants to return it and buy a TRV950, because it's got better 16:9 and 1/2 a LUX more (?) Also, he inferred the 953 is too complicated, that's why he still uses Ataris. Go figure. (That's why I'll never sell my DVL9500 cams. It don't matter, not even in a 24P HD world around the corner. I'm keeping them, and they're going in the grave with me!)

Pixel shifting or not -- do some search on the Net guys -- I learned it in a Kodak seminar; once you shift the pixels it becomes a 1-CCD camera colorwise -- the colors need to be aligned precisely to take advantage of the 3-CCD system. CineAlta and Varicam are aligned precisely; the other ones -- I don't know. But on DV it does not really matter much. One CCD or 3 CCDs -- it does not make much difference because of the color compression.

Joseph, video is far more forgiving than film---especially stills. Trust me. Even bad video can look good, not so with pictures.

Dear everyone.

I am getting dizzy reading all these posts.

But I thought DV953 / MX5000 use image shift only when you take still pictures. It doesn't use image shift when you take movies.

Am I correct on this?

I am not sure why Joseph is saying that DV953 isn't true 3CCD Camcorder.

DV953 has enough resolution to take movies (800k x 3 CCD) without shifting image and DV953 uses 3CCD to enhance 800k image to 3meg still pictures. Of course it isn't true 3meg pictures, that's because CCD is only 800k size. Image shifting enables 3meg pictures.

This thread is going around and around.

First, why should I trust the subjective visual observations of someone who buys cameras with the sole intent of just playing with them and returning them. This is suspect in my opinion, along with his testing methods.

Second, Joseph, can you provide a link to some other source for your contention that 3 CCD's act like 1 CCD when pixel shift technology is employed? If Kodak brought this up at a seminar it must be published somewhere. If you can't find a link to substantiate your claim, then I think people are getting confused over nothing.

Guys, I'm leaving town -- starting new project; don't have time. Kodak is quite advanced in CCD technology and quite helpful. Get on their site. They must have something to that effect. Do search on Google or maybe Kodak site on: Kodak CCD theory, operation, engineering, or something. I want to stress that it does not really matter much if you use 1 or 3 CCDs, especially not in DV, MPEG2 Micro DV, MPEG2 DVD, or MPEG2 HD DV. The colors are significantly compressed. It would make the most difference on 4:4:4 color system, where there is no color compression. On 4:1:1 and 4:2:0 systems you compress chroma (color) 4x compared to luma (B/W), so it does not really matter. Once you get to megapixel CCDs with resolution a lot higher than the format, it matters even less. Dalsa is 8 meg camera and is using 1CCD. Using 3 CCDs on consumer cameras has more to do with marketing and being able to take good stills. Anyway the 953 seems to be a nice camera from what the guy on the other board says and because of the manual controls it probably is the best buy in that price range, followed by PDX10, DVX100, HD10, if JVC fixes up the color problem. Al these are nice cameras for video production where you have enough light, with lux rating 15 for 953, 7 for PDX, 24 in progressive for DVX and 35 in HD for HD10. DVX is 3 Lux in interlaced but I'm adding it here because of the progressive mode, where it has increased subjective resolution over DV. 953 and PDX have increased 16:9 resolution. HD10 is HD.

Joseph, I don't understand were you get all those "strange" thoughts. (Green) pixels shift techniques have been patented about 20 years ago and has been accepted in the pro cam world as a method which allows the difficult trade-off of optical aliasing vs luma resolution limits for a given (3CCD) structure to be shifted towards a higer luma resolution picture. If you persist in yr belief that this structure is a 1 CCD equivalent you must at least know that yr 1 CCD has 3 times times the surface and fill factor(sensitivity) and three times the amount of pixels (resolution) and most important that you have to invent a 1 CCD structure with partly overlapping pixel structures(which is optically the case in 3CCD pixel shifted structures), not co-sited like in 1CCD devices with filter.

Got tired of explaining this. We are basically recording luma and 2 compressed chroma signals. On one CCD system you get these signals off of a single chip and there is a space shift among the pixels whose signal you use to combine into signal to be recorded. That is the reason that 3 CCD's are used, so you get the signal from all 3 components at the same time, in the same space. You use these signals to create the 4:1:1 signal. If the CCD does not have enough resolution for the format, like on Canon XL1/XL1s, the manufacturer selects to use physical shift among pixels. This allows higher luma resolution, but worsens the color resolution.

Example. There is a boundary of blue and green in the picture. On a true 3 CCD system all 3 sensors will see the same boundary. On a shifted pixel sensor, near the boundary one color sensor may se the blue, one the green, and one half of each. By combining these signals, of 3 sensors that are not at the same location, you get inaccurate color. It is that simple.

All 3 of the 4:2:2 signals are recorded simultaneously so you need to feed them with simultaneous information that is same in space and time.

As I said, in DV it does not really matter much. But the 3 physically shifted CCDs do resemble 1 CCD less the double green pixel count. The color accuracy is equal to 1 CCD sensor. The large pixel count naturally helps.

I had the camera quality backwards in my previous post. The low end is 953; the high end is HD10.

Since you're back in town from your project can you find a link to substantiate your post, please? I searched Kodak's site and found nothing. I hope you'll have better luck.

Just to continue the other story, the dv.com member, Lyonhart, exchanged his faulty PV-DV953, and now is happier (with the replacement). All those thousands and thousands of words he wrote, and complaining about the wasted days used to write these words, it never dawned on him that there was something wrong with his cam. Go figure!

Joseph, you are dead wrong. Andre De Clerq already explained why; and I value all his posts, since he knows what he's talking about, all of the time. He is our Adam Wilt and Peter Utz.

Jeff: If you can't find it, I'm not going to look for it. I already explained it. Talk to someone else who knows.

Note from forum administrator Chris Hurd: It is ALWAYS the responsibility of the person making the claim, and not the person reading the claim, to substantiate that claim with evidence. I am very suspicious of anyone who says "you go look for it" in response to a query asking for documentation supporting their claim. The responsibility lies with whoever makes the claim, and if they're unwilling to do the legwork to subtantiate it, then I doubt it's for real. -- CH

Frank: It appears that you don't have a clue what Andre is talking about.

Joseph, I said you were wrong: a 3 CCD cam using pixel shift is still a 3 CCD cam, which captures superior/more accurate color than a 1 CCD cam. Since you said otherwise, the task is yours alone to prove.

Re: "...3-CCD camera; the way the CCD's are placed creates the same image as a 1-CCD camera with 3x the resolution of one of the chips." ---Jospeh George

Frank Granovski: Joseph, I said you were wrong. A 3 CCD cam using pixel shift is still a 3 CCD cam, which captures superior/more accurate color than a 1 CCD cam. Since you said otherwise, the task is yours alone to prove.

Yes, you can call it a 3 CCD camera. It uses 3 CCDs, but the color, with the shift, will be same as on 1 CCD camera. The luma will have greater resolution, which will allow you to advertise it as a 2.4 megapixel digital still camera, but even that will not be 100% true, as the pixels will overlap and the image will be worse than from a true 2.4 megapixel camera, with the double No. of green pixels, etc.

The you said I said response:


Re: "Anyway. Believe what you want. You obviously do not understand the technical issues...Enjoy your little toy."

No need to get defensive. I simply said that you are wrong.

Re: "If you believe that the camera has the same color as one that has the pixels properly aligned for the purpose of color improvement, fine.

That's what you think I said, but that is not what I said. This is what I said: "a 3 CCD cam using pixel shift is still a 3 CCD cam, which captures superior/more accurate color than a 1 CCD cam."

You said: "3-CCD camera; the way the CCD's are placed creates the same image as a 1-CCD camera with 3x the resolution of one of the chips."

Joseph. if you got tired of explaining your story, which I suppose is based on some knowledge you got from amateur magazines, you should add this:

1. There is a huge difference between co-sited and shifted pixels w.r.t. the potential risk(see further) of shifted colors and color resolution under certain conditions.

2. Though you mention 4:1:1/4:2:2 you seemingly don't relate this to your story. If you considder a 1 to 1 structure (which is not allways the case) you got to know that the color components (Cr/Cb) are not based on a single pixel sampling but composed out of 4 surrounding pixels. Resulting in a general color blur which is much higher with co-sited(1CCD) structures than with overlapping or shifted 3CCD structures.

3. Though pixels are not "on the same location" you should know that in the mapped (subsampled CrCb) form, there is no problem to slighly electronically shift the pixel timings in order to get optimal Y/C outputs.

All by all Joseph, you should realise that the color resolution in a 3CCD using shifting techniques is not at all different from a non shifted one. The only thing what can happen is a sub optimal shift compensation during the further processing. And ...if you see color fringes, remember that chromatic aberations, Y/C delay mismatches in color encoders/decoders and convergence errors in CRT type displays are far more visible the your color registration story.

Again you are right the 3CCD will act like a single CCD if you can build the CCD which I mentioned earlier..and this will get you a Nobel prize.

Thanks for the clarification and knowledgable response, Andre.

Since it appears nothing new will be added and to avoid covering the same ground this thread is locked. If any members have additional questions please start a new topic.

Note from Dvi forum administrator Chris Hurd: This entire thread is an unfortunate occurence which resulted from my own inability to keep an eye on all things at all times. There is some serious misinformation here, and not only is the discussion now locked, but I'm considering deleting the entire mess. Part of my goal for this forum was to create a discussion area for reliable, accurate information.

What transpired is neither: the claim that the PV-DV953 is "not a true 3-CCD camera" because it uses an established, proven 20-year-old technology called Pixel Shift is absolutely, patently false. Of course the PV-DV953 is indeed a true 3-CCD camcorder, with the CCD block arranged in the standard arrangement of a prism directing light to individual image sensors based on wavelength. Pixel Shift technology involves a precise alignment of the green CCD which is offset from the red and blue CCD's by one-half pixel in both vertical and horizontal axes, with the offset happening either electronically or physically or both. The claim above that Pixel Shift results in inferior color reproduction due to higher sampling of the green CCD is also false... the green CCD also provides luma (brightness) information, and not "double green" as the post below suggests. Pixel Shift is not marketing hype by any stretch of the imagination, rather it is a very real and very effective and very much established Panasonic technology.

For those who need empiric evidence in order to disbelieve this most recent conspiracy theory, simply walk into to any retail store which has a Canon GL2 on display with power. Compare the superb color reproduction of the GL2 with any single-chip camcorder, and you'll see for yourself how the GL2's Pixel Shift does NOT affect color reproduction whatsoever. As I said, I'm not sure if I'm going to let the misinformation which follows stand up for much longer, but for now here it is, anyway.