HVX ccd size and pixel count, what is it?
- Thread starter mthornton
- Start date
thisiswells
Retired
mthornton said:
Anyone have a clue?
Barry_Green
Moderator
The announced details are three newly-designed 1/3" 16:9-shaped high-def CCDs.
Exact pixel count has not been announced yet.
Exact pixel count has not been announced yet.
My finger are crossed, I just hope that it is at least 720p HD pixel count per chip.
Would they try to protect the SDX-900 and short the HVX user then?
Some interesting questions running through my mind.
I am thankful for 4:2:2 color space, I guess we can't ask for too much.
Mthornton
Would they try to protect the SDX-900 and short the HVX user then?
Some interesting questions running through my mind.
I am thankful for 4:2:2 color space, I guess we can't ask for too much.
Mthornton
Barry_Green
Moderator
They will absolutely not try to "protect" the SDX900 by "shorting" the HVX -- I mean, think about it -- the HVX is high-def and the SDX is standard-def-only. If that was the only thing that mattered, then the SDX is the one getting shorted!
As far as pixel count -- I've been thinking a lot about this, and I know Graeme has as well. And the more I think about it, the more I think Panasonic should probably never reveal what the pixel count ends up being -- that way we can focus on the footage, and not on who has bragging rights to the most-dense-pixel chip.
Because people just don't seem to understand -- they want to believe that more pixels = better, but they don't understand the physics and optics that go into this. It is, I believe, absolutely impossible to design a lens that could resolve enough detail for a 1/3" 1920x1080 chip. Let me put it this way: the $60,000 high-def lenses for the "big boys", the CineAlta and VariCam, resolve about 85 line-pairs per millimeter. A 1920x1080 CCD would require a lens that could resolve over 183 line-pairs per millimeter! It cannot be done!
The limiting factor isn't the chip pixel count -- you could theoretically design a CCD with 8000x6000 pixels on it, but if the lens can only feed it 1200x800 worth of resolution, then what's the point? The most you'd get out of your 8000x6000 ccd would be 1200x800 imagery. See what I mean?
And, in fact, that 8000x6000 CCD would deliver terrible video performance -- you'd have just horrible latitude and sensitivity. Horrible. A 1200x800 CCD would be so much better that it's not even funny.
So the only thing that knowing what the pixel count is would do, would be to give someone "bragging rights" -- as in, "My ccd is bigger than yours." Even if the footage looks worse! (which it would). You want to ask about the HVX user getting "shorted"? Putting more pixels than necessary on the chip would be the way HVX users would get "shorted" -- because that would mean that the video image was suffering, just to make a concession to the marketing department. And I really hope they don't do that.
I am sure that Panasonic will design the chip around the lens -- whatever the highest, sharpest resolution the lens can deliver, they'll design the chip to capture that resolution. They'll use the biggest pixels they can get (for the most dynamic range and best sensitivity and lowest noise) and the result will likely be the highest-resolution possible given the state of today's optics. What that pixel count ends up being is, ultimately, irrelevant. It's like Rolls Royce used to list in the spec sheets for horsepower: they'd say "adequate."
Instead of worrying about pixel count on the CCDs, I'd rather see resolution charts from the camera. Let's see what really matters -- how all the elements work together. If the HVX footage is higher-resolution in 720p mode than an HD100, and higher-resolution in 1080i mode than an Z1, then I don't care if they did it with a 300x200-pixel chip!
As far as pixel count -- I've been thinking a lot about this, and I know Graeme has as well. And the more I think about it, the more I think Panasonic should probably never reveal what the pixel count ends up being -- that way we can focus on the footage, and not on who has bragging rights to the most-dense-pixel chip.
Because people just don't seem to understand -- they want to believe that more pixels = better, but they don't understand the physics and optics that go into this. It is, I believe, absolutely impossible to design a lens that could resolve enough detail for a 1/3" 1920x1080 chip. Let me put it this way: the $60,000 high-def lenses for the "big boys", the CineAlta and VariCam, resolve about 85 line-pairs per millimeter. A 1920x1080 CCD would require a lens that could resolve over 183 line-pairs per millimeter! It cannot be done!
The limiting factor isn't the chip pixel count -- you could theoretically design a CCD with 8000x6000 pixels on it, but if the lens can only feed it 1200x800 worth of resolution, then what's the point? The most you'd get out of your 8000x6000 ccd would be 1200x800 imagery. See what I mean?
And, in fact, that 8000x6000 CCD would deliver terrible video performance -- you'd have just horrible latitude and sensitivity. Horrible. A 1200x800 CCD would be so much better that it's not even funny.
So the only thing that knowing what the pixel count is would do, would be to give someone "bragging rights" -- as in, "My ccd is bigger than yours." Even if the footage looks worse! (which it would). You want to ask about the HVX user getting "shorted"? Putting more pixels than necessary on the chip would be the way HVX users would get "shorted" -- because that would mean that the video image was suffering, just to make a concession to the marketing department. And I really hope they don't do that.
I am sure that Panasonic will design the chip around the lens -- whatever the highest, sharpest resolution the lens can deliver, they'll design the chip to capture that resolution. They'll use the biggest pixels they can get (for the most dynamic range and best sensitivity and lowest noise) and the result will likely be the highest-resolution possible given the state of today's optics. What that pixel count ends up being is, ultimately, irrelevant. It's like Rolls Royce used to list in the spec sheets for horsepower: they'd say "adequate."
Instead of worrying about pixel count on the CCDs, I'd rather see resolution charts from the camera. Let's see what really matters -- how all the elements work together. If the HVX footage is higher-resolution in 720p mode than an HD100, and higher-resolution in 1080i mode than an Z1, then I don't care if they did it with a 300x200-pixel chip!
thisiswells
Retired
I agree with Barry--I have no desire to know the pixel count on this cam.
I'll spend way too much time second guessing it and thinking:
"what if they did this, then I would have better sensitivity" or
"what if they did that, then I would have better resolution" etc.
I don't really want to think that way.
Don't ever reveal it, K. Sony never revealed the recording specs on HDCAM, BTW.
It was all theorised and conjectured and deduced to what we know now is "HDCAM"
but they never printed specs on the format. Good idea.
I'll spend way too much time second guessing it and thinking:
"what if they did this, then I would have better sensitivity" or
"what if they did that, then I would have better resolution" etc.
I don't really want to think that way.
Don't ever reveal it, K. Sony never revealed the recording specs on HDCAM, BTW.
It was all theorised and conjectured and deduced to what we know now is "HDCAM"
but they never printed specs on the format. Good idea.
Barry_Green
Moderator
I mean, it's not like we're not curious, but I look at it like this:
If I want a fast car, I want one that will go zero to 60 in maybe 4.5 seconds, right? If I shopped by horsepower ratings, I could find a Mitsubishi Lancer EVO with about 271 horsepower, or a Dodge SRT-10 truck with 500 horsepower.
Which is quicker?
Duh. The Lancer would smoke the truck.
So the question is, what are you wanting the camera to do: do you want it to deliver high-resolution footage? Or do you want to brag about pixel count?
For me, if I was buying a quick car, I'd want the one that was quicker, regardless of technically how many horsepower it outputs.
For the camera, I want the one that puts out the best, highest-def footage, regardless of pixel count.
That's all I'm saying.
If I want a fast car, I want one that will go zero to 60 in maybe 4.5 seconds, right? If I shopped by horsepower ratings, I could find a Mitsubishi Lancer EVO with about 271 horsepower, or a Dodge SRT-10 truck with 500 horsepower.
Which is quicker?
Duh. The Lancer would smoke the truck.
So the question is, what are you wanting the camera to do: do you want it to deliver high-resolution footage? Or do you want to brag about pixel count?
For me, if I was buying a quick car, I'd want the one that was quicker, regardless of technically how many horsepower it outputs.
For the camera, I want the one that puts out the best, highest-def footage, regardless of pixel count.
That's all I'm saying.
thisiswells
Retired
Sometimes "adequate" is adequate enough for a printed specification.
fiercecurry
Well-known member
My concern is whether the image resolution will be true 1280x720 or less. If its less than we are back to DV territory, with an image that is being interpolated and upresed to HD. It would be the same if i took DV footage and scaled it up to 1280x720 (provided it was cropped to 16:9)or even 1920x1080 in post in a program like FCP or After Effects. Would it not?
thisiswells
Retired
My concern is that the picture looks good.
Don't think we should bother to know what's inside.
Don't think we should bother to know what's inside.
Barry_Green
Moderator
Yes, but here's the thing -- true 1280x720 resolution could be accomplished using pixel shift and a CCD with a pixel count as low as 854x480. By that, I mean that by employing a 50% pixel offset, enough luma samples could be gathered to resolve a grid of 1280 x 720 (854x1.5 = 1281, 480x1.5=720).
Pixel shift is not up-rezzing or interpolating. It's taking advantage of the lower chroma resolution that gets recorded, in order to capture maximum resolution.
Yet if people thought their new high-def camera had a CCD of 854x480, they'd probably scream "that's standard-def! That's lower than an XL2, fer cryin' out loud! My XL2 is higher-res than that!" Even if the high-def camera's resolution chart showed that it was delivering an honest-to-goodness, genuine, 1280 x 720 display, and the XL2 maxxed out at 720x480 on its frame, you'd still find "fanboy" arguments all across the net complaining about it. Sony uses a 960-pixel-wide CCD to deliver a high-def image, and if anyone thinks that an XL2 can stay in the same room as a Z1 (when viewed on a high-def TV) they're kidding themselves.
So that's why I say -- who cares what the under-the-hood process is. Let's see what the actual footage looks like. Let's see the actual resolution once all the elements in the imaging chain are taken into account -- lens, CCDs, AtoD conversions, DSPs, compression algorithms -- everything.
'Cause in the end, all that matters is the footage.
Pixel shift is not up-rezzing or interpolating. It's taking advantage of the lower chroma resolution that gets recorded, in order to capture maximum resolution.
Yet if people thought their new high-def camera had a CCD of 854x480, they'd probably scream "that's standard-def! That's lower than an XL2, fer cryin' out loud! My XL2 is higher-res than that!" Even if the high-def camera's resolution chart showed that it was delivering an honest-to-goodness, genuine, 1280 x 720 display, and the XL2 maxxed out at 720x480 on its frame, you'd still find "fanboy" arguments all across the net complaining about it. Sony uses a 960-pixel-wide CCD to deliver a high-def image, and if anyone thinks that an XL2 can stay in the same room as a Z1 (when viewed on a high-def TV) they're kidding themselves.
So that's why I say -- who cares what the under-the-hood process is. Let's see what the actual footage looks like. Let's see the actual resolution once all the elements in the imaging chain are taken into account -- lens, CCDs, AtoD conversions, DSPs, compression algorithms -- everything.
'Cause in the end, all that matters is the footage.
But wouldn't pixel shift further reduce the chroma sampling, or are the two not related?Barry_Green said:
The Sony CCD is 960 x 1,080 and therefore pixel-shifts only horizontally. Either a 1,280 x 720 or a 960 x 1,080 CCD would be adequate. The former would offer a lower pixel density, full vertical resolution, and horizontal oversampling (assuming 960 x 720) at 720p, and full horizontal resolution (assuming 1,280 x 1,080) and mild vertical pixel shifting at 1080i/p; the latter would offer full vertical resolution and mild pixel shifting horizontally at 1080i/p. I'm guessing, as have many others, that 1,280 x 720 is the resolution. Of course, they could go with 960 x 720 or . . . ah, skip it!
Barry_Green
Moderator
Pixel shift may reduce the chroma sampling, but the point is, the codec doesn't record that chroma anyway -- so what do you lose? You gain sensitivity, lower noise, latitude, and you lose something that you'd never have anyway.
If you're recording 4:4:4, then pixel shift wouldn't work. But at 4:2:2, you should be able to pixel-shift a full 50% additional resolution (in theory) with, near as I can tell, *no* detrimental effect on the picture at all. My understanding of pixel shift is far from complete, but this is what I've pieced together so far (although the BBC says that a 30% increase in resolution is more appropriate to expect).
Your last paragraph pretty much sums up where I've been ending up as well -- they're camera engineers, we're not, and second-guessing how they do what they do could drive one crazy. For example, a 1920x1080 image could theoretically be best recorded with a 1920x1080 CCD, but that presumes that you'd have a lens that can resolve 1920x1080! And if you don't, then the 1920x1080 CCD is completely wasted and is in fact harming the video quality! So with DVCPRO-HD, the largest maximum they should even consider would be 1280x1080 -- as that's the maximum the codec will resolve, so why even entertain a larger size? If, in fact, a horizontal pixel shift can be accomplished that loses *no* luma resolution, and preserves the full 4:2:2 color resolution, then why *not* pixel-shift it to the max, which means a maybe 855x1080 CCD? And if vertical pixel shift works as well, why not take advantage of that also -- which brings us to around 855x720... yet still delivering 100% of the maximum possible resolution the codec can record. 100% native 1280x1080 for 1080i/1080p.
Yet, if they were to announce that that's how they were implementing it, can you just imagine the howling over on some of the other boards?
I do not know how the chip is going to work. I don't have any inside information on it. All I'm saying is, I want the final image to deliver native full resolution of the high-def image. As long as it does that, I don't really care how they get there, and I may not even understand how it works -- just please get there!
If you're recording 4:4:4, then pixel shift wouldn't work. But at 4:2:2, you should be able to pixel-shift a full 50% additional resolution (in theory) with, near as I can tell, *no* detrimental effect on the picture at all. My understanding of pixel shift is far from complete, but this is what I've pieced together so far (although the BBC says that a 30% increase in resolution is more appropriate to expect).
Your last paragraph pretty much sums up where I've been ending up as well -- they're camera engineers, we're not, and second-guessing how they do what they do could drive one crazy. For example, a 1920x1080 image could theoretically be best recorded with a 1920x1080 CCD, but that presumes that you'd have a lens that can resolve 1920x1080! And if you don't, then the 1920x1080 CCD is completely wasted and is in fact harming the video quality! So with DVCPRO-HD, the largest maximum they should even consider would be 1280x1080 -- as that's the maximum the codec will resolve, so why even entertain a larger size? If, in fact, a horizontal pixel shift can be accomplished that loses *no* luma resolution, and preserves the full 4:2:2 color resolution, then why *not* pixel-shift it to the max, which means a maybe 855x1080 CCD? And if vertical pixel shift works as well, why not take advantage of that also -- which brings us to around 855x720... yet still delivering 100% of the maximum possible resolution the codec can record. 100% native 1280x1080 for 1080i/1080p.
Yet, if they were to announce that that's how they were implementing it, can you just imagine the howling over on some of the other boards?
I do not know how the chip is going to work. I don't have any inside information on it. All I'm saying is, I want the final image to deliver native full resolution of the high-def image. As long as it does that, I don't really care how they get there, and I may not even understand how it works -- just please get there!
fiercecurry
Well-known member
This is what the Andromeda folks are doing with the DVX100A. They are using "pixel shift" to get resolutions as high as 1540x990 and 1280x720. But here is the disappointment, I downloaded both clips at 720x480 and the 1540x990. The 1540x990 obviously was "larger" in pixels but on close inspection it actually was "softer" than the 720x480 clip. So all that you were gaining was an image that is HD in resolution size but not in image quality. Imagine taking a still photo to Photoshop, that is 720x480, than changing the image size to 1280x720. You are not going to magically gain more detail and resolution, you will just have a larger image. Thats it.
We will have to wait and see what the footage will look like, but if the pixel shift is the technique that the HVX will be employing, than we are not going to be getting native HD. Rather an HD spec, that does not have the detail anywhere close to what the F950, F900 or Varicams have. So it not just the 2/3 CCDs and cine style lens' that this camera will be short of but the pixel count as well, resulting in a more inferior HD image compared to it's bigger brothers. I know we all want to believe this is going to be an amazing camera, and in a lot of ways it is, but what can you realistically expect for $6000?
We will have to wait and see what the footage will look like, but if the pixel shift is the technique that the HVX will be employing, than we are not going to be getting native HD. Rather an HD spec, that does not have the detail anywhere close to what the F950, F900 or Varicams have. So it not just the 2/3 CCDs and cine style lens' that this camera will be short of but the pixel count as well, resulting in a more inferior HD image compared to it's bigger brothers. I know we all want to believe this is going to be an amazing camera, and in a lot of ways it is, but what can you realistically expect for $6000?
Aaron Koolen
Well-known member
Also, if you think of a "potential" 50% additional resolution, then the true 1920x1080 (square pixel) res that is 16:9 comes out to a CCD width of 960....Hmmm. Isn't the Varicam that width?
Aaron
Aaron
braw
Well-known member
fiercecurry said:
I don't believe that's a fair assumtion. The DVX CCD's and lens are designed for SD only, the Andromeda system is cheating HD, the HVX lens and CCD are supposedly going to be designed for HD. The lens and CCD will be completely different, they would have to. So what you mentioned is not an accurate example of pixel shift. From what I have read, almost all HD cams employ something like it. Even the F900.
Is this accurate?
thisiswells
Retired
Saying that Andromeda "cheats" HD is like saying the Canon XL1 "cheats" miniDV.
The chips in the XL1 are "upressed" so to speak to achieve 720x480...
This technology has been in use for several years. No-one is being dishonest here.
The Reel-Streem product is pretty impressive. The stills on their website are incredible,
if not an enormous download, unfortunately.
I'd say check it out.
The chips in the XL1 are "upressed" so to speak to achieve 720x480...
This technology has been in use for several years. No-one is being dishonest here.
The Reel-Streem product is pretty impressive. The stills on their website are incredible,
if not an enormous download, unfortunately.
I'd say check it out.
braw
Well-known member
I'm not trying to dog on what they have achieved with an SD camera, it's friggin amazing. But they are getting HD resolution from and SD sensor and lens. My point is the HVX will have a different sensor and lens that are intended to deliver and HD image, the DVX was never designed to do that.
Tapping into the sensor and bypassing the guts of the DVX isn't a fair analogy to pixel shifting on a sensor with a lens designed to deliver HD. Up rezzing isn't exactly the same as pixel shfting from what I understand. I may very well be wrong on this one but that's my understanding.
Tapping into the sensor and bypassing the guts of the DVX isn't a fair analogy to pixel shifting on a sensor with a lens designed to deliver HD. Up rezzing isn't exactly the same as pixel shfting from what I understand. I may very well be wrong on this one but that's my understanding.