smpproductions
Well-known member
I had a workshop today on the XDCAM HD and they mentioned that the HVX didn't shoot in true 1080i and actually uprezzes to 1080i, is this true?
is the HVX true 1080?
- Thread starter smpproductions
- Start date
smpproductions
Well-known member
is the HVX true 1080?
I had a workshop today on the XDCAM HD and they mentioned that the HVX didn't shoot in true 1080i and actually uprezzes to 1080i, is this true?
I had a workshop today on the XDCAM HD and they mentioned that the HVX didn't shoot in true 1080i and actually uprezzes to 1080i, is this true?
Depends on what you mean.
The CCDs are not 1920x1080. But that's not the same as saying it doesn't shoot 1080.
The pixels on the CCD do not correspond 1:1 to pixels in an image. The CCD is an analog, monochrome devices meant only to capture light. The magic happens in the processing of the light.
Functionally, the HVX shoots everything as 1080 first and then downrezzes to the other resolutions.
The CCDs are not 1920x1080. But that's not the same as saying it doesn't shoot 1080.
The pixels on the CCD do not correspond 1:1 to pixels in an image. The CCD is an analog, monochrome devices meant only to capture light. The magic happens in the processing of the light.
Functionally, the HVX shoots everything as 1080 first and then downrezzes to the other resolutions.
Jeff Anderson
Custom Title!
Probably referring to the CCD size being less than 1080x1920, I forget what the actual pixel res is though. So yes, very true. Sure is pretty though 
No its not true. It shoots 1080P infact, not 1080i. EVERY mode it shoots starts in 1080p and is downressed.
The lower pixel count chips is the result of a engineering technique. essentially think of it as a "one chip" 1920x1080 CCD, broken into three chips for the sake of Chroma seperation. ( the one chip cameras have lower resolution counts for each color too. the luma signal is composited from these chroma samples. the HVX 200 IS a 1080 camera. Its just that the chip array is GENIUS. the XDcam guys were just down talking the competition.
The lower pixel count chips is the result of a engineering technique. essentially think of it as a "one chip" 1920x1080 CCD, broken into three chips for the sake of Chroma seperation. ( the one chip cameras have lower resolution counts for each color too. the luma signal is composited from these chroma samples. the HVX 200 IS a 1080 camera. Its just that the chip array is GENIUS. the XDcam guys were just down talking the competition.
Barry_Green
Moderator
No it isn't true. "uprezzing" refers to digital interpolation. That's what Sony does with the V1U. The HVX uses spatial offset to create an actual 1920x1080 sampling array off the chips.
smpproductions
Well-known member
All I want to know is, whatever the process the HVX uses is it better or worse than what the F900 or the F350 does?
smpproductions said:
It's an ambiguous question. If you're talking about effective resolution then there's no way the HVX can even come close to either the F350 or the F900 (those are more than 10x the price for a reason). Even if you somehow define a process as being "better" in theory, you still have to look at how it's being implemented and the end results is what counts.
smpproductions said:
The HVX has about the same or better apparent resolution (shoot a resolution chart and observe the image) as the other 1/3" cams which have actual 1920x1080 chips.
Mediacre said:
What I said above. A CCD is an analog, monochrome device and the pixels on the CCD do not correspond to digital pixels in the final image. What matters is the final image.
Barry_Green
Moderator
You're missing an understanding of how CCD "pixels" get translated into image pixels. There's no one-to-one correlation. A CCD isn't a digital device, it's analog. It outputs an analog signal, like a microphone does; that analog signal gets sampled into digital pixels, like an audio A-to-D does (and yet nobody ever asks if a microphone is "16 bits" or "24 bits" or whatever, right?)Mediacre said:
The resolution of each individual chip is largely irrelevant, because the system doesn't read each individual chip. It reads all three as an aggregate imaging block. And as an aggregate imaging block, the HVX system delivers about a 2K image worth of resolution according to Juan Pertierra of reel-stream.
The "pixel shift" system is not "uprezzing". Uprezzing is a digital interpolation; that's what Sony does with its V1U. "pixel shift" is not actively "shifting" anything, it merely references that one chip is spatially offset by half a pixel, which doubles the effective sampling sites (960x2 = 1920, 540x2 = 1080) to give the system a theoretical max of 1920 x 1080 sampling sites. Juan says that he gets 2K out of it because there are more active pixels on the chips; he's reading an actual 2048 x 1105 out of the system.
To take matters further, the "pixel shift" system isn't really all that different than a bayer pattern system. Look at the Silicon Imaging SI-2K chip -- that's a 1920x1080 chip system, right? So is it "true" 1920x1080? Well, it has a bayer filter over it, which means that 1/4 of its pixels are covered with blue, 1/4 of them are covered with red, and 1/2 are covered with green. And that means that its red and blue resolutions are (guess what): 960 x 540. Its green is better, at 960x1080-ish (but "ish" because it's not an actual 960x1080, it's a zigzag pattern of half-size 960x1080 pixels). You cannot read the system as discrete pixels, because no individual chip pixel works on its own -- each blue pixel must be used in concert with a nearby red and green pixel, for example.
So does that mean that the "highest resolution" you can get from an SI-2K is 960x540? Try making that claim to them and they'll laugh you out of the room. Look at the footage -- it's obviously 2K res. So how can this be? It's just the way it is.
And spatial offset works largely the same way, but instead of having all the pixels on one big chip, they're separated out onto three separate chips (one red, one green, one blue). And the total surface area of the three HVX 1/3" chips is about 75% as large as the total surface area of the SI-2K's single 2/3" chip.
You've got to divorce yourself from the mentality that "one pixel = one pixel", because it just doesn't. For example, each pixel in the target YUV frame is made up of 60% green, 29% red and about 11% blue (or thereabouts). How does that work for a 1:1 relationship? It doesn't. Because there isn't a 1:1 relationship. It's too bad that sensor designers chose the same term ("pixel") as the RGB graphic frame people did, as that just adds an unnecessary layer of confusion.
It's a different way of getting there, but the end result is quite comparable. We'll find out just how comparable once Juan gets through with the Hydra.
In the meantime, just look at the images. Listen to the testimonials. The system works, and delivers a true high-def image that intercuts nicely with other camera systems, whether VariCam or F900.
Barry_Green
Moderator
Are you concerned about the process? Or the results?smpproductions said:
The process is different. The F900 and F350 are interlaced chipsets that use dense-pixel imagers, the HVX200 uses a progressive chipset with large-pixel imagers. But do you really care about how it gets there, or do you care more about where it gets to?
Of course a $100,000 system is going to do better than a $6,000 system. But is it that much better? Is it 15x better? Or, more likely, about 15% better?
Read Chris Oben's article about how they tested the HVX200 for integration with their F900s on the TV show "The 4400". After extensive testing they declared "The P2 material was intercut with compressed HDCam all on a DVCProHD timeline. The edit looked great. With minor tweaking I was able to bring the P2 footage into line (contrast and brightness) with HDCam. The kinetic energy of the P2 footage definitely added to the intensity of the scene. Even the breaking glass footage, which was a static shot, intercut nicely. When closely scrutinized it was possible to discern the P2 footage from the F900 by looking at the detail in the highlights." They then said that the highlight issue was because of the default settings, so they took the camera into the lab and tried to match it better: "We brought the ‘P2’ back to the studio and in a controlled lighting situation, ran through the various Matrix and Gamma settings looking for the best match to our F900 look in a tungsten environment. This test was performed by comparing captured footage output from Final Cut Pro on the G5 Framestore system to our 23” HD CRT monitor. We achieved our goal using a Black ProMist 1/2 filter a Detail setting of plus 2, Gamma ‘Cine-Like D’ and Matrix set to ‘Cinelike’."
So "The 4400" is happily intercutting F900 footage and HVX200 footage for their TV series. Shane Ross talks about how they intercut the HVX200 with the VariCam in several History Channel documentaries. You can read firsthand experiences from people who have intercut the F900 with the HVX and projected it on a 30' diagonal movie theater from a 4K projector: "It intercut almost seemlessly with the F-900. For sure there was a little extra sharpness and a little better skin tones with the F-900 but not $100,000 dollars worth." (from http://www.dvxuser.com/V6/showthread.php?t=85252)
Or you can get hung up on counting pixels on a chip. In the end, which is going to do more for you: using footage? Or arguing about some pixels?
Last edited:
Thanks for the detailed reply Barry. It explains a lot in terms of how it works and the difference between uprez and pixel shifting.
The comparison to the SI2K was also useful and I would think the same comparison could be drawn with RED since it also use a Bayer sensor?
It's here where all ends for me too. The final results. I have shot with the HVX200, Z1 and HD100 and it's clear that the HD100 makes sharper images. If pixel shift is the same as having native resolution because it's not a 1 pixel = 1 pixel ratio and the HVX shoots all in 1920x1080 why is that the HD100 is sharper? I don't believe it’s the lens because that 16x fuji stinks. OK the HVX has a fixed lens but still, I don't think it can be worse than the fuji 16x, maybe equal but not worse. So what is it? That's not something that somebody told me. It's something that I have seen with my own eyes and produced the footage. I don't think it's because the HD100 has too much image enhancement either as last time I did a test I shot with the HD100 way down in sharpness and it still came out sharper. On the top of that, every single review I have read on the HVX (outside of DVXuser of course) has said things to the effect of "The HVX200 has low rez CCDs and it shows". I think if pixel shift was really that magic of a thing more expensive cameras would be using it at the same level of the HVX and using a cheaper lower rez chip and making more profit. But they are more expensive for a reason right?
Not putting the HVX down. In terms of color, skin tones etc the HD100 can’t touch it. But it is not as sharp as the HD100.
I'm still curious to see how Juan will squeeze 2k out of it . I have never been impressed by Andromeda and in my opinion, all I ever saw from it looked basically the same as normal DVX footage, just bigger, but not necessarily sharper.
The comparison to the SI2K was also useful and I would think the same comparison could be drawn with RED since it also use a Bayer sensor?
Barry_Green said:
It's here where all ends for me too. The final results. I have shot with the HVX200, Z1 and HD100 and it's clear that the HD100 makes sharper images. If pixel shift is the same as having native resolution because it's not a 1 pixel = 1 pixel ratio and the HVX shoots all in 1920x1080 why is that the HD100 is sharper? I don't believe it’s the lens because that 16x fuji stinks. OK the HVX has a fixed lens but still, I don't think it can be worse than the fuji 16x, maybe equal but not worse. So what is it? That's not something that somebody told me. It's something that I have seen with my own eyes and produced the footage. I don't think it's because the HD100 has too much image enhancement either as last time I did a test I shot with the HD100 way down in sharpness and it still came out sharper. On the top of that, every single review I have read on the HVX (outside of DVXuser of course) has said things to the effect of "The HVX200 has low rez CCDs and it shows". I think if pixel shift was really that magic of a thing more expensive cameras would be using it at the same level of the HVX and using a cheaper lower rez chip and making more profit. But they are more expensive for a reason right?
Not putting the HVX down. In terms of color, skin tones etc the HD100 can’t touch it. But it is not as sharp as the HD100.
I'm still curious to see how Juan will squeeze 2k out of it . I have never been impressed by Andromeda and in my opinion, all I ever saw from it looked basically the same as normal DVX footage, just bigger, but not necessarily sharper.
Barry_Green
Moderator
Well, comparison in that you're not going to get a 4.5k 4:4:4 image out of it, that's true. But you do get absolutely stunning full-color images that obliterate anything else on the market. So again, yes, you could claim that the Red has "only" 1920x1080 res in the red & blue, so pixel-counters could argue that it's "no more than 1080p" but once you look at the footage it's obvious that it's startlingly higher def than 1080p. Red starts with about a 4.5k image, so even though some res is lost in the de-bayer process the end result is still an extremely detailed, extraordinarily high-def picture.Mediacre said:
I've shot extensively with all of them as well, and it's clear to me that the HD100 doesn't make sharper images, they are at best very comparable.
Pixel shift isn't necessary "the same", it's just another means to get a similar result. And the HD100 isn't sharper. I mean, it is if you're comparing 720 to 720, but that's because the DVCPRO-HD recording format prefilters the 1280-pixel-wide lines down to 960 for recording. If you're comparing the live images, they are extremely comparable. If you're trying to get the sharpest images out, the HVX makes sharper images in 1080 than it does in 720; if you downrez its 1080 down to 720 you'll find that the images are every bit as sharp and perhaps sharper than the HD100.
For an equivalent price point a fixed lens will beat an interchangeable one all day long, any day of the week. The HVXs lens is substantially superior to that Fujinon.
Depends on what you mean by "way down". Did you turn it to "off"? There's a massive difference between "MIN" and "OFF", even at "MIN" it's very highly edge enhanced; the difference becomes obvious when you set it to "off"; you can see just how much edge enhancement is being applied when you toggle between "MIN" and "OFF". I mean, if by "way down" you mean you turned it down to, say, -6... be aware that at -6 the JVC is applying as much edge enhancement as the HVX does when it's at full maximum +7! The HD100 is extremely oversharpened.
The HPX500 uses it, and the FX1 and Z1 use it, and the XHA1 uses it and the XLH1 uses it... what camera are you thinking of that *doesn't* use it? Only the HD100, as far as I know. Granted the HVX does it more aggressively than the others because the HVX (and HPX500) use it vertically as well as horizontally, but basically every camera on the market is using it to increase their horizontal resolution. Even the $27,000 HPX2000 uses it to get its 1080 image. It's a far more widespread and common technique than most people seem to realize.
Last edited:
smpproductions
Well-known member
Thanks for the info Barry. I was just wondering how the 2 processes were different from how the HVX does it to how XDCAM HD does it. During my workshop on the XDCAM HD I heard alot of slamming on Varicam and HVX and how they really only shoot 720 so I just wanted to get my facts straight.
Thanks again!
Thanks again!
The Varicam shoots only 720. But the HVX shoots 1080, 720, and 480.
There are a number of distinct differences between a "pixel shift" system and a bayer pattern.Barry_Green said:
Most significantly, in a bayer pattern the red and blue matrices are such that red and blue pixels are physically in different places, and hence each can contribute uniquely towards effective (luminance) resolution. In a 3chip design they are optically in the same place (via the beam splitter) and hence each can't add extra resolution information relative to the other. So can a pixel shift system of 3x960x540 chips be then considered equivalent to a Bayer system of 2x960x540 pixels?
No, because in a bayer pattern there is no overlap between red, green and blue pixels - each pixel occupies a totally unique space. With pixel shift, the size of each pixel means each green pixel has a considerable overlap over optically adjacent red/blue pixel sites.
Those two points together mean very good for sensivity for a given sensor size, less so for resolution versus a bayer pattern as spoken of here. Pixel shift is a powerful and valid technique, and gives a worthwhile resolution improvement over what may be expected from the native chip pixel count, there's no question otherwise. But compared to a bayer pattern of the same sensor size, (and all else equal) it should give better sensitivity, but less resolution - you never get anything for nothing.
The SI-2k chip is, of course, far larger than the 1/3" of the HVX200, which brings it's sensitivity back up.
