HVX 3ccd @ 1080?

Emanuel

Cross-Examiner
Barry (Green)'s quote @ http://www.dvinfo.net/conf/showthread.php?t=54479:
«Some people had been speculating that the CCD would be a 720p CCD and "uprezzed" to 1080; this page is making it sound like it's the opposite, that it's native 1080 and "downrezzed" to 720. It also means that the 1080/60i would be created from a 1080/60p scan, so interlace and progressive will probably have the same sensitivity! And, they put in big bold letters that it has a great signal-to-noise ratio.»

So, and now my (rumor) fellows?
 
Well the Japanese translation in one of the other threads says it samples at 60p allright. But the actual size? Who knows...
And if the image delivers, who cares? It's wait 'n see.
 
I looked and looked and couldn't find anything that implied the real CCD resolution of the chips.

My opinion was that somebody jumped the gun with all that, but maybe I didn't dig deep enough.
 
Well, since the cat is partially out of the bag due to recent
Japanese translated HVX200 literature, Jan mentioned that the
HVX200 does have a 1080 60p CCD.

Here's the link: 1080 60P

Steve
 
Stevet said:
Well, since the cat is partially out of the bag due to recent
Japanese translated HVX200 literature, Jan mentioned that the
HVX200 does have a 1080 60p CCD.

Here's the link: 1080 60P

Steve

This is what Jan said a few posts later in this thread:

The signal that is captured from the CCD is a 1080P signal. The CCD is an analog device. There is not a line for line readout in any CCD camera. The CCDs on the HVX200 are not 1080 X 1920, the performance in low light would be unacceptavble. However the output is captured at 1920X1080 and then is converted to the format that is needed, like 1080i or 720P or SD. The 1080P signal only gets recorded in the 2 1080P progressive formats of 24P and 30P and these are segmented frames in a respective pulldown to 1080i. All of this happens before compression and subsampling to bring the signal to DVCPRO HD.
 
stephen birdsong's quote @ here:
«I spoke with an engineer at HD Expo, the chips are native 720p, and the 1080 mode is simply upsampling. This is the same reason why the HD XL camera is 1080i, native 1080p chips are too expensive to put into a prosumer camera. Variable framerate 720p is still pretty impressive at $6k.»

Graeme Nattress' quote:
«Pixelshift is not upsampling. It's real rez you get, and you get lower noise / better dynamic range benefits too. Indeed, you'll end up with a better, contrastier picture this way.

Other affordable HD cameras might have "resolution" but they don't have "definition", and certainly not "high definition" in the way that say, a varicam or HDCAM image just pops off the screen (and that's before post production).

Panasonic really seem to be doing the engineering right on this one, from all we've heard. But the proof is in the pudding, and that's just about baked.
»

What CCD native resolution for our HVX (1080 or 720)?

So...quoting Mike Curtis...the saga continues?
 
Native CCD resolution is irrelevant. What matters is the definition of the resolved image you see on your TV screen, which will be either full 720p, full 1080p ot 1080i depending on camera mode. Pixelshift is very valid engineering that works.

Graeme
 
"The CCD is an analog device. There is not a line for line readout in any CCD camera."

Huh? A CCD has a line for line (well, row per row) readout in its nature. The voltages are read-out analoguely, but the matrix-dividing is done beforehand, making it somewhat digital, i.e. quantisized in nature.

960x1080 like the sony makes 1080i, so 960x1080 at 60p, would make 1080p without having the full res 1920x1080 CCD's (just uprezzing in one dimension).
 
Hm. I figured from the original statements that the camera functioned as a 1080i mode machine first that it was definitely a 1080-line sensor. Knowing now that it's a 1080P sensor, the horizontal dimensions are left to pixel shift.

Given that the final output of the machine is at best 1280x1080 4:2:2, it is not worth the machine having true 1920x1080P sensors, which has indeed been shown the case. The question remains as to what the horizontal dimensions are and how much pixel-shifting there is.

The theoretical maximum improvement for a constant-position pixel shift is about 1.41 (sqrt of 2). However, a 1/2-pixel shift results in a factor of 2 sampling. The lowest resolution the sensors could be to achieve 4:2:2 at 1280x1080P is 640x1080P.

Given all the temperature problems that appeared with the JVC HD100 resulting in the split-screen problem, I wonder what Panasonic's solution will be... Going to 1280x1080P sensors might be asking for trouble. I'm already a little irritated by the 1280 horizontal dimension with the PA of 1.5 in DVCPRO-HD... that's a lot of scaling. If they use pixel-shifting to get half the horizontal resolution, I expect the images will look very soft in the horizontal dimension when played back at full resolution.

-Spiff
 
As IANAE*, I ask what's the difference between pixelshift and upsampling?

[1]
When stephen birdsong said that speaks with an engineer at HD Expo and he told that «the chips are native 720p, and the 1080 mode is simply upsampling». What does it mean?

[2]
And when you, Graeme, you said: «Pixelshift is not upsampling», it means that the method used in HVX is pixelshift and not upsampling?

[3]
What does it mean regarding that "Resolution" is not the same than "Definition"?

*I Am Not An Engineer
 
Upsampling is basically a scaling algorithm of any kind that increases the resolution. They can be smart (like the one I'm working on) or stupid, but they don't use any helper information.

Pixelshift achieves the same result, ie, higher resolution, but the pixelshift gives the algorithm extra "helper" information that means it's not guessing, but reconstructing a higher resolution, and hence, should look practically identical to that of what the higher resolution would have looked like if they'd used higher resolution chips, but with better noise and dynamic range.

for 1280x1080p, you need about 905x763 chips with horizontal and vertical pixelshift. For 1440x1080p 1018x763. For 1080i, you'd need a lot less, about 534 vertical due to the 70% interlace factor.

Given that the camera needs 1440x1080p for DvcproHD25p (PAL), I'd say that the pixel count will be something like 1000x800 at most and 960x720 at least.

Graeme
 
Thanks Graeme!

Graeme_Nattress said:
Upsampling is basically a scaling algorithm of any kind that increases the resolution. They can be smart (like the one I'm working on) or stupid, but they don't use any helper information.

Pixelshift achieves the same result, ie, higher resolution, but the pixelshift gives the algorithm extra "helper" information that means it's not guessing, but reconstructing a higher resolution, and hence, should look practically identical to that of what the higher resolution would have looked like if they'd used higher resolution chips, but with better noise and dynamic range.
So, the method used for HVX is pixelshift, right?

Graeme_Nattress said:
for 1280x1080p, you need about 905x763 chips with horizontal and vertical pixelshift. For 1440x1080p 1018x763. For 1080i, you'd need a lot less, about 534 vertical due to the 70% interlace factor.

Given that the camera needs 1440x1080p for DvcproHD25p (PAL), I'd say that the pixel count will be something like 1000x800 at most and 960x720 at least.
In terms of 24p (1280x1080) and 25p (1440x1080) comparative, as you said, because: «What matters is the definition of the resolved image you see on your TV screen, which will be either full 720p, full 1080p ot 1080i depending on camera mode. Pixelshift is very valid engineering that works.» will not to be any noticeable difference, right?

And it's the same than saying to big screen @35mm film-out or digitally projected, isn't it?
 
Doesn't matter how it's displayed, it's the end result that counts. Panasonic could use magic CCD's with 4 pixels on them for all I care - if the result looks good, it is good.

Depending on the resolution of the CCD, there could be a difference for PAL over NTSC, but it's more complex than just resolution as slightly different compression plays a part too. I'd doubt they'd look night and day different, and it could be that the lens of the camera is more of a limiting factor than the CCDs, like in the other cheap HD cameras. Who knows eh?

Graeme
 
for 1280x1080p, you need about 905x763 chips with horizontal and vertical pixelshift
Graeme, I don't think that pixel shift sampling provides a significant increase in chroma resolution. In order to supply the 4:2:2 the HVX200 promises, the resolution-increasing pixel shifting operation can only be in the horizontal dimension, and indeed requires a 1080-line sensor.

It sounds like the chip won't have vertical pixel shift (other than to do minor image enhancement). The 1080 mode will be sampled in 1080 rows, and the 720P mode will be digitally interpolated from that, given that CCD technology requires each row to sampled independently.

So, the method used for HVX is pixelshift, right?
The pixel-shift operation will be similar to the shifting on the Sony FX1/Z1U. If they used 960x1080P CCDs, it would be practically ideal for getting the 1440x1080 required of PAL land, the 1280x1080 required in NTSC land, and the 720P modes. The only question is if the internals of the camera can handle the bandwidth.

We know that the combination of pixel shift and the lens on the FX1/Z1U is not really capable of resolving the full 1440 horizontal pixels alotted to HDV. It will be interesting to see how well the lens on the HVX200 does.

-Spiff
 
You're right that you'll not quite get 4:2:2 in 1080p, but I doubt you'd notice this in the final product because of the strong DVCproHD compression - it's not like DigiBeta where the chroma channels are practically perfect. Obviously, there's enough rez to get full 4:2:2 chroma in 720p and 1080i, and really that's more than good enough.

Still, you could be right and it's a 1080 line sensor, but I doubt it.

Graeme
 
Graeme_Nattress said:
Doesn't matter how it's displayed, it's the end result that counts. Panasonic could use magic CCD's with 4 pixels on them for all I care - if the result looks good, it is good.

Depending on the resolution of the CCD, there could be a difference for PAL over NTSC, but it's more complex than just resolution as slightly different compression plays a part too. I'd doubt they'd look night and day different, and it could be that the lens of the camera is more of a limiting factor than the CCDs, like in the other cheap HD cameras. Who knows eh?

Graeme
Thanks Graeme! As I've been saying, including to other members, your POV is gold here as Barry one (unfortunately, we don't have the Nattress plug-ins or even your Standards Conversion @PC platform). I can confess that I had already thought transfer me for FCP side because (unique) products like yours. If it was not a 8 bits system and not so universal and affordable like PC...

But your bet is in a little bit difference between PAL over NTSC, isn't it?
I hope so for the price and lack of 24p.

Graeme_Nattress said:
because of the strong DVCproHD compression
I'm asking myself because I know a lot of HDCAM examples @big screen but not (yet) even one from DVCPRO-HD, as I said today in another thread (despite Barry's enlightenment, useful as always).
 
There are not really any big-screen DVCPRO-HD movies because up until the HVX, there hasn't been a 1080p camera in DVCPRO-HD. 720p's great, but it isn't 1080p. So for 1080p the only option has been the HDCAM CineAlta.

When the VariCam II arrives, with its rumored 1080p capability, we should see a lot more bigscreen movies shot in DVCPRO-HD.

When stephen birdsong said that speaks with an engineer at HD Expo and he told that «the chips are native 720p, and the 1080 mode is simply upsampling». What does it mean?
It means that the engineer he talked to was wrong, or he heard it wrong. Everything Panasonic's said is that the image starts at 1080/60p and other versions get downconverted from there.
 
Back
Top