HVX-200--Sensor Speculation at DVinfo.net

Barlow Elton · July 30th, 2005, 01:37 PM

I wonder what the likely resolution of the CCD will be? I am currently co-editing a feature shot in HDCAM 24P and converted to offline in DVCPRO HD 24P (23.98) through a Blackmagic Decklink HD card. Internally, FCP sees the material as 1280x1080 resolution, but it plays in QT player as a full 1920x1080 image. To my eyes, the format is astonishingly good for the bitrate, (11.4MB/s) and it easily plays off a single Firewire drive. There are some minor compression artifacts (noise in shadow areas mainly) but it's simply amazing considering it's a second pass of compression on HD material.

My curiosity is just what resoution the CCD will likely need to be to accomodate full 1080 images in the HVX-200? It seems to me that the minimum res would be just what I'm working with right now--1280x1080. That's something like 1.3 million pixels, which is somewhere in the neighborhood of half a million more pixels than full 1280x720 images, ala JVC HD100. Just enough horizontal sampling to stretch out properly to the full 1920 spec...it would seem.

I don't fully understand how 1280 squeezed horizontal pixels UNSQUEEZES to a full 1920 when played back in apps such as QT7, and looks great!

Is it likely that the CCD will be 1280x1080 in order to maximize sensitivity, or does the camera need to oversample more pixels than the format accomodates in 1080 recording, (like an F900 does--2.2 MP sensor I think) in order to increase image quality?

Just curious...

Barlow Elton · August 9th, 2005, 12:34 PM

Even though my thread was a dud, (isn't anyone interested in what the likely res of the CCD will be?) I wanted to add a prediction--Because the new HDX400 1080i camera only records 720 60p and upconverts and writes to tape as 1080i, I think that answers the question of whether the HVX200 will actually record a full, native 1080 lines. Why would Panasonic let the 200 trump the 400, or the Varicam, for that matter, in actual resolution?

Steven White · August 9th, 2005, 01:00 PM

Quote:

I don't fully understand how 1280 squeezed horizontal pixels UNSQUEEZES to a full 1920 when played back in apps such as QT7, and looks great!

It doesn't "unsqueeze". The information is gone and no miracle of math can bring it back. 1280->1920 is the simple option of digital upsampling for LCD/digital display, or in the case of a CRT, multiplying the analog signal by a constant. It doesn't look nearly as great as when the image is a native resolution, square pixel image with suitable bandwidth.

-Steve

Stephen L. Noe · August 10th, 2005, 03:24 PM

Barlow,

I'm not sure anyone knows the answer to that question at this point. We'll see as soon as the specs 'officially' come out. I'm interested in the native CCD size as well.

take care hombre...

Barlow Elton · August 10th, 2005, 05:59 PM

Quote:

Originally Posted by Stephen L. Noe

Barlow,

I'm not sure anyone knows the answer to that question at this point. We'll see as soon as the specs 'officially' come out. I'm interested in the native CCD size as well.

take care hombre...

Thanks for the response, but I wasn't looking for total answers, just mildly fun speculation. So far, no one seems to be refuting the likelihood that the HVX200 won't actually have any more true resolution than the higher end Panny HD cameras. (I really hope I'm wrong) It seems to me that a true 1080i/p acquisition (at least vertically) in such a diminutive little camera could really ruffle some feathers and upset the pecking order.

Dan Vance · August 10th, 2005, 06:51 PM

Well, just for the sake of fun speculation, we can work backwards from some assumptions and see where that leads. From what I've read, the minimum practical pixel pitch (center of one sensor site to the next) is about 4um (microns). Below that, manufacturing costs rise rapidly and light sensitivity gets unacceptably low. For a 1/3" camera with 16:9 chips, the image area is 2.94mm x 5.12mm. With 4um pitch, the maximum pixel count is then 735 x 1307 (which strongly suggests the final HVX chips will be 720 x 1280). To get a vertical count of 1080 pixels in a 2.94mm chip height, the pitch would be only 2.7um, which, if even do-able, would have awful light sensitivity. From all the whining that goes on about low-light performance of cameras, I can't imagine that they (Panasonic, or any manufacturer) would sacrifice sensitivity for that resolution.
Of course they can always claim "pixel shift" and artificially exaggerate the numbers so that buyers can believe they are getting more resolution anyhow.
But I'm betting the actual pixel count will be 720 x 1280.

Brian Sveum · August 10th, 2005, 07:14 PM

I do enjoy the numbers and speculation. Can someone tell me how much the pixel shift would bump the resoluton up to in theory. Does pixel shift allow for 1080p? I'm not very clear on this whole concept. I know Barry Green broke this down once.

Barry Green · August 10th, 2005, 08:16 PM

According to the BBC, the maximum gain in resolution one can expect from a 1/2-pixel spatial offset/pixel shift design is about the square root of 2, so 1.414 times as much res. The Sony Z1 uses spatial offset to try to get a 50% resolution boost from its CCDs, and it does come pretty darn close. The BBC said that a more likely real-world scenario would be to get a 33% increase in resolution over the native pixel count.

Barlow Elton · August 10th, 2005, 11:39 PM

Dan's droppin' some serious science, yo! Cool. I will definitely be shocked if Panasonic pulls off a *true* 1080i/p that isn't an upconvert. It seems like there are both scientific and corporate reasons why it just won't do it.

Steven White · August 11th, 2005, 06:34 AM

Quote:

To get a vertical count of 1080 pixels in a 2.94mm chip height, the pitch would be only 2.7um, which, if even do-able, would have awful light sensitivity. From all the whining that goes on about low-light performance of cameras, I can't imagine that they (Panasonic, or any manufacturer) would sacrifice sensitivity for that resolution.

Well, the Sony HDR-HC1 is equipped with a 1/3" type, 1920x1440 (4:3 square pixel) CMOS sensor, and is already on the market for significantly less than the Panasonic HVX200 will cost. It doesn't have great low-light, but the sacrafice as was made.

-Steve

Thomas Smet · August 14th, 2005, 08:44 AM

For a camera aimed at the film market why would low light need to be a huge concern. Any "Serious" film maker would usually use some kind of lights for a shoot except for very rare situations.

I personally think a possible resolution for the camera could be 1280 x720 just like the other Panasonic HD cams. Even their 1080i camera uses a 1280x720 set of chips and somehow up converts the 720 to 1080i to get DVCPro HD 1280x1080i. I e-mailed Panasonic to find out how they do this but naturally I didn't hear back from them.

What this means is that we know Panasonic knows how to get at least to 1080i from the same chips that give them 720p with very good results. What this doesn't tell us is if and what their method will be to be able to get 1080p.

Since Panasonic does pixel shift very well I am sure we will see it on the HVX200. 1280x720 times 1.5 gives exactly 1920x1080. For 1080 30p it would only need to take every other frame from the 60p chips. Then the DVCPro HD codec will end up with 1280x1080. Upscaling to 1920 and going back down to 1280 would give a slight illusion to having more detail just as with the DVX100A using pixel shift but still recording 720x840. Since the DVX100A already has enough pixels for SD video it wouldn't really need pixel shift but it does help to give a little bit more detail.

Juan's Real Stream product has shown us how good 480p chips can look when upscaled to close to 1080p because of pixel shift. Imagine how much better starting with 720p would look.

I really do hope Pansonic does go with this method because then it would mean that for uncompressed capture from the component outputs we could end up with a very nice pixel shifted 1920x1080 raster.

Dominic Jones · August 14th, 2005, 12:25 PM

Whilst I'm not doubting the HVX will have a 1280x720 sensor (seems very likely it will, although of course we'd all love a true 1080 vertical resolution), the assumption that it's not practical to have a pixel size small enough to allow 1080 on a 1/3" chip simply cannot be correct - the FX-1/Z1, after all, are 1080 ona 1/3" chip...

Thomas - are we assuming that the HVX will have pixel shift oriented CCD's in *both* axes?? I'll happily and readily admit to being *way* behind yourself and many others in the low-down-n-dirty tech of all this, but surely that would have some negative ramifications on the 720p mode of the camera, which seems to be the mode Panny are selling it on?

I'd love to get a better understanding of this, but please explain in laymans terms, or I'll get lost!!!

Graeme Nattress · August 14th, 2005, 01:41 PM

The FX1 / Z1 uses 960x1080 chips, and due to it being interlaced, will get better noise performance due to line averaging (is that where the better noise comes from?) which is needed to avoid interlace twitter, and gives you about 70% of the 1080 vertical resolution. It uses pixel shift to achieve 1440x1080, but doesn't look that high resolution. It tends to look a little better than 960, but not too much, on real footage, to my eyes.

The issue with making the photosites on the sensor small, is that you get poorer dynamic range, poorer noise levels (and noise is the enemy of compression) and you don't necessarily get higher resolution, because smaller photsites means you need a sharper lens, and that is going to get expensive, although physically smaller, which is cheaper, but I don't think that much cheaper to get you the higher resolution you need, if you know what I mean.

Basically it's all one big compromise. Panasonic stated that they're going to optimise for resolution, noise, and dynamic range, and given that the constraint is that of a 1/3" sensor, I think that's going to mean a 1280x720 chip size, with horizontal & vertical pixel shift. A 1280x720 sensor without shift is "enough" to get practicaly perfect 1080i due to it's necessary interlace factor, and with the pixel shift, it should be good enough to produce as good a 1080p as the lens is capable of, and indeed, I'd guess the lens would be the limiting factor in this, rather than the chip itself.

But it shouldn't be long now before we get to see for ourselves.

Graeme

Barlow Elton · August 14th, 2005, 02:03 PM

Quote:

Originally Posted by Thomas Smet

I personally think a possible resolution for the camera could be 1280 x720 just like the other Panasonic HD cams. Even their 1080i camera uses a 1280x720 set of chips and somehow up converts the 720 to 1080i to get DVCPro HD 1280x1080i. I e-mailed Panasonic to find out how they do this but naturally I didn't hear back from them.

What this means is that we know Panasonic knows how to get at least to 1080i from the same chips that give them 720p with very good results. What this doesn't tell us is if and what their method will be to be able to get 1080p.

Since Panasonic does pixel shift very well I am sure we will see it on the HVX200. 1280x720 times 1.5 gives exactly 1920x1080. For 1080 30p it would only need to take every other frame from the 60p chips. Then the DVCPro HD codec will end up with 1280x1080. Upscaling to 1920 and going back down to 1280 would give a slight illusion to having more detail just as with the DVX100A using pixel shift but still recording 720x840. Since the DVX100A already has enough pixels for SD video it wouldn't really need pixel shift but it does help to give a little bit more detail.

Anyone looked very carefully at the DVCPROHD demo DVD that came with FCP 4.5? The 1080i stuff is razor sharp in QT7. (You do need a hi-res monitor-1600x1200 minimum- and a fast Mac to play out smoothly) It's shockingly good if it's an upconvert. Some of the scenic mountain shots are unbelievably crisp. The Varicam 720p stuff is nice, but somewhat soft by comparison. I would be very interested to see some good uprezzing to 1080p done with that footage. Did Panasonic have a true 1080i camera when the DVD was made? That stuff seems like it was a true 1280x1080 acquisition.

If Panasonic can deliver 1080i/p acquisition with the HVX200 at 90-95% of the sharpness and quality of that demo DVD, it will absolutely smoke the competition. It might even make the obscene cost of P2 taste like cherry flavored Kool-Aid, if you know what I mean. DVCPRO HD is a very robust format--the workflow is tried and true, not kludgey like HDV. Kudos to Panasonic for taking the high road and doing entry level, professional HD their own way.

I sure hope there's a FireStore alternative right off the bat though. And I hope you can do thumbnail previews off of a HD setup just like you can with P2.

Here's hoping they don't drop the ball.

Marty Baggen · August 14th, 2005, 02:41 PM

Barlow,

as one of many whom are contemplating the move to HD (HDV or otherwise), I swear, it depends on the day of the week as to which camera I am going to buy.... either Sony, JVC, or wait for the Panny.

Can you explain what you mean by robust format for the DVC ProHD? And can you enlighten on the current post production options for it and P2 hardware?

July 30th, 2005, 01:37 PM	#1
Barlow Elton Major Player Join Date: Mar 2005 Location: UT Posts: 945	HVX-200--Sensor Speculation I wonder what the likely resolution of the CCD will be? I am currently co-editing a feature shot in HDCAM 24P and converted to offline in DVCPRO HD 24P (23.98) through a Blackmagic Decklink HD card. Internally, FCP sees the material as 1280x1080 resolution, but it plays in QT player as a full 1920x1080 image. To my eyes, the format is astonishingly good for the bitrate, (11.4MB/s) and it easily plays off a single Firewire drive. There are some minor compression artifacts (noise in shadow areas mainly) but it's simply amazing considering it's a second pass of compression on HD material. My curiosity is just what resoution the CCD will likely need to be to accomodate full 1080 images in the HVX-200? It seems to me that the minimum res would be just what I'm working with right now--1280x1080. That's something like 1.3 million pixels, which is somewhere in the neighborhood of half a million more pixels than full 1280x720 images, ala JVC HD100. Just enough horizontal sampling to stretch out properly to the full 1920 spec...it would seem. I don't fully understand how 1280 squeezed horizontal pixels UNSQUEEZES to a full 1920 when played back in apps such as QT7, and looks great! Is it likely that the CCD will be 1280x1080 in order to maximize sensitivity, or does the camera need to oversample more pixels than the format accomodates in 1080 recording, (like an F900 does--2.2 MP sensor I think) in order to increase image quality? Just curious...

August 10th, 2005, 11:39 PM	#9
Barlow Elton Major Player Join Date: Mar 2005 Location: UT Posts: 945	Dan's droppin' some serious science, yo! Cool. I will definitely be shocked if Panasonic pulls off a true 1080i/p that isn't an upconvert. It seems like there are both scientific and corporate reasons why it just won't do it. Last edited by Barlow Elton; August 11th, 2005 at 11:42 AM.

August 14th, 2005, 01:41 PM	#13
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	The FX1 / Z1 uses 960x1080 chips, and due to it being interlaced, will get better noise performance due to line averaging (is that where the better noise comes from?) which is needed to avoid interlace twitter, and gives you about 70% of the 1080 vertical resolution. It uses pixel shift to achieve 1440x1080, but doesn't look that high resolution. It tends to look a little better than 960, but not too much, on real footage, to my eyes. The issue with making the photosites on the sensor small, is that you get poorer dynamic range, poorer noise levels (and noise is the enemy of compression) and you don't necessarily get higher resolution, because smaller photsites means you need a sharper lens, and that is going to get expensive, although physically smaller, which is cheaper, but I don't think that much cheaper to get you the higher resolution you need, if you know what I mean. Basically it's all one big compromise. Panasonic stated that they're going to optimise for resolution, noise, and dynamic range, and given that the constraint is that of a 1/3" sensor, I think that's going to mean a 1280x720 chip size, with horizontal & vertical pixel shift. A 1280x720 sensor without shift is "enough" to get practicaly perfect 1080i due to it's necessary interlace factor, and with the pixel shift, it should be good enough to produce as good a 1080p as the lens is capable of, and indeed, I'd guess the lens would be the limiting factor in this, rather than the chip itself. But it shouldn't be long now before we get to see for ourselves. Graeme __________________ www.nattress.com - filters for FCP

August 14th, 2005, 02:41 PM	#15
Marty Baggen Major Player Join Date: May 2005 Location: monroe, or Posts: 572	Tell me more.... Barlow, as one of many whom are contemplating the move to HD (HDV or otherwise), I swear, it depends on the day of the week as to which camera I am going to buy.... either Sony, JVC, or wait for the Panny. Can you explain what you mean by robust format for the DVC ProHD? And can you enlighten on the current post production options for it and P2 hardware?

August 9th, 2005, 12:34 PM	#2
Barlow Elton Major Player Join Date: Mar 2005 Location: UT Posts: 945	Even though my thread was a dud, (isn't anyone interested in what the likely res of the CCD will be?) I wanted to add a prediction--Because the new HDX400 1080i camera only records 720 60p and upconverts and writes to tape as 1080i, I think that answers the question of whether the HVX200 will actually record a full, native 1080 lines. Why would Panasonic let the 200 trump the 400, or the Varicam, for that matter, in actual resolution?

August 10th, 2005, 03:24 PM	#4
Stephen L. Noe Trustee Join Date: Apr 2005 Location: Chicago Posts: 1,214	Barlow, I'm not sure anyone knows the answer to that question at this point. We'll see as soon as the specs 'officially' come out. I'm interested in the native CCD size as well. take care hombre...

August 10th, 2005, 06:51 PM	#6
Dan Vance Built the VanceCam Join Date: Apr 2004 Location: Prescott Valley, AZ Posts: 109	Well, just for the sake of fun speculation, we can work backwards from some assumptions and see where that leads. From what I've read, the minimum practical pixel pitch (center of one sensor site to the next) is about 4um (microns). Below that, manufacturing costs rise rapidly and light sensitivity gets unacceptably low. For a 1/3" camera with 16:9 chips, the image area is 2.94mm x 5.12mm. With 4um pitch, the maximum pixel count is then 735 x 1307 (which strongly suggests the final HVX chips will be 720 x 1280). To get a vertical count of 1080 pixels in a 2.94mm chip height, the pitch would be only 2.7um, which, if even do-able, would have awful light sensitivity. From all the whining that goes on about low-light performance of cameras, I can't imagine that they (Panasonic, or any manufacturer) would sacrifice sensitivity for that resolution. Of course they can always claim "pixel shift" and artificially exaggerate the numbers so that buyers can believe they are getting more resolution anyhow. But I'm betting the actual pixel count will be 720 x 1280.

August 10th, 2005, 07:14 PM	#7
Brian Sveum New Boot Join Date: Dec 2004 Location: Minneapolis, MN Posts: 6	I do enjoy the numbers and speculation. Can someone tell me how much the pixel shift would bump the resoluton up to in theory. Does pixel shift allow for 1080p? I'm not very clear on this whole concept. I know Barry Green broke this down once.

August 10th, 2005, 08:16 PM	#8
Barry Green Barry Wan Kenobi Join Date: Jul 2003 Location: North Carolina Posts: 3,863	According to the BBC, the maximum gain in resolution one can expect from a 1/2-pixel spatial offset/pixel shift design is about the square root of 2, so 1.414 times as much res. The Sony Z1 uses spatial offset to try to get a 50% resolution boost from its CCDs, and it does come pretty darn close. The BBC said that a more likely real-world scenario would be to get a 33% increase in resolution over the native pixel count.

August 14th, 2005, 08:44 AM	#11
Thomas Smet Trustee Join Date: Mar 2004 Location: Milwaukee, WI Posts: 1,719	For a camera aimed at the film market why would low light need to be a huge concern. Any "Serious" film maker would usually use some kind of lights for a shoot except for very rare situations. I personally think a possible resolution for the camera could be 1280 x720 just like the other Panasonic HD cams. Even their 1080i camera uses a 1280x720 set of chips and somehow up converts the 720 to 1080i to get DVCPro HD 1280x1080i. I e-mailed Panasonic to find out how they do this but naturally I didn't hear back from them. What this means is that we know Panasonic knows how to get at least to 1080i from the same chips that give them 720p with very good results. What this doesn't tell us is if and what their method will be to be able to get 1080p. Since Panasonic does pixel shift very well I am sure we will see it on the HVX200. 1280x720 times 1.5 gives exactly 1920x1080. For 1080 30p it would only need to take every other frame from the 60p chips. Then the DVCPro HD codec will end up with 1280x1080. Upscaling to 1920 and going back down to 1280 would give a slight illusion to having more detail just as with the DVX100A using pixel shift but still recording 720x840. Since the DVX100A already has enough pixels for SD video it wouldn't really need pixel shift but it does help to give a little bit more detail. Juan's Real Stream product has shown us how good 480p chips can look when upscaled to close to 1080p because of pixel shift. Imagine how much better starting with 720p would look. I really do hope Pansonic does go with this method because then it would mean that for uncompressed capture from the component outputs we could end up with a very nice pixel shifted 1920x1080 raster.

August 14th, 2005, 12:25 PM	#12
Dominic Jones Major Player Join Date: Mar 2005 Location: Brighton, England Posts: 225	Whilst I'm not doubting the HVX will have a 1280x720 sensor (seems very likely it will, although of course we'd all love a true 1080 vertical resolution), the assumption that it's not practical to have a pixel size small enough to allow 1080 on a 1/3" chip simply cannot be correct - the FX-1/Z1, after all, are 1080 ona 1/3" chip... Thomas - are we assuming that the HVX will have pixel shift oriented CCD's in both axes?? I'll happily and readily admit to being way behind yourself and many others in the low-down-n-dirty tech of all this, but surely that would have some negative ramifications on the 720p mode of the camera, which seems to be the mode Panny are selling it on? I'd love to get a better understanding of this, but please explain in laymans terms, or I'll get lost!!!