Confused: Why are we calling this a 1920x1080 sensor?

[David Ziegelheim]

The spec seems to say it is a 960x1080 sensor that interpolates the missing pixels. The 960x540 HVX200 does a shift to get approximately higher resolution, but it isn't called 1280x720.

This camera has lots of interesting features: probably the HDMI being of most interest to me. I'm just confused why we always seem to talk about it as a 1920x1080.

Thanks,

David

[Douglas Spotted Eagle]

Because that's what the sensor is providing.

It's not a resample or shift, it's combinant of existing pixels. It's one of the benefits of CMOS that CCD can't provide; individually addressable pixels that can be tapped for more than one piece of information.

You might look at the Sony site for some simple diagrams as to how this functions; it's confusing until you actually sit down with a piece of graph paper and hand-draw it, or at least that's how it was for me. The graphics make sense, but until I drew it out for myself, the reality wouldn't sink in.

[Heath McKnight]

The sensor sizes of the Sony cameras are 960x1080, but through pixel-shift, it goes to 1440x1080 and easily to 1920x1080. The HVX200, another great HD camera, has a sensor size of 960x540; I'm not 100% sure how it works since Panasonic hasn't totally revealed their tech, but it goes to 960x720 (and then to 1280x720 on HDTVs) and 1280x1080 (and then, again, to 1920x1080 on HDTVs).

Spot is the man when it comes to this stuff. Either way, it always looks good on my 1080i HDTV and my best friend's 1080p LCD HDTV.

heath

[David Ziegelheim]

The Sony site says:

The signals from the four surrounding photodiodes are used to reproduce one more signal with the Enhanced Imaging Processor™.

They talk about "1920x1080p video signals" and "960x1080 effective pixels." That they do it before image processing rather than in image processing is semantics. They only have 960x1080 cells of information. Arguably, if they did the interpolation in a shifted array they may have more information. The Sony doesn't claim a shift, rather offset rows where the interpolated pixels are surrounded on all sides.

This should still leave the resolution edge with the Canon's 1440x1080 in interlaced mode, Or leave a horizontal vs vertical tradeoff in progressive mode. To me, that seems to support the resolution images on their site, where the Canon seems to have more horizontal resolution, and the V1 more vertical resolution. To me, at least.

[Chris Hurd]

Originally Posted by David Ziegelheim

That they do it before image processing rather than in image processing is symantics.

Incorrect. It is not "semantics." How they're getting more sampling points per pixel does not matter nearly as much as how the image actually looks on an HDTV display. Instead of getting hung up on numbers, what you need to do is evaluate the actual image for yourself and base your decision on what you see. There are a variety of ways other than the Pixel Shift process to achieve sub-pixel sampling. The Sony site provides a useful explanation of what's happening.

[David Ziegelheim]

Well, I agree with the spelling of semantics. And if you are saying the quality of the electronics and image processing software is important, I'll agree to that to. However, how do call it a 1920x1080 camera? If we only cared about the size of the image in the image processor, we would have to call all these cameras 12 or 14 bit. The big news is the uncompressed, 4:2:2 upres'd output via HDMI.

The quality of the image processing may be best addressed in a shoot out with the A1 (to HDV) or maybe G1 to compare digital HD output. When both camera's are individually tweeked.

[Heath McKnight]

Originally Posted by David Ziegelheim

Well, I agree with the spelling of semantics. And if you are saying the quality of the electronics and image processing software is important, I'll agree to that to. However, how do call it a 1920x1080 camera? If we only cared about the size of the image in the image processor, we would have to call all these cameras 12 or 14 bit.

Because you end up with a 1920x1080 image size. When I output as a full-rez QuickTime movie in Final Cut Pro, the size is 1920x1080, not 1440x1080.

heath

[David Ziegelheim]

Originally Posted by Heath McKnight

Because you end up with a 1920x1080 image size. When I output as a full-rez QuickTime movie in Final Cut Pro, the size is 1920x1080, not 1440x1080.

heath

Than you can call the Z1, A1, etc. 1920x1080 cameras. What we are talking about is the resolution of the image coming from the sensors. If there are 960x1080 cells, you have 1,036,000 (x3) pieces of information. If you had a 1920x1080 sensor, you would have 2,073,600 (x3) pieces of information. Interpolating the missing resolution at the sensor before image processing may actually be a less satisfactory algorithm, being unable to use edge detection, etc.

We have three things here: what is captured (sensor, lens, etc.), how it is processed (A/D converter, image processing), and how it is accessed (HDV, HDMI, HD-SDI, analog component, DVCProHD, AVCHD). It is the combination, with ergonomics, price, etc, that determine the result.

The V1 appears to be a great camera with a significant new feature in this price class (HDMI output). However, the A1 is also a great camera, and probably gives up little or no resolution to the V1, in contrast to what an actual 1920x1080p sensor may provide.

[Heath McKnight]

Small question, which A1? Sony's 1-cmos or Canon's 3-ccd?

heath

[Seun Osewa]

All prosumer "HD" cameras except the cinealta and maybe the JVC cameras are hopelessly far from the resolutions they claim to provide. Their "high resolution" frame grabs are almost as fuzzy as upsampled 24p DV frame grabs.

[Douglas Spotted Eagle]

Originally Posted by David Ziegelheim

Well, I agree with the spelling of semantics. And if you are saying the quality of the electronics and image processing software is important, I'll agree to that to. However, how do call it a 1920x1080 camera? If we only cared about the size of the image in the image processor, we would have to call all these cameras 12 or 14 bit. The big news is the uncompressed, 4:2:2 upres'd output via HDMI.

The quality of the image processing may be best addressed in a shoot out with the A1 (to HDV) or maybe G1 to compare digital HD output. When both camera's are individually tweeked.

1. It's not semantics. Semantics would imply that they're taking information from one pixel, shifting it or transforming it to something else, which is what every other camera in the industry does at some level, right up to CineAlta. Instead, it takes multiple bits of information from the same pixel and surrounding pixels.

You can call it whatever you like but the fact remains, it's actual samples. If you're going to go by your definition, then there aren't cameras built outside of the 100k plus cams that have imager resolution equal to final output resolution, other than a couple.

2. The output from the HDMI isn't uprezz'd, unless you consider that they convert the PAR from 1.333 to 1.0 on HDMI output. The 4:2:2 is taken from the stream pre-encoder, and as far as uncompressed, well... even the very inexpensive palm-corders have 4:2:2 uncompressed HD pipes. Just as most every DV camcorder has had 4:2:2 uncompressed SD pipes. The HDMI is just another means of packaging and delivering it.

The HDMI aspect of the camera is being significantly overblown, IMO. Yes, you get an uncompressed stream from the camera. That's exciting, but old news. The "cool" factor is that you can now keep it digital. Except...I'll gladly wager that we can compare 4:2:2 uncompressed component from the cam (or any other cam) and 4:2:2 uncompressed digital from the cam, capture them, compare them, and either you won't be able to tell the difference, or you'll like the component better. Having already done this with the AJA Xena card and beta hardware, that's been our result.

For the benefit of Sean, the V1 is quite reasonably close to XDCAM HD in terms of resolution, and not that far off of the FW900 in terms of actual resolution. The compression is what makes the differences, not to mention the significant differences of glass and imager size, but that more goes to sensitivity than resolution.

All that said, at the end of the day, view the media from the camera on a monitor capable of displaying that resolution without resampling. Make your determinations from that. Numbers are effectively meaningless, and merely provide measuring points. While we need to know how many inches are in a foot, we can't measure the value of that foot scientifically; it's aesthetic, not quantifiable.

[Greg Boston]

Originally Posted by Douglas Spotted Eagle

All that said, at the end of the day, view the media from the camera on a monitor capable of displaying that resolution without resampling. Make your determinations from that. Numbers are effectively meaningless, and merely provide measuring points. While we need to know how many inches are in a foot, we can't measure the value of that foot scientifically; it's aesthetic, not quantifiable.

Agreed, forget the numbers! I had the opportunity to see the awesome capabilities of this little package in NJ this past week.

All this measurbating gets old after awhile.

-gb-

[David Ziegelheim]

Let try this one at a time.

1. Sony is not sampling a pixel and the ones around it, just the ones around. The shift takes advantage of having three sensors, so there is another value to use. Because this is only part of the phase of getting the final image, it is only part of the story. The Silicon Imaging camera, with a forum on this site, is way under $100k and has nearly all the good stuff, at least to me.

2. The V1 HDMI output is captured at 960x1080, is upres’d to 1920x1080, is down res’d to 1440x1080, then upres’d to 1920x1080. What do you call that? The component is an interpretation of the digital signal in the camera, so that may affect the subjective view of the image. However, the HDMI captures is also much less expensive, more in line with a sub-$5k camera. Personally, I’d love to see an SI-2k Cineform solution in a $5k camera (smaller sensor, fixed lens).

3. Monitors are just now becoming native 1080p, and the electronics is just (as reported by Steve Mullen) handling the de-interlacing and up-rezing adequately. In a few years this will be the norm.

4. While I haven’t seen it personally, I find it hard to believe that an uncompressed 4:2:2 image with uncompressed audio doesn’t work much better in post than a heavily inter-frame compressed 4:2:0 image with MPEG1 audio.

Heath: I was thinking about the Canon A1. I don’t believe (although I may be wrong) the Sony A1U has the optics, quality of image control, and maybe the image processing as these cameras. Additionally, it is interlaced.

David

[Tony Tremble]

Greg

A good antidote to the measurebating are clips and good descriptions of footage that has been seen.

What are the awesome capabilities of this package as you see them?

From what I have seen there is "something" happening with the Sony V1 image. Even in a wide angle and loads of depth of field there is a real sense of depth, for want of a better explanation, a 3D feel. Is this something you've seen? I don't know how or why this should be. But it's very nice.

TT

[Douglas Spotted Eagle]

Originally Posted by David Ziegelheim

2. The V1 HDMI output is captured at 960x1080, is upres’d to 1920x1080, is down res’d to 1440x1080, then upres’d to 1920x1080. What do you call that? The component is an interpretation of the digital signal in the camera, so that may affect the subjective view of the image. However, the HDMI captures is also much less expensive, more in line with a sub-$5k camera. Personally, I’d love to see an SI-2k Cineform solution in a $5k camera (smaller sensor, fixed lens).

Your viewpoint isn't accurate, as several of us have tried to point out.

But in the words of the famous Wil Rogers, "A man convinced against his will..." and all that.