IBC: Sony announces HVR-V1e

[Jack Zhang]

Once it does come to America, NASA might buy one to fly up to the space station to replace the Z1 already up there.

[Paul Rickford]

Has anyone been or going to IBC today to have a first hand look at the V1?, I would be interested in some 'hands on' and some more info as to what's in or missing from the menu (Black Stretch, full scan, timecode etc.) Shame i'm busy or I would have jumped on the plane myself.
Can we get up a section for the FX7/V1 soon as this is going to be a long thread!

Cheers

[Stu Holmes]

Originally Posted by Jack Zhang

Once it does come to America, NASA might buy one to fly up to the space station to replace the Z1 already up there.

Aha ! didn't know that - very interesting. thanks for that snippet of info.

[Boyd Ostroff]

I'm just guessing that Jack is indulging in a little wishful thinking here and not reporting an actual news story...

[Steve Mullen]

Originally Posted by Boyd Ostroff

I'm just guessing that Jack is indulging in a little wishful thinking here and not reporting an actual news story...

He certainly is! These are step-down models from the Z1/FX1 and step-up models from the HC1/HC3/A1.

I have no idea why any anyone on either of the threads is comparing the new camcorders to their FX1 or Z1. They don't really compare to any of the Canons either.

They give Sony a lower priced 3 chip HDV camcorder that has much less bulk. They fill the class Sony has always had below their top consumer and pro (PD170) models. The hallmark of these models has always been medium size, light weight, with poorer (but not poor) low-light performance.

What's new is the technology. IF the V1's DSP is, as claimed by Sony, "1920x1080p" then it's reasonable to assume that this chip is getting 3 million pixels EITHER 25 or 30 OR 50 or 60 times each second from the CMOS chips. This fills half of the 6 million cells in the DSP's memory (3x1920x1080).

This raises 2 questions:

1) Does the "p" refer to the CAPABILITY to run in progressive mode for 1920x1080/25p and 1920x1080/30p? If so, when running at 50i and 60i the DSP is running at 1920x1080/50i or 1920x1080/60i.

Or, is Sony claiming that the chips and DSP are ALWAYS running in progressive mode at 50Hz or 60Hz? I raise this question because if the "p" refers only to the CAPABILITY to operate at 25Hz and 30Hz, then I doubt we'll see 24p -- as that requires a different clock rate for everything.

However, if the chips and DSP are always running at 50p or 60p -- then 24p can be obtained easily. No clock rate change is needed -- just the addition of pull-down. 24 frames can be selected using a 2:3:2:3 cadence from 60p. Thus, every 5 frames yields 2 frames. (25p and 30p use a 2:1 cadence, where every 2 frames yield 1 frame.)

However, these 2 frames are carried in 4 fields, and there are 60 fields every second. Thus, within every 12 fields, either an upper or lower set of lines from a 24p frame is carried. This is a long-way of saying that 2:3:2:3 pulldown is used to covert 24p to 60i. This 60i looks just like film converted to video. And, reverse pull-down can be used to obtain 24p for editing.


2) Are the unfilled 1 million cells along the horizontal axis? Or, along the vertical axis? In other words, what's the chip's resolution aspect-ratio?

Everyone has assumed the CMOS are 960x1080. However, they could be 1920x540? Either is 1 million pixels. In the former case, then HOW do the missing 960-columns get filled? In the latter, how do the missing 540-rows get filled?

Some claim the 960 must be "scaled" to 1920. I doubt this. Given the diagonal pattern, it's possible the "intermediate" upper or lower (or both) CMOS elements are used to fill the missing columns. Since every VIDEO ROW has a similar pattern, the DSP may be able to obtain 1920 with no scaling. However, this type of interpolation will NOT yield the resolution that a would come from a 1920-wide CCD. In fact, effective resolution might be lower than from a 960-wide CCD or a single higher-rez CMOS. Hopefully, it will be equal.

If the DSP and chips are always running at 50Hz or 60Hz -- and the 6 million cell buffer DSP is being filled with progressive images, how are 50i and 60i obtained? In these cases, only 540-rows are output by the DSP at 50Hz or 60Hz -- as fields -- that have 1440-columns. Odd- and even-rows are output alternately.

To output 24p, 25p, or 30p -- pulldown (2:3, 2:1, 2:1) is used to output 540-rows by the DSP at 50Hz or 60Hz -- as fields -- that have 1440-columns. When these fields are viewed as frames, they will have 1080-lines with no interlace artifacts.

Thus, if I'm correct, the V1 offers technology that supports both interlace and progressive using interlace HDV as carrier. And, it will be the first Sony HDV camcorder to offer progressive with no loss of vertical resolution. This is made possible by the ability of CMOS chips to be read-out at very high-speed.

[Paulo Teixeira]

The Sony V1 is a perfect replacement for the Canon GL2 and the Canon XH-G1 is the perfect replacement for the Sony Z1. Both companies need to stop replacing each other’s camcorders and start replacing their own.
This is not funny by the way.

[Paul Rickford]

One thing I noticed, Sony are making a big thing about being shorter/ smaller than the Z1, but the battery in the pic of the V1 sticks out a long way v being recessed on the Z1.

[Wayne Morellini]

Originally Posted by Marvin Emms

clearvid and pixel shift are performance tradeoffs wrapped in snakeoil explanations. In this case a pixel shift would be genuinly useful, in that noone uses high resolution colour information, and the camera doesn't record it anyway, but clearvid desperatly needs increased resolution on the diagonals which could be achieved by placing the center of the green pixels at the middle of the cross of the other two colours. Equivalent to a movement of half a pixel in both directions on a normal system.

I saw a Sony page listing clearvid cameras, including the HC1. So, maybe they have redefined clearvid to mean the cmos technology in general now, and not specifically the tilted pixel sort. If this is the case the V1 might not be a loss and have proper regular pixels.

[Marvin Emms]

Steve,

"Everyone has assumed the CMOS are 960x1080. However, they could be 1920x540? Either is 1 million pixels."

I specifically did not assume this and pointed out that I couldn't see on the Sony site any claims to support this. I have also stated that assuming a resolution of 960x1080 is not meaningful for the clearvid system, and the answer is no, a clearvid with a resolution of 960x1080 does not have 1 million pixels, neither does one at 1920x540, not even close.

Ive worked through some simple but very dull geometry and anyone who expects me to explain it better plug up their ears to stop their brain escaping. For now I'll just tell you the results.

A clearvid sensor with square pixels at 45 degrees and a resolution of 1920x1080 will have an aspect ratio of 16:9 and a pixel count of (1920 x 1080) / 2. Yes you read that right, divided by two. This is where the 1Mpixel value comes from and this is why internally the DSP works at full 1920x1080 resolution - it is the native horizontal and vertical resolution of the sensor.

As Ive allready said, I would expect the green to be pixel shifted by half a pixel in both directions diagonally, or a 'full clearvid pixel' horizontally or vertically (actually half a diagonal pixel, which amounts to the same thing). This way it may be possible to teese back some of the lost diagonal luma resolution at the expense of reduced colour resolution. This would be a good trade off under the circumstances.


Wayne,

All the signs point to the camera having the twisted pixel structure. If the page lists HC1 as having clearvid, may I suggest this is most likley a mistake or a misinterpretation.

[Thomas Smet]

If the chips were 1920x1080 and tilted I would think SONY would want to claim 2 Mpixels. It might not look like 2 Mpixels in the final image but they really could get by trying to claim it has that many pixels. From a certain point of view they would be telling the truth and it would make the specs that much more impressive to sell. If they say 1Mpixel I would think it really means each chip only really has 1Mpixels or else knowing SONY they would exploit the heck out of the fudged number game.

I'm not if this is the case with cmos chips but in the JVC there was the issue of not being able to handle that many pixels due to heat which is why they had to do the split screen thing. Would a 3 chip cmos camera suffer from the same issues and not be able to do 1920x1080p even if they really wanted to? Since the chips are 1/4" (I think) they are even smaller than 1/3" which could make the problem even worse. I do know cmos uses less eneragy so maybe this isn't even an issue here. If it is an issue however I would think the chips would have to be either 960x1080 or 1920x540. Any resolution above that at 25p could cause the heat issue.

[Wayne Morellini]

Originally Posted by Marvin Emms

Wayne,

All the signs point to the camera having the twisted pixel structure. If the page lists HC1 as having clearvid, may I suggest this is most likely a mistake or a misinterpretation.

Or it is just a change in marketing strategy to lump all the cmos cameras behind a fancy name. I've seen this sort of post rewrite strategy in times past. Still if they are tilted pixels then that is a complete disaster, and the Sony Execs are welcome to buy the cameras all by themselves. If they were trying to be unpopular with me, bent pixels is the way to go. I'll tell you what, next they could do pixels the shape of the Startrek insinuate, I'm sure that all those pointy bits can go into most of the gaps (I know, mean, but you could get it to work, but would you want to).

[Marvin Emms]

Thomas,

You seem to misunderstand.

The chip physically has a resolution of 1920 pixels on one axis, and physically has a resolution of 1080 pixels on the other axis, but you cannot multiply these together to get the total number of pixels. That only works with an ordinary grid.

If the pixels were triangular, or hexgonal, or shaped like some sort of Escher drawing, you can nolonger multiply one axis by another to get the total number of pixels.

The total number of pixels, physical points on the silicon sensitive to light, for the 1920x1080 chip in clearvid is 1'036'800, give or take a few effective lines. Having worked through the math and found an answer that matches so closely the internal specs I have no doubt this is exactly what they are doing.

Measure the resolution of the camera with a video chart and multiply and you end up with a fake answer that is far too high. This is a perfect snakeoil solution, you get to design a 1Mpixel camera that tests like a 2Mpixel camera so long and none looks carefully at the quality.

Why do this? bigger pixels == less noise for a given size of sensor. Who cares if the result has diagonal banding issues? Sony don't.

Wayne,

I really doubt Sony are backtracking on ClearVid. For a start, so few technical sites have understood the trick. The mistake Sony made was in releasing the HC1 in the first place, it was too good.

[Heath McKnight]

I still think a lot of this points to Sony releasing the V1 (US version) as a 60i/30p/24p unit. I think they can add 24p and will do so, to compete with JVC and Canon. As far as 50i/25p, not sure if that will be on the V1, since only the Z1 (not the A1) has 60i/50i capabilities. But maybe.

heath

[Steve Mullen]

Originally Posted by Thomas Smet

If the chips were 1920x1080 and tilted I would think SONY would want to claim 2 Mpixels. If they say 1Mpixel I would think it really means each chip only really has 1Mpixels or else knowing SONY they would exploit the heck out of the fudged number game.

Sony certainly would be hurting themselves to quote a lower number given Panasonic is doing the reverse.

However, I too had considered adding a thrid option is my post in which the chip was neither 960x1080 nor 1920x540. So I'm certainly not backing any one resolution. In fact, the real point of my post was the concept of the CMOS and DSP always running at 50p or 60p! By discarding every other field -- 50i/60i is recorded. By discarding every other frame -- 25p/30p is recorded (with each remaining frame carried in 2 fields). And, by discarding frames using 2:3 pull-down -- 24p can be recorded as 60i.

My problem is the statement that "The chip physically has a resolution of 1920 pixels on one axis, and physically has a resolution of 1080 pixels on the other axis," but you cannot multiply these together to get the total number of pixels. That only works with an ordinary grid."

I'm more than happy to believe any claim if it is backed up by an explanation.

The following "The total number of pixels, physical points on the silicon sensitive to light, for the 1920x1080 chip in clearvid is 1'036'800, give or take a few effective lines. Having worked through the math and found an answer that matches so closely the internal specs I have no doubt this is exactly what they are doing." is simply a restatement of the assertion.

So take us through your math because you may very well be correct. In fact, I suspect you are because, as I told Barry, I doubt 960 is scaled by 2X to 1920 only to be down-scaled to 1440.

It also shows how the Z1 replacement may work, but using 1/3-inch CMOS. Sony will have a CMOS/DSP engine working at 1920x1080p. And if JVC buys sensors from Sony as they have in the past -- given we already know they want to go CMOS -- will they get these babies?

This CMOS/DSP engine gives Sony what they need -- a single engine for the world because Europe still wants 1080/50p. And this engine can ouput both 1080/50p and 1080/60p to Bluray.

This changes everything if we are both correct.

-------------------------
Steve Mullen
My "Sony HDV Handbook" is available at:
www.mindspring.com/~d-v-c

[Douglas Spotted Eagle]

Originally Posted by Steve Mullen

Sony certainly would be hurting themselves to quote a lower number given Panasonic is doing the reverse.

My problem is the statement that "The chip physically has a resolution of 1920 pixels on one axis, and physically has a resolution of 1080 pixels on the other axis," but you cannot multiply these together to get the total number of pixels. That only works with an ordinary grid."

I'm more than happy to believe any claim if it is backed up by an explanation.

[/url]

Given that you've already done the math, you probably can realize that due to direct mapping of pixels in CMOS, plus the added speed of CMOS, "additional" pixels can be read by combining two or more pixels together serving double duty. This is one place that CMOS simply kicks over CCD, because you can't directly address pixels with a CCD chip vs having that ability in CMOS. You are correct, 960 is not doubled to 1920 to be downconverted to 1440.
I dunno if Sony's OEM division will be selling these chips to other companies or not. I somewhat doubt it in the short run, but I've been wrong before. No one at Sony is talking about it, at least not to me.