"As I told you last week Sony is going to have a big announcement day on March 13th! We know that the A57 is coming for sure but what about other cameras and lenses? Trusted sources now told me that a Sony NEX-C3 successor is coming soon (NEX-F3?) and a new NEX camcorder will also come right before the NAB show in April. I have no final specs right now but this is going to be a very busy time here on SAR…so stay tuned!Oh, and I really hope Sony will announce the new NEX lenses too! We alreayd have plenty of NEX cameras but so few NEX lenses."

PS: 4K2K to compete with JVC or a finally enhanced VG30?

Couple of thoughts....

- One, although the FS100 has clearly beaten the AF100 in the $5,000 camera field, I'd guess that sales have started slow a bit as the camera's been out for nearly a year now. Meaning early adapters, rental houses, TV shows, and those unhappy with the AF100's performance have already bought into the system.

- Two, those sales will be further cut into by Nikon's D800 - and presumably Canon's Mark III - in the next half year, as the overall video quality should be close to equal, with the added plus of a full frame sensor...

All for about $1,500 cheaper then the FS100 as it stands now.

- The VG20 seems to have been a near total flop - at least in English speaking countries where not one big name shooter/website/magazine has bothered to review the camera. Would be buyers were turned off by both lack of advertised (and industry wide accepted standard) features, and by the poor customer relations generated by the months long false advertising for those missing features.

A lack of ND wheel (also a failing of the FS100) and no XLR inputs are one thing; the inability to reduce sharpening or adjust picture profiles is something else altogether. Add to that a hideous rolling shutter and the 3X cheaper NEX 5n - which has more features to boot - and it's a safe bet that the VG20 is already headed to the dust heap of video camera history.

- All of which leads me to believe that it's totally possible that the pro division will release a $3,000-ish camera that essentially will come with the features advertised for the VG20, plus maybe 1-2 XLR ports, fewer pixels, and improved rolling shutter. No SDI (the FS100 lacks this feature as well, strange as it is). It wouldn't be unprecidented: Canon's XF100 already has oodles of pro features - including 4:2:2 50Mbps internal recording - albeit in a fixed lens, 1/3" format, for $3000.

Frankly, I think both the FS and AF100 were grossly over-priced (as the subsequent rebates and ebay auction prices bare out) at $5,000. I like the AF100's form factor and features, but dislike the MST format and underwhelming sensor. I like the FS100's image (particularly in low light), but dislike it's fragile looking form factor (one that doesn't even have a proper viewfinder).

A $3K camera that sort of saddles the two cameras would, in my opinion, sell like hotcakes if the performance was there. I do not think Sony's ready to jump into the 4K game yet - and certainly not at the $3K price point.

"I do not think Sony's ready to jump into the 4K game yet - and certainly not at the $3K price point."

Good point.

Sony's 16MP chip is about the smallest they make. More likely they bump to 24MP even though not needed. Seems every "new" camera must have more pixels. :(

Sony's 16MP chip is about the smallest they make. More likely they bump to 24MP even though not needed. Seems every "new" camera must have more pixels. :(

I hate that trend. and it's a clear sign that manufacturers aren't listening too closely to the needs of customers. Besides 4k/5k, nobody is really asking for more megapixels. Everyone is asking for better image quality and better lowlight performance, which could be achieved by improving technology at a given pixel desnity, not jamming more in for relatively the same quality as the previous generation.

- The VG20 seems to have been a near total flop - at least in English speaking countries where not one big name shooter/website/magazine has bothered to review the camera. Would be buyers were turned off by both lack of advertised (and industry wide accepted standard) features, and by the poor customer relations generated by the months long false advertising for those missing features.



I just got back from Maui and took the VG20 along. I was going to a friends wedding and I took the T2i and the VG20. I have no doubt the VG20 shot most situation better than I could get out of the T2i. I shot the camera on a whale watch, and in a background wedding shoot. (I gave hired videographers the respect and leeway they needed.)

I ran into expected problems with bright back ground situation shooting the VG 20. I think if I could have had access to contrast reduction and sharpness reduction on this camera, it cold have helped in some of situations. And I was hand holding everything, so it is hardly fair to compare with a tripod mount wedding shoot. However, I still think the VG20 does a great job in about 98% of the circumstances I will shoot in, and will mostly outshoot the DSLRs I have worked with in those situations.

I will post some footage on this soon.

And there is the paradox. Those with a VG20 won't upgrade unless the VG30 offers a lot more because the VG20 offers enough.

I would expect an OLED VF/LCD and the missing adjustments. Plus, a 24MP sensor. That would give the VG30 all the features of the NEX-7.

I just got back from Maui and took the VG20 along. I was going to a friends wedding and I took the T2i and the VG20. I have no doubt the VG20 shot most situation better than I could get out of the T2i... However, I still think the VG20 does a great job in about 98% of the circumstances I will shoot in, and will mostly outshoot the DSLRs I have worked with in those situations.

I'm not saying that the VG20 can't produce some nice images - it can. I've posted some footage from around the web that's pretty dang good... But most of that good footage were static shots during brilliant daylight hours - in other words, ideal conditions for video cameras.

Unfortunately, I've seen zero good footage from complicated lighting conditions such as you might find in a narrative film. This might be from a lack of "serious" film makers embracing the camera, but part and parcel of that is the intentionally crippled nature of the camera. Seriously, NO ability to alter the contrast settings in a $1,600 camera (w/o a lens) released in the year 2012? No presets other than the awful "cinetone?" No way even to buy firmware to upgrade those missing (and advertised) features?

Beyond greedy.

Far as a possible new camera goes, I don't think Sony gives squat what VG20 buyers think one way or another. They certainly haven't cared about VG10 buyers who have been clamoring for a 24p firmware upgrade since that camera came out.

Want more proof? If this rumored new camera comes out from the pro division, I doubt they'd care in the slightest what happens to a camera from the consumer division. Given the giant losses Sony's suffered, I'd guess it's every man/division for itself. Even if they did care, I doubt the VG20 has sold in any kind of numbers requiring them to worry too much.

No, the camera they'll try to protect is the FS100 - which is why the VG20 was crippled in the first place - meaning they'll probably build in some other bizarre limitation. Even if they don't, is a non-crippled version of a VG20 worth $3,000? I'm not sure that even a non-crippled VG20 is worth $1,600, seeing how lightly used AF100s are going for $3,000 or less right now.

No, Sony will have to do more then give us what they promised the VG20 was going be in the first place if they really want to spark sales. If Sony wants me to spend my otherwise earmarked Canon Mark III money in two months, they better at least give me one feature the Canon/Nikon DSLRs won't have, like a legit XLR port, or better codec.

I wish they make a vg20 with a removable top handle with xlr. I missed have that option from my a1u and that's the only thing holding me back from getting the vg20. Picture style would be nice too.

Unfortunately, I've seen zero good footage from complicated lighting conditions such as you might find in a narrative film. This might be from a lack of "serious" film makers embracing the camera, but part and parcel of that is the intentionally crippled nature of the camera. Seriously, NO ability to alter the contrast settings in a $1,600 camera (w/o a lens) released in the year 2012? No presets other than the awful "cinetone?" No way even to buy firmware to upgrade those missing (and advertised) features?



I have to say that "serious film makers" can overcome the issues better than non-serious film makers. I know enough about the camera to know that if you take control of your lighting situation, and fill properly, with lights and/or reflectors, you can manage to get a great film out of this camera. I also believe that as it is presented, it has a decent dynamic range to work in. Where it will disappoint many is the shooters inability to extend dynamic range by manipulating contrast and saturation and that is my complaint. I don't have many others about this camera, as I compare it to my Canon 5D and my Canon T2i. I think images are generally cleaner, higher resolution and less grainy in a given low light situation than either the 5D or the T2i. And the grain that is there, seems to have a more film like feel that that blotchy grain in the Canon cameras.

I have to say that I am really enjoying the VG20, I do a lot of travel and the 25p and fantastic stills just blow away what I was getting from the VG10, yes those 'missing' settings would be nice, but I can count on one hand the number of times i actually used them on the VG10 in just over a year.

For my use with using a big chip camcorder, once the proper white balance is set, good footage straight from the camera with no need to keep on playing around with the settings gives me the freedom and time to concentrate on good composition and controlling the depth of field.

No complants here

Guys, I appreciate what you're saying - there is, despite everything, a lot to like about the camera - it's the cheapest large sensor "true" video camera on the market; it can make very nice pictures, it's small, and it is simple to use.

But it's not like the missing features would have overwhelmed anyone - the 5 advertised presets, a contrast lever and a sharpening lever - that's all they had to add (or in this case, not subtract).

And you're right as well - perhaps I should have not used the words "serious film maker." I know some Chris's work, and I respect it greatly. This new short is also what I'd call "serious":
Water Don't Stop on Vimeo

What I should have said was that there seems to have been a lack of interest from commercial and full length film producers/shooters. Of course, this might change as time goes by.

Prolly the best indoor footage I've seen so far:
Lighting Tests with the Sony NEX-VG20 - YouTube

Looks like Sony marketing does feel the need to have a bigger number for each new model:

"NEX
NEX-F3 2012 (March) -> The NEX-C3 successor
NEX-5G, 18.3 million pixels, 2012 (September) -> The NEX-5n successor
Fullframe NEX in late 2013."

So a VG30 may have 18.3MP.

That, however, is an interesting number. I'm writing a 2 part story for Broadcast Engineering about the really big guns, like the F65. The F65 just happens to have an 18.3MP (effective) chip. Not saying the VG30 will have the 8MP diagonal pattern of the F65, but Sony has used their ClearVID pattern on prosumer camcorder in the past.

Doing so could bump the H REZ up to about 1000 pixels. I THINK that might reduce aliasing if the OLPF stayed the same.

Steve, any idea when the VG30 would actually hit the streets? I'm looking to upgrade soon...

If the announcement is at the end of March, it might be by May -- my guess.

But, it all seems odd because the NEX-7 isn't even shipping. A new NEX this spring seems too soon. I would have expected late summer.

Maybe Sony will finally let a NEX have the features available in its prosumer camcorders.

Black compress, black expand, an adjustable knee, and a full set of gamma curves.

Looks like about a dozen new cameras are coming -- all using 18.2MP chips. 18 must be the new 16.

Supposedly a new APS-C sized sensor coming for a Sony SuperZoom.

Something happen March 13.

It's hard to believe that Sony would use an 18mp sensor for a pro video application. Pros hate interpolation and moire pattern. Grossly oversampled video is bad for resolution and bad for photosite size too. It's pretty much a bad thing no matter how you look at it for video.

If there is a new NEX coming at NAB this year, it's got to have the current super 35 sensor.

Although, I cant imagine what this new NEX would look like. It's impossible have a right hand zoom rocker, (no NEX or Alpha lens has motorized zoom) so I cant imagine it being a good "run and gun" camcorder.

The FS100 already fills the void so why would Sony release a new FS100 competitor?

A "pro" version of the VG20 with XLR audio and super 35 sensor?

It's odd.

18Mp gives marketing a "bigger number" which is needed to get idiots to think of upgrading. Ye, it makes no difference in image compared to 16MP.

Under cover of a "new" sensor Sony could sneak in the firmware changes we all want. Of course, that could be done for the VG20, but Sony seems to like cameras every year.

If you worry about 18MP -- then you won't like their other option -- the NEX-7's 24MP sensor. Sony sees prosumers as consumer sheep who are fooled by bigger numbers. From their point of view WE should be willing to spend the $$$ to get an FS100. Keeping a big gap is part of their way of increasing profits.

PS: In writing a 2-part story on the true high-end (RED, F65, F3 with field recorder) I've come to the conclusion that most every videographer is making an error. We focus on DOF and sensitivity.

These are real issues, but I think latitude is even more important. That forces a move to 12-bit to 16-bit linear or 10-bit log. To do this cameras must get past AVCHD et al, to pure H.264/AVC.

I'm not sure RAW is needed. And, I'm not sure 4K is really needed. RED went much further than is necessary for video shooters. But, a QuadHD RAW camera would be nice.

Steve, I agree - hard to see how paying full price for what is sure to be the last of the 8-bit gen high end cameras makes much sense, unless you have immediate need.

Just picked up a VG20 - figure it will do enough of what it needs to do (which for me is have S35 chip + interchangable lenses) - at a reasonable price. I love the image from the FS100, but lord is that a lot of money for an 8 bit cam w/o ND wheel, proper handle, or a viewfinder.

I'm willing to spend a little to hold on until the 10/12 bit cams come down the line... which may be announced as soon as April at NAB. The rest of my dough is going towards lenses and other support gear.

The Chinese are building what they call an "S8" RAW camera. The chip is wider than a Super8mm camera as it is 2.5K. Uses Adobe CinemaDNG (each frame is a TIFF) so very short recording times -- like real film. Use all those wonderful 1960's Euro C-mount lenses. The really fast ones DO give a shallow DOF.

If it is only $2000-$3000 ready to shoot, it sure would be fun to play with. It's what I think Scarlet should have been. Great for film schools.

They are also doing an S16 and an S35. The later two use the CineForm codec. The company makes astro-cameras.

See: KineRAW-S35 (http://www.kinefinity.com/index.html)

As we all have noticed, the last decade has pushed camcorder prices from $3000 to $6000-$9000. Maybe the Chinese can bring prices back to "reasonable" levels. Supposedly the S35 will be $8000. That, of course, is big question. Quality, I don't see as an issue. Support, however, might be.

Some frame grabs from the S8.

These are from a fairly tiny chip with 2.5x1.1MP -- yet to my eyes they have a real film look. So are we paying too much attention to a big chip and megapixels?

What do you think?

Steve, 1958 called -- they want their camera back!

Hey, retro is *in* -- I think it looks pretty cool!

In the pause before NAB, I decided to learn about CinemaDMG which is the RAW 12-but file format recorded to disk by the F8. First I added an Adobe Labs plugin to Premiere CS 5.5. Then downloaded a sample of ACAM dll "footage." After stumbling around I final found the magic adjustments to RAW panels. Wow. Want to design your own camera?

I'll post a sample soon where I chose a NEW White Balance to apply to the RAW data. This overrode the metadata on the camera. SEE:

Ikonoskop ACAM dll camera's CinemaDNG test - YouTube


Next I'll try the import via Resolve Lite and output a ProRes file for FCP X.

While this is all OT, it is just one more sign of what's coming beyond big chips and 4K2K.

PS: If these really sell for $2000 I think I'm a buyer. But, what did folks think of their image quality?

KineRAW-S8p (tm) lens test reel #A2 w/o links on Vimeo

KineRAW-S8p (tm) 2.5K lens test reel B1 on Vimeo

...These are real issues, but I think latitude is even more important. That forces a move to 12-bit to 16-bit linear or 10-bit log. To do this cameras must get past AVCHD et al, to pure H.264/AVC....



In the pause before NAB, I decided to learn about CinemaDNG which is the RAW 12-but file format recorded to disk by the F8. First I added an Adobe Labs plugin to Premiere CS 5.5. Then downloaded a sample of ACAM dll "footage." After stumbling around I final found the magic adjustments to RAW panels. Wow. Want to design your own camera?

I'll post a sample soon where I chose a NEW White Balance to apply to the RAW data. This overrode the metadata on the camera. SEE:


To my mind the biggest problem with CinemaDNG and most other RAW formats so far are the incredible data rates/sizes. It would be far more practical to roll out a 16 bit variant of MPEG-2 and/or MPEG-4 + metadata at 50 Mbps - 100 Mbps and it would likely do 99% of any other RAW format in a far smaller foot print.

The S8 applies no compression. But, most all other cameras that record RAW use wavelet compression to reduce file size. So one moves beyond MPEG-2 and MPEG-4.

CineForm Wavelet compression is used by the S16 and S32 and RED.

Storage size -- which used to be a big concern -- is rapidly becoming a non-issue. 2.5K 24fps is about 100MB/s which is 6GB/m. A 12 minute run -- like a 400 foot magazine -- is only 72BG. An SSD can hold 128GB, so that's almost 20 Minutes. (Speed would need to be about 200MB/s.)

I've imported CinemaDNG into Premiere for editing.

I've also imported into AE and exported as ProRes 422 HQ. Then imported into FCP X which automatically generates ProRes Proxy in the background.

I've also exported, from AE, ProRes 4444 which nicely carries RGB as 12-bit values. This also works very nicely with FCP X because the proxy is ProRes Proxy.

Since AE outputs no Alpha data, no bits are used in the file. Thus the data are 12-bit4:4:4 RGB. This is a really cool way to work because FCP6/6 can do more than 5 streams of PR$$$ in real-time.

The nice thing is that CinemaDNG is free. I believe after importing into Photoshop one can make a template of the RAW correction. Then use that in AE. So much to learn!

Products | DigitalBolex.com (http://www.digitalbolex.com/products/)

2500$ Starter price

Specs
http://www.digitalbolex.com/wp-content/uploads/2012/02/d16specsheet.pdf

Resolution 2048 x 1152 (Super 16mm mode) + 1920 x 1080 pixels (16mm mode)
Format Adobe Cinema DNG, TIFF, JPEG Image sequences
Colour depth 12 bit – 4:4:4
File size 2 to 3 MB per frame in RAW
Sensor Kodak CCD: 12.85 mm (H) x 9.64 mm (V) – Similar to Super 16mm
Pixel Size 5.5 micron (compared to the 4.3 micron size of many DSLRs)
Framerate up to 32 fps at 2K, 60fps at 720p, 90 fps at 480p
Sound Balanced, 2 channel, 16 bit, 48 kHz via XLR
Viewfinder 320×240, 2.4” diagonal, with Focus Assist
Video out 640 x 480 B&W via ⅛” video jack (HD-SDI avail in separate unit)
Ports ⅛” video, headphone, USB 3.0, Audio XLR (2), 4-PIN XLR
Data Storage Dual CF card slots, SSD (buffer drive)
Power Internal battery, 12V External via 4 pin XLR port
Body Milled steel and hard plastic
Size (body) Approximately 5”H (without pistol grip) by 4”W by 8”D
Size (grip) 5”H by 2”W by 5”D
Lens mount C-mount comes standard; Optional PL, EF, B4
Weight 5lbs
ISO Options 100, 200, 400
Also in the box pistol grip, USB 3.0 cable, internal battery, 4 pin XLR Battery, cable, video cable, transcoder/raw conversion software

The S8 applies no compression. But, most all other cameras that record RAW use wavelet compression to reduce file size. So one moves beyond MPEG-2 and MPEG-4.

CineForm Wavelet compression is used by the S16 and S32 and RED.

Storage size -- which used to be a big concern -- is rapidly becoming a non-issue. 2.5K 24fps is about 100MB/s which is 6GB/m. A 12 minute run -- like a 400 foot magazine -- is only 72BG. An SSD can hold 128GB, so that's almost 20 Minutes. (Speed would need to be about 200MB/s.)


With some frequency I record events that run 45 minutes to 2+ hours in length, continuous. Factor in multiple cameras and would you care to re-evaluate the difference between your approach & mine and the proposition of storage being a non-issue?

Keep in mind that for this price point the solution has to be one that will fly in the high-end consumer space. With all due respect, I do not believe the RAW formats we have seen so far pass that test.

If you shoot things this long there is NO film camera you can use. I doubt even an F65 will work for you. You must live with the quality "video" provides. I could call these legacy cameras because they use compression technology that's a very weak link when compared to sensors. :)

I'm finding RAW very easy to work with. Import a 12-bit RGB CinemaDNG clip into AE and adjust RAW. THIS IS FUN. Set the desired color temp. Adjust sharpness. Create your own gamma curve. For example, create the amount of black compress YOU need.

Drop clip into a 16-bit composition and export to 12-bit RGB ProRes 4444 (no alpha). Every bit from the sensor has been processed in floating-point.

Import into FCP X where a ProRes Proxy is automatically made. You edit with the Proxy which is fully RT. (In fact, ProRes 4444 is multi-stream RT in FCP 7 on a MBP.)

After your edit, export to 12-bit RGB ProRes 4444 (no alpha). Tests show every bit from import to export is identical. Effects are rendered to 12-bit ProRes 4444.

Bottom-line, other than the deBayering process performed in floating-point -- done in your camera using 14-bit integer), there is perfect fidelity from sensor to production export.

There is no way any "video" camera can even hope to provide 4:4:4 color through a 4:2:0 codec. But, it's not just quality -- try passing a 12 stop dynamic range through an 8-bit codec. And, unless your camera's port outputs 12-bit 4:4:4 (like the F3) not even an external recorder can provide RAW quality. These are definitely not consumer cameras. They only cost a bit more than a VG20.

When RAW is combined with 4K2K, what we are working with now will look like DV.

What makes progress possible is summarized by Moore's law -- who I knew when we were both at DEC a long long time ago -- storage quadruples every 18-months. So, in fact, 2-hours will be possible.

PS: Of course, you can import CinemaDNG into Premiere and edit directly.

I think there could be a "new law" pertaining to data storage... Not sure if it's "data expands to fill the storage provided" or perhaps "storage expands to accomodate the data".

Come to think of it, that's pretty much how "suff" and "space" has always worked, at least in my closet and garage...

The demand for storage of 1's and 0's is certainly growing at a fast rate - I joke about my first computer with the 30 MEGABYTE hard drive (yeah, and maybe 128K RAM)... I've got a micro SD that's so tiny you can lose it if you blink, and it's 32 GIGABYTES... seemed like a better deal than buying a 16G that I'd fill up fast even with "consumer" AVCHD video at 1080/60p.

I started with computers with 4096 words of 12-bit core. That's 4K. No disk. Only paper-tape. Big expansion, another bank of 4K words. Naturally there was no operating system.

Then the wonder of a 4K hard disk and an operating system.

My feeling is that AVCHD was speced assuming high storage costs. By the time it was fully to market, storage had become so cheap that we now want/need 36Mbps to 72Mbps.

Canon has taken advantage of lower cost storage by offering I-frame only H.264. But, it's still 8-bit 4:2:0.

So a VG30 may have 18.3MP.

That, however, is an interesting number.
And another interesting number would be 18,662,400 - whatever that works out to in megapixels.

Why? Because it's (3x1920)x(3x1080) and could make for an easy direct read to full 1080 resolution with no need to pixel skip - similar to the C300 but in blocks of 9 rather than 4. For many reasons it's not as good as the 2x2 approach (uneven nos of R,G,B photosites block to block) but for a camera at this level it could lead to a very good compromise between stills and video performance.

It assumes a 16:9 sensor, or at least to have that many photosites in the 16:9 window of a larger sensor.

Interesting approach to use 9 rather than 4. How does that work?

I'm still stuck on the idea that the F3/FS100 chip that has 3.6 mega "pixels" but which Sony will not provide the specs for -- really is has many times more photosites.

I've got to think how 9 would work into 18.6 to get to 3.6 (for 1080p) and 4K2K.

Steve

So I wasn't the only one thinking on the possibility of a "Superpixel" approach with a 3x3 matrix that could then be used to improve the accuracy of the "interpolated pixels" (you could also possibly overlay each superpixel in such a matrix, and get lots of overlapping data points).

think of it like this:

xxx
xXx
xxx

The center "pixel" is the "target", but you can draw additional information from the surrounding 8 pixels in a higher resolution chip, plus exchange info beteen adjoining "superpixels". Combined with some fancy mathematics, could yield some interesting results...

My thoughts were that with sufficient processing capability and the rigtht algorithms, you could do quite a bit with improving interframe accuracy, as you'd have some motion data that could be used to "predict" moving objects in frames.

Ideally, I guess there needs to be an "upgrade" to the AVCHD codec to allow for the extended information in both colorspace and motion. As with all things computer related, it's a function of how much processing horsepower you can throw at the "problem", at what speed, tied to how much data storage is available, and again how fast 1's ands 0's can be processed/written.

It only gets "better" as technology progresses, because some of us just can't resist pushing the envelope!

Interesting approach to use 9 rather than 4. How does that work?
I was thinking of a much simpler approach than Daves last post - use 3x3 blocks, but directly read the R,G,B values of the photosites (as the C300 does), binning and averaging all the R,G,B within the block.

So:
RGR GRG RGR
GBG BGB GBG
RGR GRG RGR

GBG BGB GBG
RGR GRG RGR
GBG BGB GBG


etc
If you read Alan Roberts report on the C300 he speculates that the C300 may be deriving it's 720p output by a similar method. In other words, derives 1080 from 2x2 basic blocks, 720 from 3x3. See http://blog.creativevideo.co.uk/wp-content/uploads/2012/01/canon_c300_alan_roberts.pdf , bottom of page 13.As he says:
Figure 7 shows the Bayer pattern with a 1280x720 grid superimposed. Each pixel group contains either 4 or 5 green photo-sites, 1 2 or 4 red, and 1 2 or 4 blue. Clearly, it is not possible to read from the sensor in a simple way to get a 1280x720 structure, and more complex processing is needed. In principle, it would be possible to sum relevant photo-sites in each pixel and normalise the values pixel by pixel, but this would produce a fixed pattern of coloured noise which would probably not be acceptable.
What "probably wouldn't be acceptable" for a broadcast standard camera and a consumer stills/video camera may be two very different things! It may give a very acceptable compreomise between stills and video performance.
I'm still stuck on the idea that the F3/FS100 chip that has 3.6 mega "pixels" but which Sony will not provide the specs for -- really is has many times more photosites.
From charts I've seen, I'm in no doubt that the F3 has a Bayer chip of about 2456x1372 (or 3.36megapixels) - have you seen this thread? http://www.dvinfo.net/forum/sony-xdcam-f3-cinealta/496135-imager-info-service-manual.html

Apart from what the service manual says, the F3 charts show all the signs of a Bayer matrix of that resolution. Diagonal lines give slight out of band magenta-green aliasing, and horizontal and vertical give cyan-yellow - exactly as would be expected. The frequency at which this occurs is where it would be expected if the dimensions were 2456x1372.

I believe it to be doing a full deBayer, and if we assume that'll give about 80% of the sensor dimensions as luminance resolution, that would give about 1965x1100 - exactly what you'd hope a 1080 camera to give. What makes you think it has a lot more 3.36MP? From the (relatively high) power requirements of the F3, I'd take that as further evidence od deBayering and downconversion.

I confess the FS100 does puzzle me though. I can't think of a mechanism by which that same chip could give the results it does in the FS100. The much lower than F3 power requirements seem to indicate it's not deBayering like the F3, and that's backed up by the apparent total lack of coloured aliasing. It would seem that the FS100 must be employing a simpler readout mechanism than the F3, but is still managing to turn in far better results than I'd expect, even if not up to F3 standard. Can the FS100 and F3 be using different chips? Seems unlikely, Sony have said it's the same chip and it seems most logical it would be for economies of scale. I'm sure I'm missing something, but can't think what.....?
I've got to think how 9 would work into 18.6 to get to 3.6 (for 1080p) and 4K2K.
Sorry, I don't follow your meaning?

My FEELING that the chip had many more pixels was Juan's comment that Sony won't allow him to give the sensor spec. Why I asked myself? My GUESS was that the the chip had enough photosite resolution to support 4K2K in a future camera and Sony didn't want to make it obvious they would have such a camera announced in the next year (NAB 2012).

I was mainly looking at the FS100 which -- as you said -- seems to be "different."

But the Service info seems to eliminate my idea of a super high rez chip. pixels = photocites -- I guess. :)

So same chip -- perhaps two radically different ways of debayering? The F3 can output RAW S-log 4:4:4 data. That means the debayer likely can't be done as part of the readout process. It would be a post readout process on a FRAME of sequential RGB data. By hving a complete frame of data both the debayer and downscale can be performed over many pixels in all directions. A powerful DSP could do real magic in turning these data into 2MP of YCrCb values.

The FS100 might not do this -- so less heat and power and lower cost. It would need only a few video-lines of memory and so the debayer interpolation would be restricted to nearest neighbor.

Since the sensor line is longer than a recorded line, YCrCb vales must interpolated to create a downscale. That's EZ. Harder is the fact lines must interpolated to downscale the number of lines. This might require 3 lines of memory.

I was mainly looking at the FS100 which -- as you said -- seems to be "different."

But the Service info seems to eliminate my idea of a super high rez chip. pixels = photocites -- I guess. :)

So same chip -- perhaps two radically different ways of debayering?
The more I think about all this, the more significant I feel it may be.

What is crucial is to think fully about the sort of alising seen from the F3. It's slight enough to not be a problem in use - but it does act as a "fingerprint" to inner workings. Just imagine a Bayer sensor illuminated by a pattern of horizontal lines such that alternate sensor rows are white, alternate black. It follows that in one case the "white" rows will be RGRGRGRG etc, the "black" rows will be GBGBGBGB etc. It further follows that the resultant will consist solely of red and green (hence yellow) with no blue. Move the pattern down by one line and the "white" photosites will all be green and blue (hence cyan) and no red.

But this situation will be a function solely of the sensor geometry - not how it is read. If only red, green photosites are illuminated then no matter what the read mechanism you'll NEVER see any blue in the output.

And cyan/yellow aliasing on hor/vert lines is EXACTLY what you see with the F3, and EXACTLY where you'd expect to see them with dimensions of about 2456x1372. It agrees with the service manual info, and with what Sony were originally saying about the sensor.

So why aren't we seeing the same on the FS100 charts? By the previous reasoning, the readout method shouldn't matter.

But what if we assume the FS100 and the F3 DON'T have the same sensor? It's what has been generally believed for the last year - but what if it's wrong? It would explain the conflicting measurements.
And it may even give a clue to why Sony initially said the F3 sensor was 3.36 megapixel and then suddenly became coy. Easier to say nothing at all. The reason for the coyness about the F3 sensor is not to do with the F3 - it's that they don't want to release details of what the FS100 sensor is, as you suspected. If the general assumption is that the F3/FS100 have the same sensor - well, that's up to the people who think it - it's not actually been said.

Is this real? Is the assumption that the F3/FS100 have the same sensor really been wrong all along? I don't know, I admit that. But if the sensor is the same, it's very difficult to reconcile some of the measurements that have been made. If it isn't, a lot of pieces of a puzzle start fitting into place.

Of course, the question then would be if not 3.36 megapixel, what is it? It's reasonable to assume "a lot more" so that the cyan/yellow alias effect will occur at a far higher frequency and not be seen in the examples we've seen. It may not even be Bayer. Can I ask what originally made you suspect a high photosite count, and what Juans comment was?

I'll add a "clue" from Sony cameras I'm familiar with - the current P&S line, which all are SUPPOSED to use the same 16.2 Mpixel sensor... for some reason I'm getting cleaner output and about 1-1 1/2 stops better low light (subjectively) out of the TX100V than ANY of the other TX, HX, or WX cameras I've had in hand to play with.

IOW, SAME sensor, likely the same processing, yet one camera performs enough differently from the others to be quite noticeable. Some of the cameras (top of each T/H/X line, the TX100 being the top of that heap) have 60p, but this doesn't account for the differences from my testing.

I'm always impressed by the results the TX100 delivers in "bad" light - lower noise, with little or no "blotchy" or blue noise (something which I really notice ever since the HC1... it seems to come and go in Sony products!), sharper overall picture (I'd chalk that up to lens differences, maybe), and just overall better performance (better even than the late model year issue TX55 I'm using because it's super compact).

My point being that there may be different "teams" within Sony who "tweak" the various hardware somewhat differently. I've noticed that in the past, they seem to "fine tune" sensors later in the product cycle, sometimes managing to squeeze a little more out of "identical" hardware, but there also may be some internal competition to make each product line "better" than the other product lines or something.

I only add the experience as a "FWIW"...

Now that I have the latest 1080p Apple TV connected to my 63" HDTV -- I've been exploring camera tests.

1) Don't kill me, but the NEX series are -- to my eyes -- painfully soft. The NEX-7 is equally soft, but has aliasing on any thin line.

2) Interestingly, the NEX do look more detailed when Zeiss primes are used. This suggests the obvious -- the sensor is better than the stock lenses.

3) After watching many clips I decided that "softness" MAY BE a matter of taste. For many, the softness is what makes them feel the NEX shoot images more like film. But then I watched several NEX vs Canon 5D II comparisons.

Canon 5D Mark II vs Sony NEX-5 (Low Light Test) - YouTube

Sony NEX-5 vs Canon 5D-MkII - YouTube


4) Certainly no one has claimed the 5D looks like video! Yet, it has far more detail. (But, yes it has the same aliasing as the NEX.) In other words, clarity doesn't signal video. So, IMO, lack of clarity should not be taken as a a sign of film.

5) Nevertheless, there is a very clear difference between the Canon 5D and the JVC PX-10 (shooting 1080p60). In every test, the JVC blows all the others away in terms of detail and zero aliasing.

It clearly doesn't look like film. It has the "looking through a window with no glass" look. It looks, perhaps, too much like video. I'm looking for something in the middle. The stills from the S8 have that look.

PS: The NEX-5n, the A-adaptor, and a set of Zeiss primes MAY be a better way to spending $2000+.

I'll add a "clue" from Sony cameras I'm familiar with - the current P&S line, which all are SUPPOSED to use the same 16.2 Mpixel sensor... for some reason I'm getting cleaner output and about 1-1 1/2 stops better low light (subjectively) out of the TX100V than ANY of the other TX, HX, or WX cameras I've had in hand to play with...
Interesting, but somewhat different to the case I was referring to. I can well believe that (subjective) low light sensitivity will vary with processing as well as fundamental sensor. It could be as simple as lower noise processing chips, or be down to noise reduction processing. (Think of the Panasonic cameras, the HPX370 and 250. They appeared at first to be astonishingly low noise with a 1/3" chip compared to such as an EX, but then the "noise ghost" issue came up. Interframe noise reduction was being used to give a very clean result - the problem being that it works well on static scenes, but falls down with movement. Hence the noise ghosts.)

In the case I referred to, such as zone plates give a look directly at characteristics of the sensor itself. You'd expect the

I'm sorry David. Rather than start another OT thread -- which this is t0 some degree -- I tossed my observations on the NEX into it.

The way it connected -- was not stated.

Bottom-line, if I had to buy a camcorder right now I would go with the JVC PX10 because of it's amazing detail. BUT, I won't buy it because -- like the NEX series -- it simply can't handle "normal" NV or India sunlight. This failure suggests that 12-bit RAW may be necessary in addition to fine detail and shallow DOF.

Of course, an F3 is beyond many of our budgets. So, I've been posting about what would be at the $3000 to $6000. I don't think Japan is going to give us that, although maybe the Canon 5D III comes close.

I'd be REALLY careful about using those video "tests" from UTooB... there are some pretty obvious methodology flaws... MP4?!?! at 1440x1080?? And the first video just plain looks out of focus in several spots, with improperly set exposure (Sonys need to be pulled back a bit on exposure... nature of the beast).

I don't have the NEX5n (though if I decide to "test" the E mount waters, it's a logical choice), but Sony designed these to shoot 1080 60p - THAT would be a proper side by side, and not reduced/degraded resolutions...

As for the handling of high brightness situations, I think you have to consider a variable ND for most any of these big chippers! Thanks for the reminder, as I've got an SLT-A65 inbound (darn vintage Minolta glass collection!). Cheapest way I found to get hands on the new 24Mpixel sensor!

The point is not one or two comps, but watching many as one would watch a BD 1080p movie -- on a big HDTV. The NEX images simply are less clear than that from other cameras. The lens does matter -- the terribly soft NEX in the church is the 16mm pancake. Go to any PHOTO review and the comments are IT'S SOFT. But Sony had to know that because it tried a revision before shipping.

Tonight I looked at Panasonic GH2 and it matched the clarity of the 5D II, but with no aliasing. Once again this is what others say, but until I could feed 1080p to an HDTV I saw no point at looking.

Yes, an ND filter is a must, but does nothing about dynamic range. Without sensor, processing, and codec dynamic range -- all one can do is bring the brights down to prevent over-exposure -- while crushing the rest of the picture. (And, looking at the aliasing, why do you want 24MP?)

Of course, if the VGs had proper CAMCORDER controls one could adjust knee to help. (Better to have no knee and record either log 10 bit or 12-bit RAW.)

PS: The GH2 looks far better when it's been hacked to record at 65Mbps and even better when running at 176Mbps. Amazing detail increases. Which may be why the JVC at 36Mbps looks so clear.

As I said earlier, I think a whole bunch of video shooters followed Sony down a carefully guided path to big chips -- for shallow DOF -- only for some to realize that many other critical shooting aspects were sacrificed to keep the FS100 safe. They've given us a great sensor and that's it.

But, attach a $1000 Atomos ProRes 422 recorder, an adaptor for all those fine Minolta lenses, run full manual, and I suspect you've got an FS100 at a far lower cost. :)

An alternative is what's looking like a great buy -- the GH2 hacked to about 65Mbps.

PS: Watching the Digital Bolex samples last night, for better or worse, they look just like what I used to shoot. This could be a very fun camera!

I own both the Canon 5D II, and the Sony VG 20. I will confess that I have had other things going on since I bought on the first day the Sony was available, so testing was limited to a few situations, but I feel I can comfortably make a judgement between the two.

I use the same lens system on both cameras, with the exception of course, of the Sony stock zoom. In testing I have done, I have no doubt that the Sony VG20 has a little bit more low light capability than the 5D, and that in the low light situations, its not near as noisy. In addition, the noise seem to be a fine grain on the VG20, as a opposed to the more blooming looking grain in the 5D.

In a normal lighting situation, the cameras come closer. However, I think the VG20 resolves better, in tests with a two sister Nikons, a 35mm for the VG20, and a 50mm for the 5DII, I showed the Canon resolving somewhere around 550 lines while the Vg20 resolved around 700. As a result, images appear to provide finer detail in foliage and background. (Of course some of that might be attributable to the shallower depth of field in the 5DII.)

I saw several 5D II tests where the shallow DOF made the background go very soft.

Your number seem right for the VG because it is somewhat less than the FS100 which is somewhat less than the F3. All these cameras are less than provided by the EX3 -- at 1000. Of course, you were using a Nikon lens.

Your number seems wrong for the 5D. It seems unlikely a camera that can capture only 550 lines could be used by filmmakers when the RED provides 1600 lines (1.78:1). 550 is HDV resolution!

My GUESS would be your Nikon lens is not as good as the lenses those shooting with the 5D normally use.

But, in any case, my point is that capturing 1000 to 1600 lines yields a picture that is significantly clearer. And, log 10 bit 4:4:4 RAW or 12 bit 4:4:4 RAW record dynamic range and color that 24Mbps AVCHD can't. I feel the market is shifting away from looking at only chip size and toward lens-quality and avoiding low bit rate codecs -- or codecs entirely.

Steve, I think it your numbers seem a bit high. Attached is a chart Zacutto did in the shootout in 2011. But I am ready to be schooled if we are talking about different things. The charts in the shoot out were shoot with lenses that were capable of resolving 4000 lines.

Attached is a chart Zacutto did in the shootout in 2011. But I am ready to be schooled if we are talking about different things.
One point about that chart is the units used: "Line PAIRS per picture height". It's more usual to talk about "lines per picture height" - lpph. Hence 540 Line pairs per picture height means the same thing as 1080 lines per picture height.

And for a 1080 sytem camera, the absolute maximum it could resolve is 1080 lpph. If you fed any more into it, and saw a result, it's aliasing. A bad thing. A 1080 camera CANNOT resolve better than 1080 lpph, and in practice 1000 lpph is more desirable and realistic.

It's also not the whole story by any means to just give a figure for the resolution and leave it at that. As example, look at Alan Roberts zone plate for the 5D - http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034-ADD39_Canon_5D_DSLR.pdf . At what point does the "real" resolution die away and the aliasing take over. If you pointed it at a chart with 1200lpph I'm sure it would give a result - but an undesirable, aliased one. It should give plain grey for an "ideal" 1080 camera.

I'm very sceptical of many of the figures in the Zacuto chart, and for another example take the AF100 figure of 509 line pairs measured resolution (=1018 lpph). That is complete nonsense. Look at Adam Wilts comparisons of the F3 and the AF100.

Look at the first chart and especially the horizontal or vertical lines in the wedge. At the outside, they are converging (as on the original chart) but around 650 lpph they start to DIVERGE. The lines on the chart don't do that in reality, of course - the result proves the AF100 aliases above about 650 lpph. (Adam quotes 680 lpph, which I'd say was a bit lenient, but we're in the same ballpark.) That's equivalent to 325 line pairs (340 with Adams figures).

You can see from the chart just how ridulous the Zacuto figure of 509 line pairs is. It equates to 1018 on Adams chart - or just inside the 1000 line ring. What should be very fine horizontal and/or vertical lines are very coarse and blurred there - nothing like the chart original. (Compare with the F3.) Even more distinct is the next wedge with lines at 22 deg to the axis - instead of the fine radial lines of the chart the aliasing has turned it in to lines at 90 deg to what they should be!! 509 line pairs my *&^?*.........!

Just to make it more fun -- in the cinema world with different aspect-ratios. So only the horizontal resolution is measured, in LINES.

So a RED measures 3200 LINES. If you divide by1.78 for a 16:9 ratio, it falls to about 1800 Lines p/height.

For a 3 chip cameras, 1000TVL/ph is the maximum.

UPDATE:

Curious about tests I found this. "In our test, the PX10 was able to record a horizontal sharpness of 900 lw/ph and a vertical sharpness of 800 lw/ph—both of which are excellent results. Looking back at the performance crops on our Color and Noise page, you can see how sharp the image captured by the PX10 really is."

When you consider this is a single chip camera, 900 is amazing! If you ignore the 1.78:1 aspect ratio, it's 1068 LINES. Now if JVCs QuadHD camera doubles this, that's about 2000 lines.

Of course, 36Mbps doesn't hurt either for shooting 1080p60.

I think I'm coming to a personal conclusion. I want SUPER clarity with ZERO image artifacts. Lo-light and MINIMUM DOF just are not my concern. And, while I see the need for a film look for narrative work, I've just never been interested.

My experience is based on some personal testing and experience.

David Heath said: "One point about that chart is the units used: "Line PAIRS per picture height". It's more usual to talk about "lines per picture height" - lpph. Hence 540 Line pairs per picture height means the same thing as 1080 lines per picture height."

I had assumed this to be just a difference in terminology over and not a difference in the actual measurement. No matter what though, the numbers in in the Zacutto testing reporting present a scale for the different cameras which is consistent with my experience.

I have test shot foliage in wide angle shots and other fine detail objects, using stills and grab frames from video with my Canon 5D. Whatever the reason, its just not that good with that fine detail in which to me echos the resolution testing. I don't really care about pixel counting and numbers, I just look at what I see. There is absolutely no doubt that my Nikon primes have the resolving power necessary to meet any video need when you compare the stills to the video grabs and so the real issue with the 5D Mark II is resolving power of the chip, and of course, any thing the camera does to create the final video file.

I have always felt that the pure DSLR video shooters were over emphasizing things in expressing their desire to shoot with the absolute highest resolution and best lenses. It seemed to me that for pure video purposes, medium quality lenses with proper chromatic aberation correction will do as well as well as the expensive L line lenses..

Steve Mullen said: "I think I'm coming to a personal conclusion. I want SUPER clarity with ZERO image artifacts. Lo-light and MINIMUM DOF just are not my concern. And, while I see the need for a film look for narrative work, I've just never been interested. "

I have to agree that is exactly what the video for film debate is all about. I feel that the push to 4k will actually result in the loss of film feel for many applications. I saw "Acts of Valor", the film shot 70% with the 5DMark II, about three years about, even before the camera had 24p capability, as I recall. As a lowly operator who has used the camera, I saw some of the issues I had experienced with this camera. It certainly was not perfect. But next to me on each side, were friends or family that have no connection with the camera or film making, and I never heard one thing about lack of focus, or poor highlight or shadow detail or bad dynamic range. They just enjoyed the film for what it was.

I had assumed this to be just a difference in terminology over and not a difference in the actual measurement.
Well, the difference between "line pairs" and "lines" is simply a factor of two. "Line pairs" assumes alternating black/white lines and counts the number of pairs, "lines" counts the number of total lines, black and white.
No matter what though, the numbers in in the Zacutto testing reporting present a scale for the different cameras which is consistent with my experience.
If that is so, it's likely to down to co-incidence! The figures on the Zacuto chart are seriously flawed, to the extent of being useless in my mind, and I've no idea how they were come by. The Zacuto figures show the F3 as having a measured resolution of 540 line pairs versus 509 for the AF100. It implies the F3 is better than the AF100 - but not by that much. Adam Wilts charts that I linked to earlier show a far different story, and a very significant difference. I'm in no doubt that somewhere around 325 line pairs (650 lines) is the accurate figure for the AF100.

But as far as the 5D2 goes then there's a zone plate at http://thebrownings.name/WHP034/images/WHP034-ADD39-5D2-zone-plate.png At a glance, the resolution looks to be about 500 lines (250 line pairs) but it's pretty obvious that's far from the whole story. It's not just that it may be a bit softer than the AF100, but it's the really bad COLOURED aliasing that is likely to give the worst differences. The AF100 aliasing is not good, but at least it's monochrome. Matters like that don't register in things like the simple Zacuto tables.
I have test shot foliage in wide angle shots and other fine detail objects, using stills and grab frames from video with my Canon 5D. Whatever the reason, its just not that good with that fine detail in which to me echos the resolution testing.
I may agree with the trends - it's the absolute values that Zacuto gives that are wrong.
I don't really care about pixel counting and numbers, I just look at what I see. There is absolutely no doubt that my Nikon primes have the resolving power necessary to meet any video need when you compare the stills to the video grabs and so the real issue with the 5D Mark II is resolving power of the chip, and of course, any thing the camera does to create the final video file.
The trouble with the "just look at what I see" approach is that your conclusions can be influenced as much by original subject matter as camera hardware. Point at one scene and camera A may look better than B - point at another scene and it may be the other way round. Use charts and it takes the guesswork away and gives a scientific basis to opinions. That said, I'd be the last to pixel count to the ultimate degree - it would be very wrong to simply say "camera A resolves 900 lines and camera B 950 - so B is obviously superior". (If B had a lot of coloured aliasing as well and A didn't that could be a far more important factor, for example.)
I have to agree that is exactly what the video for film debate is all about. .........As a lowly operator who has used the camera, I saw some of the issues I had experienced with this camera. It certainly was not perfect. But next to me on each side, were friends or family that have no connection with the camera or film making, and I never heard one thing about lack of focus, or poor highlight or shadow detail or bad dynamic range. They just enjoyed the film for what it was.
This is the "script or technical quality" chestnut again. And my opinion is that they are two different matters and it's silly to confuse them. Given the choice between a boring script shot in perfect quality and a gripping story shot on a DSLR, then I'll be the first to choose the latter. But it doesn't have to be an either/or choice.

Why can't we have the gripping story shot with the better camera? Two completely separate issues which shouldn't be confused.