If you're going to be at NAB, please stop by the Sony booth and say hi.
I'll be working in the F3 / FS100 area in the afternoons.

In addition, I'll be doing 2-hour presentations on the F3 & FS100 from 10:00 to Noon on Monday, Tuesday, Wednesday in a classroom off the show floor. I'll be showing quite a bit of footage shot with both cameras. If you're not sure what the differences are between the F3 and FS100, please join us. Even if the class says it is full, come anyway and we'll try to accommodate everyone.

Sony | Broadcast and Business Solutions (http://pro.sony.com/bbsccms/ext/BroadcastandBusiness/minisites/NAB2011/nab2011_seminars.html?XID=E:nab2011seminarinvite_clickhere:nab)

See you here in Vegas.

This is from the latest FS100 spec sheet:

Camera Specifications Detail:

Imaging Device Exmor Super35 CMOS Sensor

Camera Section Specifications Detail:

Minimum Illumination 0.28 lx (1/30 shutter, IRIS F1.4, Auto GAIN)
Optical System Single Super35mm size sensor
Pickup Device "Exmor" Super35 CMOS Sensor

Exmor R Image Sensor Specifications Detail:
Video Captured Image Resolution hd: 1920x1080

Not since Panasonic refused to give the resolution of their 1/3-inch HD chips has a company provided less useful information!

We have known for months that the sensor is Super35 in size. So is the chip in a $500 NEX-3. We also know it is a single sensor and is CMOS. This is not "DETAIL!"

The term "Video Captured Image Resolution" is useless.

If Sony means the camera is a 1920x1080 camera -- we don't need a "spec" to tell us that.

Does "capture" mean the number of chip pixels used for recording? If it does, then Sony has a big problem because that's only 2MP and not the 3MP needed before debayering. The pix will have low resolution and will look much softer than an EX1. And, indeed it does not measure well. A step backwards from the EX1.

Are the captured pixels contiguous -- meaning are columns and/or rows dropped from the pixels on the chip? If not contiguous -- what is the skipped/discarded ratio? When pixels are skipped/discarded chroma aliasing results. Very not good for the F3 at it's high price. And, obviously demo material will have been carefully chosen to not show aliasing.

Moreover, since we know the capture aspect ratio is 16:9 -- what is the aspect ratio of the chip? What are the Effective and Gross pixel counts?

This simple information has never been hidden before. Why now?

===============

Imaging Device: It's got a BIG chip that's "Super" and yet not as big as "Full."

Good gods, calm down. They released the spec months ago. Gross Pixels Approx. 3,530,000 pixels. Effective pixels of moving pictures in 16:9 Approx. 3,370,000 pixels. Effective pixels of moving pictures in 4:3 Approx. 2,530,000 pixels.

The pixal count is in the camera spec section, the F3 & FS100 use the same sensor.

http://www.sony.co.uk/res/attachment/file/59/1237480643259.pdf

Perhaps the confusion could be coming from the Alan Roberts BBC test.

Xxxxxxxx c cc

Yuyyyyyyyyy

The pixal count is in the camera spec section, the F3 & FS100 use the same sensor.

http://www.sony.co.uk/res/attachment/file/59/1237480643259.pdf

Perhaps the confusion could be coming from the Alan Roberts BBC test.

Allan is not very likely to be as confused as Sony marketing. After the VG10 mess there is no reason to expect Sony industrial marketing to understand the difference between a PIXEL and a PHOTOSITE.

I believe the number of pixels upon which debayering is performed is indeed 3.3MP. With every other camera that has so big a chip, the number of PHOTOSITEs is 3 to almost 8 times greater. This is partially because these chips are 4:3. And, partially because photocites cannot be too huge.

ALL factors must tie together mathematically. And the measures confirm the math. Just publishing a single number doesn't verify the number.

And this is the case with the new 65 where we get a number of numbers. I expect that Broadcast group marketing has not made an error. I think industrial marketing group has.

If there are only 3.3 million photocites, this can be confirmed by Sony publishing the photocite size, spacing, and chip aspect ratio.

And I'd suggest everyone reread Allan's paper and point out where he has gone wrong.

The Arri Alexa has 4,66 MP, which is 1.33 MP more than the quoted Sony F3 figure, rather than 3 to 8 times greater. I'm not saying it's significant, but just that not all large sensor digital cameras are 10 MP plus.

According to Arri data sheet:

The IMAGE OUTPUT is 2880 x 1620 -- which is over 4 million PIXELS.

The chip gross is 7,462,400 million PHOTOSITES.

So all I'm wanting is the same information provided by Arri and RED. Telling the FS100 buyer it's the "same chip" as used in F3 isn't providing real information. And, the Sony specs are not confirmed by measures.

If you are wondering why anyone might care -- there are two reasons.

1) If, as I suspect the chip has 14-16MPS, then it can also be used in "4K" "4KHD" and "4K2K" cameras. (These could be an FS200 or a VG20.) These cameras are a major step-up. Buying an SD camera the NAB prior to the NAB that led to your buying all new HD equipment may have been a costly error. Buying an HD camera the NAB prior to the NAB that leads to your buying all new 4K2K equipment may be a costly error. (Or, a VG20 at IBC time!)

2) If, as I suspect the chip has 14-16MPS, then -- beyond the fact this might be the same as used in the $2000 VG10 -- the critical question is HOW are these PHOTOSITES reduced to get PIXELS?

The HOW determines HOW to place the FS100 into perspective at its $6,000.

Moreover, the HOW gives us a hint at whether a Sony 4K" "4KHD" and "4K2K" camera will actually -- beyond the 4X bump in resolution -- perform like a RED. If to get a working 4K" "4KHD" and "4K2K" camera, Sony employs column and/or row skipping/discarding -- then it will have aliasing that the RED does not have. That makes the 8X zoom scarlet VERY attractive as ones next investment.

They did mention the size of each photosite didn't they? Something like 4 times bigger than average or something like that.

I can't find much of anything.

But, I did find a brief comment in TV TEK about the F65 chip.

Alex Shapirio is quoted.

"Our competitors claim to have 4K cameras, but this (F65) camera actually delivers full 4K performance."

Alex is very smart so I'm sure he is not claiming 4K cameras don't exist.

Rather he is saying different technologies can be used to build a 4K camera and the nature of the technology determines it's "performance" by which he means image quality.

This is why I'm so keen to know the numbers and technology used by the F3 and FS100.

H size = 23.6 mm

V size = 13.3 mm

Diag = 27.1 mm

These give us the chip aspect ratio = 1.77

The gross H photosite count = 8768

The gross V photosite count = 2324

The gross number of photosites = 20.38 million

The number of H pixels = 4384, recorded as 4096

The number of V pixels = 2324, recorded as 2160

The total number of pixels = 10.19 million

Total recorded pixels = 8.85 million

WHY is this called a 4K camera? Because it has 4K pixels per line. It can really be called a 4K2K camera because it has 2K pixels per column.

And, in the F65 PDF Alex gives the explanation for why there are 2X more H photosites that one would expect: an extra Green photosite for each of the 4384 pixels in a line = 4384 extra per line

And, from all this, he supports his claim that this is a better 4K camera. NO DEBAYERING is need to obtain luma! Chroma can be obtained by debayering (interpolate from nearby photosites) or by allowing chroma to have half the bandwidth of luma.

Using the same exact logic, if the Sony supplied F5 and FS100 numbers are correct, I can make the case these cameras offer inferior performance to that of the RED scarlet because they do not have the 2X, in each direction, greater number of photosites to interpolate down to FullHD. We are talking about lower than the VG10's resolution! They have only enough extra photosites to support debayering.

I'm not really a resolution-phile, I for example am totally happy with 1080p. One thing I know for sure about the F3 is the super low noise image that so far has appeared very sharp.

What I can also say about the FS100 is that NONE of the sample videos from the prototype have looked as good as the F3. That's fine I suppose because it is so much cheaper but doesn't really make me want to run to B&H and preorder a FS100. I keep waiting for that one video that shows the FS100 being great. I got that early on with the EX1, it was apparent right away that is was beautiful. Same thing with the F3.

Just saying.

Erik,

You're in Vegas, right? Why don't you come to my 10:00 - Noon workshop in room N102 (just a few yards from Sony's booth) on Wednesday morning. I'll show you some footage on a 20' screen that rivals the F3 -- plus some comparisons of different bit rates with external recorders, etc. It will be worth your time. The workshop covers the F3 and FS100. Would love to see you there.

Aww Doug, man that would be awesome! You don't know how bad I want to go. Unfortunately I am working tomorrow and it would cause quite a bit of trouble if I called off. I work for an Armored Car service and they would have to scramble to get someone who knows my route.

Sorry about that. Not as sorry as me though! Thank you very much for the offer!

After seeing this F3 footage, I can believe the chip MAY be relatively low resolution because of its very high sensitivity.

See:

Holding a Candle to the Sony PMW-F3 | CineTechnica (http://blog.abelcine.com/2011/04/07/holding-a-candle-to-the-sony-pmw-f3/)

At 800ISO with no gain, the F3 is 2-stops faster than the VG10 and with 3 stops of useable gain -- the chip compensates for the slow 18-55 and 18-200. This enables a useful ISO 6400.

Using my 120-watts of light test, the VG10 needed +24dB gain at wide. That's 4-stops of gain on top of 200ISO -- only 3200ISO. That extra stop would be very nice to have when one zoomed in and the aperture went to f/5.6. For me this answers the question is the FS100 better than the VG10. In terms of sensitivity -- yes it is. Assuming,of course, the F3 test is valid for the FS100. Which it may not be.

That leaves me with going back through the BBC report and trying to see how Allan's data can be coming from a low rez chip. Something isn't right.

I'm not really a resolution-phile, I for example am totally happy with 1080p. One thing I know for sure about the F3 is the super low noise image that so far has appeared very sharp.

What I can also say about the FS100 is that NONE of the sample videos from the prototype have looked as good as the F3. That's fine I suppose because it is so much cheaper but doesn't really make me want to run to B&H and preorder a FS100. I keep waiting for that one video that shows the FS100 being great. I got that early on with the EX1, it was apparent right away that is was beautiful. Same thing with the F3.

Just saying.

I'm not sure one needs to be a "resolution-phile" to understand the advantage of a very wide bandwidth signal. There are many reasons why one does not want to shoot and record at the same resolution as one is exporting.

Likewise, there are many reasons why one wants to record compressed RAW or Log-S.

If you want to see why Sony has built the F65 fort filmmaking -- watch this video.

fxguide | vfx, mograph, and production news (http://www.fxguide.com/)

It's about the RED Epic, but when it talks about IMAGE QUALITY it says everything Sony would say about why you really want an F65. Except, of course, that the F65 does it better!

It seems the F3 may hit a sweet spot the FS100 does not, for me.

From the presentations on the Sony's 2K/4K sensor developments

http://pro.sony.com/bbsc/video/channels-cinealta/video-prenab2011_pmwf3_35mm_camcorder_training/

PS I watched/listened twice, but couldn't figure out where the current S35 (as per the F3 and FS100) stands in this...

Are the captured pixels contiguous -- meaning are columns and/or rows dropped from the pixels on the chip? If not contiguous -- what is the skipped/discarded ratio? When pixels are skipped/discarded chroma aliasing results. Very not good for the F3 at it's high price. And, obviously demo material will have been carefully chosen to not show aliasing.

Moreover, since we know the capture aspect ratio is 16:9 -- what is the aspect ratio of the chip? What are the Effective and Gross pixel counts?

This simple information has never been hidden before. Why now?

===============

Imaging Device: It's got a BIG chip that's "Super" and yet not as big as "Full."

OK Steve - your doubts are well founded (especially in the light of BBC's Alan Roberts' report). I am an engineer, but have nothing to do with optics/electronics - so please forgive the naivety of this question:

- why "skipped/discarded"? Why not "accumulated"?

If you accumulate many tiny "photosites" into a single "pixel", you'll get the high S/N and sensitivity Sony claims...

Just a thought,

Piotr

Erik,

You're in Vegas, right? Why don't you come to my 10:00 - Noon workshop in room N102 (just a few yards from Sony's booth) on Wednesday morning. I'll show you some footage on a 20' screen that rivals the F3 -- plus some comparisons of different bit rates with external recorders, etc. It will be worth your time. The workshop covers the F3 and FS100. Would love to see you there.

Hey Doug,
Could you confirm the HDMI out specs for us. I saw the interview with you on Vimeo where you stated and they later confirmed that it's 4.4.4 but looking at the Sony website it says 4.2.2. Either way I think this will be the camera to beat this year especially when you combine it with an external recorder.

Hi Steve,

The HDMI output of the FS100 can either be 4:4:4 8-bit or 4:2:2 8-bit.
The external recorder and the camera will shake hands and determine automatically what is the highest possible setting they can share. No need to change any menus.

Also, the HDMI output has embedded timecode. Not sure how exactly that will be used right now in the real world. FYI, I could not get the TC to trigger recording on my NanoFlash the way it does with SDI. In the future I'm sure it will.

Doug

OK Steve - your doubts are well founded (especially in the light of BBC's Alan Roberts' report). I am an engineer, but have nothing to do with optics/electronics - so please forgive the naivety of this question:

- why "skipped/discarded"? Why not "accumulated"?

If you accumulate many tiny "photosites" into a single "pixel", you'll get the high S/N and sensitivity Sony claims...

Just a thought,

Piotr

OK, so from the same source you can now watch a presentation on the FS100 by Juan Martinez himself:

Sony VideON | NEXFS100U Camcorder - Sony Pre-NAB 2011 Training | Most Recent (http://pro.sony.com/bbsc/video/related-recently_added/video-prenab2011_nxcam_nexfs100u_training/)

Basically what he is saying is that:

- there is no line skipping on the S35 sensor
- the S35 sensor's pixel is 4-times larger that that on a DSLR sensor

This should answer a lot of doubts presented by Steve, and those created in our minds by the review of BBC's Alan Roberts.

Hi Steve,

The HDMI output of the FS100 can either be 4:4:4 8-bit or 4:2:2 8-bit.
The external recorder and the camera will shake hands and determine automatically what is the highest possible setting they can share. No need to change any menus.

Also, the HDMI output has embedded timecode. Not sure how exactly that will be used right now in the real world. FYI, I could not get the TC to trigger recording on my NanoFlash the way it does with SDI. In the future I'm sure it will.

Doug

Outstanding, thank you Doug!

I find it strange the choice 444 hdmi as there are no recorder for that and none that I have heard planning for it. For the little gain for 99% of shooting of 444 vs 422, a 10 bit output should have been much better and made a whole lot of differences.

(posted on wrong thread)

Hi Steve,

The HDMI output of the FS100 can either be 4:4:4 8-bit or 4:2:2 8-bit.
The external recorder and the camera will shake hands and determine automatically what is the highest possible setting they can share. No need to change any menus.

Also, the HDMI output has embedded timecode. Not sure how exactly that will be used right now in the real world. FYI, I could not get the TC to trigger recording on my NanoFlash the way it does with SDI. In the future I'm sure it will.

Doug

Hi Doug,

Does this mean the F3's HDMI can be 444 as well?

Man, Sony is really confusing. Why provide 444 in HDMI when there isn't any hardware that can take advantage of it? I just checked BlackMagic and Aja and everything is 422 YUV over HDMI except a BM Mini Converter which is an HDMI to SDI converter. HOWEVER, even though it can accept 444 RGB over HDMI, it can only output 422 YUV via SDI.

OK, so from the same source you can now watch a presentation on the FS100 by Juan Martinez himself:

Sony VideON | NEXFS100U Camcorder - Sony Pre-NAB 2011 Training | Most Recent (http://pro.sony.com/bbsc/video/related-recently_added/video-prenab2011_nxcam_nexfs100u_training/)

Basically what he is saying is that:

- there is no line skipping on the S35 sensor
- the S35 sensor's pixel is 4-times larger that that on a DSLR sensor

This should answer a lot of doubts presented by Steve, and those created in our minds by the review of BBC's Alan Roberts.

Somehow nobody care to comment on this, but nevertheless here you can find a nice summary of the FS100 presentation by Mr. Martinez:

Notes On Video: Sony NEX-FS100 Presentation - Juan Matrinez (http://notesonvideo.blogspot.com/2011/04/sony-nex-fs100.html)

"We're grossly oversampling for HD, and we're not allowed to say the specification of the sensor, but I can say that we are grossly over-sampling high definition"

Hmm....

- there is no line skipping on the S35 sensor
- the S35 sensor's pixel is 4-times larger that that on a DSLR sensor

I'm sure everyone wishes that cleared everything up. But, alas it doesn't.

Pretend you are an intern at Sony and told IF you can do the following, you get a job.

1) You can't use the 14MP sensor used in current NEX cameras and camcorders. (This let's us claim the FS100 doesn't use the same chip as the VG10.) You must use the Sony CMOS sensor used in the A55. It is a 16MP chip. It will be used in 2011 NEX cameras. When shooting a 16:9 image, the chip uses 4912 by 2760 pixels -- a total of 13.56MP.

2) You may use the current Bayer filter or you may use a different one.

3) The chip must output a pixel that is "effectively" 4X larger in order to increase performance.

3a) Sensitivity must be increased by 12dB to 18dB. We want an 800ASA sensitivity in the F3. (The VG10 has about a 150ASA to 200ASA sensitivity.)

3b) Noise must be lowered by 6dB to 9dB.

4) In order to run the chip at 60Hz rather than 30Hz, data coming from the chip must be reduced by half. This must be done without skipping or discarding.

5) You must obtain a 3.36MP before debayering. That means 4912 columns must become 2456 -- actually 2444 pixels per line. And, 2760 rows must be reduced to 1380 lines -- actually 1374 lines.

6) You must explain what that pesky BBC guy will find when he tests the F3 or FS100 for aliasing.

Before I post how I think this can be done, let's see if others can design the F3/FS100 chip.

PS: Bonus points if you can show the chip could be used for a 4K2K VG20.

Contest closes on Saturday night.

"We're grossly oversampling for HD, and we're not allowed to say the specification of the sensor, but I can say that we are grossly over-sampling high definition"

Hmm....

Hmm, is right!

Sorry we can't tell you the chip spec. -- like why? -- but we will tell you we record 2MP and the pre-debayering image is 3.36MP.

First, most single chip cameras need about 3.3MP so that after debayering one has 2MP. I've never seen these necessary pixels called "oversampling."

Second, I don't know what Juan calls "gross" but I sure don't call 68% (1.68X) gross. Now were a chip to have 6.78X more photosites -- yes, that would be gross. But, depending on how the "reduction" is made, I'm not sure I would call it over-sampling.

OK, so from the same source you can now watch a presentation on the FS100 by Juan Martinez himself:

Sony VideON | NEXFS100U Camcorder - Sony Pre-NAB 2011 Training | Most Recent (http://pro.sony.com/bbsc/video/related-recently_added/video-prenab2011_nxcam_nexfs100u_training/)

Basically what he is saying is that:

- there is no line skipping on the S35 sensor
- the S35 sensor's pixel is 4-times larger that that on a DSLR sensor

This should answer a lot of doubts presented by Steve, and those created in our minds by the review of BBC's Alan Roberts.

Amazing video, thanks for that. Makes me really want an F3 :)