Upcoming hands-on comparison of F3, FS100 and AF100 from Philip Bloom

[Mark David Williams]

Okay Alister so you don't know what the camera processor does only that someone led you believe it was 12bit.

QUOTE
There seems to be an assumption being made that the DSP and the output bit depths are the same
---------------------------------------------
No Not at all. The assumption is that WHERE information is provided the Sony FS100 is said to have 8 Bit signal processing and you said it has 12 bit. Now you say you were led to believe this and not the statement of fact you originally made.

[Brian Drysdale]

Originally Posted by Mark David Williams

Would be nice to walk into a store take the camera outside and test maybe spend a few hours hook the camera up to a monitor film in low light test the highlights do some colour correction try some gamma curves Hook up to a recorder etc The likliehood though is it will be plonked on the counter shown where the buttons are and sweet talked by a salesman who shows you the glossy brochure and tell explains why it is better than the competition. .

I'd go to one of the professional dealers, they usually have demo cameras and you can get good price deals from a number of them. Unless the dealer knows you, they mightn't let you take it away, but you can arrange to shoot some tests on their premises and take them away

Not all cameras are as simple to pick as the EX series, they've been around for a numbers of years and still regarded as a good bank for buck camera.

The only other place I recall seeing 8 bit mentioned is Philip Blooms blog, but it was rather loosely used rather than referring to the sensor DSP as such. I'd be surprised if that particular section is 8 bit.

[Mark David Williams]

Personally I don't know what the DSP is but it looks as if it could be 8 bit Phillip Bloom probably got his info from Kanta Yamamoto from Sony Europe. However I could be wrong Hopefully we will find out.

Sony NEX-FS100
The FS100 shares its sensor with the F3, however the similarity ends there because its signal processing is 8-bit

Sony FS100?just my first impressions | Bengske.com - Photography: How TO Tutorials, Tips, Tricks and Techniques.
t has the same Super 35mm sized sensor as the F3 but different processing. It has no SDI out, just HDMI and it is only 8 bit 422 even though HDMI can do 10 bit 422 because the processor is only 8 bit. Compromises have had to be made to slash the price down from the F3.

[R Geoff Baker]

Well if the processing is 8 bit then the output is necessarily less than that -- which would show up very quickly as banding in areas of solid colour/varying brightness. Blue sky, for instance, would rapidly pick up banding if the output was only 6 bit. There is no way around it.

So I think someone misspoke. Just my 2 cents.

Cheers,
GB

[Brian Drysdale]

In that case, I would take the 8bit to be the back end processing rather than front end at the sensor. They usually would define it as being the sensor (eg 12 bit A/D conversion) rather than just in general terms. You can have different rates, eg the F900 was 10bit and then processed at 8bit

[Mike Marriage]

Originally Posted by Mark David Williams

Personally I don't know what the DSP is but it looks as if it could be 8 bit Phillip Bloom probably got his info from Kanta Yamamoto from Sony Europe. However I could be wrong Hopefully we will find out.

Mark, even if Sony wrote it in the manual, it would still be VERY unlikely that the signal processing was 8 bit. Unless it is some new process being used, I would presume the DSP is at least 12 bit.

The chances of it being a miscommunication are FAR higher.

[Alister Chapman]

My source is a Sony engineer, I don't wish to say any more as that engineer may have revealed more than he should and I am under NDA. With the greatest of respect to Phil Bloom, he is not an engineer and technical stuff is not his forte. In addition knowing Kanta and his less than perfect english, it is very possible that there could have been a miss-understanding.

[Mark David Williams]

Exactly some new process could be the answer and partly the reason why power consumption is so low.

Time will tell.

[John Jay]

I am curious to what everyone thinks gives the best film look, seeing that these camers have now been compared.

To my eyes the 5d2 looks the most filmic with the AF100 looking the most video-ish, or as some have coined it *the reality look"

To place things into perspective I think the Panavision Genesis also looks video-ish.

[Brian Drysdale]

Early days yet, I suspect the widest range of options are with the F3 shooting S-log and you'll get a closer film look possibilities with that once you get into post.

Most film looking camera, the Arri Alexa. Unfortunately, Kodak have rather bland stocks these days, which are tending to look more and more like HD.

[Steve Mullen]

Originally Posted by Alister Chapman

My source is a Sony engineer, I don't wish to say any more as that engineer may have revealed more than he should and I am under NDA.

Alister, I think you are spot on about the F3 and FS100. Why?

I have a decade old math model that I use to predict measured resolution based upon published sensor specs. So I updated it for the RED ONE. It predicts 3193 LW/ph and the consensus is 3200LW/ph.

So I input your F3 and FS100 measures plus Sony's 3.37MP (2456x1372) frame. We do not use any row or column skipping. We simply downscale H and V by 64% for the F3 and 55% for the FS100 which is the "over-sampling Juan talks about. During the down-scale, all pixels are de-bayered.

An HD progressive Kell factor of 0.97 is used.

The estimated F3 resolution will be 858 LW/ph by 852 LW.

The estimated FS100 resolution will be 737 LW/ph by 732 LW.

So the photosites are 4X larger. And, the estimates seem to match your numbers quite well.

There remains one issue -- why if Sony has told us the chip is 3.37MP (2456x1372) frame, why does Juan say he can't tell us the sensor's specs? What's left that we don't know?

PS1: My math model also estimates EX1 1080p at 1048 LW/ph and 1048 LW.

PS2: I suspect the VG10 may work differently. Need to do more work.

[Alister Chapman]

Steve, couple of questions.

What is TW/PH and where do the downscale numbers come from?

Could the bayer matrix be twisted somehow?

[Brian Bang Jensen]

Now I am confused!!

In my understanding a picture pixel is a picture element consisting of 3 elements R+G+B.

Are you now telling me that Sony has pulled the same old gimmick they used to use to with the flat screens, using the combined numbers of RGB elements to lead me as a costumer to believe that the product has higher specifications that it actually has.

I have actually preordered a FS100 in the believe, that I got a genuine HD2 camera!!
If it for some reasons only resolves 750 lines horizontally I can live whit that, but if Sony is twisting the specifications to lead me to believe the camera is something it isn’t. I feel betrayed.
If the sensor is only 1.1 mp coming from 1.1mpR+1.1mpG+1.1mpB.
Then the specification should say sensor pixels used 1.1mp!!!

Does any of you actually knows the provided specks from Sony is the combined numbers of R+G+B on the sensor, putting the camera in the HD1 category, or is it something you assume from the resolution you can see from the camera?

Now you probably says it doesn’t matter if the sensor is HD1 or HD2 if the resolution is only HD1. Maybe it is so in theory, but my experience in real life is that a sensor with a high native pixel count is going to produce an optically nicer image than a sensor with lesser pixels. Giving they is mounted in a camera with the same optical resolution. What I am trying to say is, opsampling is bad, downsampling is fine.

[Brian Drysdale]

I would take it that the sensor has a standard Bayer arrangement, same as the RED and the Arri Alexa, which has 2880 x 1620 Pixels for the image area.

Bayer filter - Wikipedia, the free encyclopedia

I'd assume that the Alexa has a lot more processing going on for its higher resolution figures.

[Steve Mullen]

Originally Posted by Alister Chapman

Steve, couple of questions.

What is TW/PH and where do the downscale numbers come from?

Could the bayer matrix be twisted somehow?

Sorry Alister, my typo. Should be LW although I prefer TVL.

And yes, the secret may not be -- as I have been assuming -- the photosite count of the F3 sensor, but the arrangement of the photosites on the silicon die.

1) Some think the chip could be a Super 35 version of the ClearVid design we both know.

2) Some think the chip could be a Super 35 version of the Q67 design. Here is an edited comment on it.

"After complaining about Bayer Pattern sensors in digital cinema cameras, Sony magically think that rotating a Bayer Pattern 45degrees somehow makes it better. When you look at Sony's "Simulation" image, their Q67 is using 17.7mp compared to their Bayer example of 8.8mp. The Q67 is basically using twice as many pixels."

3) Some think the chip could be a Super 35 version of the "RGB Stripe" design. Here is an edited comment on it.

"About Sony's "Full RGB" solution they use in the F35. At high levels of detail, the RGB Stripe pattern produces rainbow patterns due to the lack of alignment between the three channels. This is plainly visible."

==========

The difference between the 64% value for the F3 verses the 55% for the FS100 may come from a pure marketing decision, or given there are many ways to debayer and some ways yield both higher resolution and lower artifacting -- I'm inclined to guess the F3 uses a more powerful algorithm. This likely requires a more powerful DSP. Conversely, the FS100 (and VG10?) use a simpler algorithm that requires a less powerful DSP.

I have a question for you about resolution measures for the VG10. Did you obtain any?

And, does anyone have from Panasonic the AF100's actual sensor row by column numbers?