Sony 1" or Panasonic HC-X1000?

[David Heath]

Originally Posted by Cliff Totten

The Fs7, F5/F55 and FS700 uses 13+MP. Again, not too crazy.

Not so - all those cameras (plus others like the s35 Varicam) have sensors with active pixel counts of (4096)3840x2160 - what are generally referred to as "8 Mp" sensors. If 8Mp is good enough for a F55.......??

Originally Posted by Cliff Totten

The AX100 and other 1 inch-type models use almost 17mp to start with and scale down. So, that's close to a 2X over sample (2:1) as opposed to Panny's 1:1.

But the same oversampling can give rise to more problems than it solves. It CAN.... which is not to say it WILL....

It must involve a downsampling of the deBayered intermediate, and digital downsampling needs a lot of processing to do well.

In terms of the question posed, then I agree that 1" is far more likely to give better results than much smaller sensors - but that doesn't necessarily mean it's down to pixel count, period. For starters there are factors like diffraction softening, and don't forget more photosites means smaller photosites.... even more of a problem for smaller chip sizes.

[Cliff Totten]

I stand corrected. I could have sworn that I had seen a higher pixel count listed on these cameras in the past. huh?

Well? Sony's web site has these specs listed:

FS7 "11.6 MillionTTL 8.9 in 4K; 8.3 Million in UHD snd HD"

Alister wote this about it:

http://www.xdcam-user.com/2012/07/ne...d-sensor-size/

It's widely speculated that the FS7 uses the same sensor as the FS700.

The FS7 is cropping an 8mp scan from the middle of the sensor? What about the "unused pixels?"

Need to look into this further.

Very interesting. Thanks for pointing that out....

[David Heath]

Originally Posted by Cliff Totten

I stand corrected. I could have sworn that I had seen a higher pixel count listed on these cameras in the past. huh?....

What you say above has always been the case - about 8.3Mp (effective) for QFHD, and about 8.8 Mp for 4K - and it's the "effective" count that's relevant.

As Alister says in the link you gave: "It is normal to have some extra pixels that are used for setting black levels etc, but this is a massive difference between the number of actual pixels on the sensor and the number that are used to create the pictures." All very true.

Theories I heard at the time ranged from to give the ability for outside recorded image viewfinding in a future camera, to giving flexibility to compensate for dead pixels (and so decrease the number of unusable chips), to their being used as part of the stabilisation system.

Which is correct, if any? I don't know, but in the comments at the end of that link, then usage for stabilisation seems the favourite. Practically, it may be interesting to speculate, but from a usage point of view it should just be considered as 3840(4096)x2160. It's always the "effective" number that's important from a usage point of view.

[Piotr Wozniacki]

Interesting discussion; even though I'm an engineer my background on this particular subject is nothing compared to yours so I can't add anything of value. But coming back to my original question, here are 2 screen grabs from an X1000 file I got a link to as being native and without any recompression. While I can understand the softness in some (well - most of) the YT/Vimeo samples (heavy compression etc) - these grabs come from what is supposed to be a native AVC, High@5.2, MP4-wrapped, 4:2:0 3840x2160 clip at 24p with some 40 Mbps data rate - straight from the Panasonic camera. BTW, for some reason I was unable to upload any of those 3840x2160 PNG pictures, so I'm just providing a link to them on another server.

What do you make of the muddiness of the distant foliage? OK - partially it's due to blur from panning or zooming present at the moments I took the grabs - but still... It looks more like SD, not even HD - not to mention Ultra HD to me! Is this really how the Panasonic X1000 records? In the same clip there some scenes with lots of details and the UHD "wow" factor - but they are in minority... Was the camera faulty (back-focus?), was it an operator error - or is the codec implementation totally screwed up?

http://www.moldcad.idsl.pl/images/x1000.png
http://www.moldcad.idsl.pl/images/x1000-2.png

[Ron Evans]

The X1000 and the AX1/Z100 are effectively 1/3 " cameras so using iris above about F4.6 or so will lead to soft image. ND needs to be switched in quite early, the camera tells the user !! On my NX5U and AX1 I try to keep iris at about F4 or larger (ie F2.8 ) for a good image. So unless the files all say exactly how they were shot it is almost impossible to compare. I can make all my cameras look great or absolutely terrible just by changing the exposure parameters.. True for my NX5U, NX30U, FDR-AX1 , the AX100 has a lot more range before it goes soft. As an example above F5.6 and the NX5U starts to look like a very cheap camera and has a hard time competing with my Xperia cell phone !!! Which may have the same sensor as the AX1/Z100 and/or the X1000 !!! I am sure the HX30V still camera I have does have the same sensor but a very much smaller lens as does the cell phone.

It is useful to look at the data code from a pure automatic shot to see what the engineers say should be the correct exposure. Most of the time on the Sony's I have that iris value is around F2.8 for the 1/3" cameras when possible. For the smaller cameras without ND the shutter speed is increased before iris closes down below F4. A good indication that I have followed to get sharp images.

For cameras with full manual controls unless they are compared with the same settings under the same conditions then it is not a lot of use !!! 4K is very difficult to focus and to me both those images are out of focus.

Ron Evans

[David Heath]

Originally Posted by Piotr Wozniacki

What do you make of the muddiness of the distant foliage? OK - partially it's due to blur from panning or zooming present at the moments I took the grabs - but still...

I wouldn't want to draw any conclusions either way. One just looks out of focus (or showing softness due to diffraction effects?), the other simply looks rough due to camera movement (right hand side is sharp, left very blurred, so I'd suspect zooming and panning at the same time?) They don't look good quality but whether it's anything intrinsic to the camera, or simply the way they are shot is impossible to say.

It all goes to show why the best tests use such as charts or special test scenes - they don't just give an idea of how "good" or "bad" a result is - they're specifically designed to tell why.

[Cliff Totten]

Originally Posted by David Heath

What you say above has always been the case - about 8.3Mp (effective) for QFHD, and about 8.8 Mp for 4K - and it's the "effective" count that's relevant.

As Alister says in the link you gave: "It is normal to have some extra pixels that are used for setting black levels etc, but this is a massive difference between the number of actual pixels on the sensor and the number that are used to create the pictures." All very true.

Theories I heard at the time ranged from to give the ability for outside recorded image viewfinding in a future camera, to giving flexibility to compensate for dead pixels (and so decrease the number of unusable chips), to their being used as part of the stabilisation system.

Which is correct, if any? I don't know, but in the comments at the end of that link, then usage for stabilisation seems the favourite. Practically, it may be interesting to speculate, but from a usage point of view it should just be considered as 3840(4096)x2160. It's always the "effective" number that's important from a usage point of view.

Yeah,...this is quite interesting. I looked at the Sony FS7 brochure and this is what is stated:

"The FS7 is equipped with a Super 35 “Exmor” CMOS Sensor
with approximately 11.6 million total pixels and 8.8
million effective pixels. The high image readout speed
of the image sensor allows the FS7 to support 4K motion-picture
shooting and Super Slow Motion. The sensor
also realizes a high sensitivity of ISO2000 and a wide
dynamic range of 14 stops. Thanks to its full-pixel readout
capability without pixel binning and sophisticated
camera processing, jaggies and moiré are minimized"

Are we 100% sure that Sony is reading only 1:1 for UHD and DCI 4K?

There are a couple of points I can see here:

11.6 total pixels and 8.8 effective...what is Sony doing with those 3+ million pixels? The FS7 does not read the full sensor width?

Sony mentions "full pixel readout and no binning" This strongly suggests the camera IS reading all the pixels and scaling. Otherwise, a standard 1:1 readout doesn't need this to be stated. And of course there would naturally not be any binning if it's 1:1 either. Sophisticated camera processing? This doesn't sound to me like a standard 1:1 readout and 8 megapixel deBayer process. It makes me think "sophisticated SCALING engine" Maybe its just suggestive marketing talk?

I'm not saying I know the answer, but this language suggests to me that it's scanning all 11.9 and using the entire sensor width and delivering a "final" 8.8mp image after it's all processed.

I know Sony likes to oversample these days. If a resolution target is "X", they generally aim for 20% (or more) over that target to compensate for the Bayer pattern loss and arrive close to the "X" resolution they want. (This commonly results in noticeably more detail but a bit more moire too).

It's very hard to make a razor sharp and detailed UHD/4K (8+MP image) if your green channel only has 4 million pixels (HALF UHD/4K raster) and your red and blue channels are only around 2 million pixels each too...(25% UHD/4K raster).

I don't know at all how Sony is doing any of this. It's cryptic to say the least.

[David Heath]

Originally Posted by Cliff Totten

Are we 100% sure that Sony is reading only 1:1 for UHD and DCI 4K?

There are a couple of points I can see here:

11.6 total pixels and 8.8 effective...what is Sony doing with those 3+ million pixels? The FS7 does not read the full sensor width?

In short, yes, 100% sure!! :-)

Key is the word "effective", that unequivocally means "the number in use". It's been confirmed in testing, and in answer to what it's doing with the other 3 million, then well... that's what is not known.

Originally Posted by Cliff Totten

It's very hard to make a razor sharp and detailed UHD/4K (8+MP image) if your green channel only has 4 million pixels (HALF UHD/4K raster) and your red and blue channels are only around 2 million pixels each too...(25% UHD/4K raster).

I don't know at all how Sony is doing any of this. It's cryptic to say the least.

Using a Bayer sensor of 3840x2160 is considered the minimum that can be used and still legitimately be called "4K". But yes, a 4K Bayer sensor won't give as much resolution as the 4k system is capable of carrying, that's true.

But it's not as bad as you fear. It's basic theory that after deBayering the luminance resolution from a Bayer sensor is likely to be about 75-80% of the sensor dimensions, and the chrominance resolution about 50%. (Which is why, in a progressive system, and using such a sensor, 4:2:2 sampling is irrelevant and won't give any advantage over 4:2:0. The extra chroma resolution it could record is not there in the first place. Worse, it just means extra bandwidth - or higher compression - for no benefit.) It's a question of where you draw lines. All I'll say is that in resolution terms, even if that is less than the 4K system is capable of, it's a lot better than 1080