F55: F5/F55 Sensor Size Discussion

[Tom Roper]

I’ve been under the impression the S35 sensor is 1:1 pixel readout, thus 8.8M pixels for 4k-DCI, but the math doesn’t work out for that. Sony literature says the sensor is 11.6M photo sites total, and the sensor dimension of 24.7 x 12.7 mm. That would mean for square pixels the spacing would be 5.2 microns. If the sensor readout was a 1:1 pixel crop, the 16:9 image circle would only be 22.91 mm which is just 80% of S35, and thus the crop factor for lenses would be 1.875:1. That’s more crop factor, more magnification than is there, so it seems instead the image is getting over sampled from 11.3M pixels of image area. That would still only be 91% of the S35 image circle for a crop factor of 1.648:1 which seems more in line with what people are seeing.

 

[starcentral]

Looks like Sony "wastes" 25% of the photosites likely for supersampling to compensate for some resolution that you use debayering. The net effect is a "sharper" 4K net image as opposed to what an actual 4K sensor can deliver.

I guess if Sony wanted to, they could market the F5/F55 as a 5K sensor/camera :beer:

 

[Viddovation]

I’ve been under the impression the S35 sensor is 1:1 pixel readout, thus 8.8M pixels for 4k-DCI, but the math doesn’t work out for that. Sony literature says the sensor is 11.6M photo sites total, and the sensor dimension of 24.7 x 12.7 mm. That would mean for square pixels the spacing would be 5.2 microns. If the sensor readout was a 1:1 pixel crop, the 16:9 image circle would only be 22.91 mm which is just 80% of S35, and thus the crop factor for lenses would be 1.875:1. That’s more crop factor, more magnification than is there, so it seems instead the image is getting over sampled from 11.3M pixels of image area. That would still only be 91% of the S35 image circle for a crop factor of 1.648:1 which seems more in line with what people are seeing.

Take a look at this Sony document, particularly page 11. It specifies that the F5 has a sensor with a total of 11.6 million photosites, but only 8.8 million *effective* photosites.
https://pro.sony.com/bbsccms/assets/files/show/highend/pdf/F5_Camera.pdf

The 11.6 really is a marketing misnomer. It's technically true, but it isn't meaningful to us users. Image sensors virtually always have masked photosites that don't compose part of the image. There a variety of reasons for this, and cameras use these masked photosites for various internal tests.

So it really is an 8.8/8.9 (depending on how you round) million photosite sensor for our purposes.

What's less clear is whether the sensor size given of 24x12.7mm represents the size of the 11.6 million photosites, the 8.8 million effective photosites, or possible the outer dimensions of the physical sensor chip.

So the math definitely works out for 1:1, 8.8 pixels for DCI-4k. As for the physical size, I'm not finding a definitive reference.

 

[Viddovation]

Looks like Sony "wastes" 25% of the photosites likely for supersampling to compensate for some resolution that you use debayering. The net effect is a "sharper" 4K net image as opposed to what an actual 4K sensor can deliver.

I guess if Sony wanted to, they could market the F5/F55 as a 5K sensor/camera :beer:

Since it's only 8.8million effective photosites, the first part of this doesn't sound right. The chip does probably do diagnostics with the masked pixels, but not super sampling.

There may be enough photosites to offer a 5k camera with the F5 sensor chip, but that doesn't mean the F5 hardware is wired to do so - whether or not there is processing power. The Red One MX chip got reused in the Red EPIC, and made use of 5k by delving into the photosites that had been masked on the Red One MX.

 

[Tom Roper]

James the only way the math works out for 8.8M is if 24x12.7mm is the dimension for the cropped portion of the sensor, in which case the full sensor is larger.

 

[Viddovation]

James the only way the math works out for 8.8M is if 24x12.7mm is the dimension for the cropped portion of the sensor, in which case the full sensor is larger.

Then there you go. Just like every other 17:9 4096x2160 capturing camera its using 8.8MP to generate the image. No super sampling from 11.6MP. So that answers your sensor size question by default then.

 

[starcentral]

Yeah after reading the PDF it seems that perhaps the sensor could be "square" but only a super 35mm size from it is used.

 

[Ebrahim Saadawi]

You're under the impression that the fs700/fs7/f5/f55 sensor is 11.6mp all in a 17:9 chip. I believe the 11.6 mp sensor is 3:2ish aspect ratio and with a 17:9 crop being 8.9 mp, as on the a7s 12mp sensor with 8.3mp native UHD crop. The extra masked pixels have many uses and exist in most cameras.

Put a lens on the camera (fs700/fs7/f5/f55), put it on any s35 camera or DSLR, and you'll find it is indeed s35, just like c100/300/alexa/etc and not a crop from s35. That what matters, real results of image size.

(Of course there are tiny variations among s35 sensor sizes but not noticeable and it's not the subject here)

 

[Ebrahim Saadawi]

Here are the numbers I am not into math so you guys figure out the aspect ratio and crop factor:

Total pixels: 4352 x 2662 = 11.59mp

Physical chip dimensions: 24.2mm x 14.8mm

Area used for 16:9 s35: 23.6mm x 13.3mm

 

[Tom Roper]

Here are the numbers I am not into math so you guys figure out the aspect ratio and crop factor:

Total pixels: 4352 x 2662 = 11.59mp

Physical chip dimensions: 24.2mm x 14.8mm

Area used for 16:9 s35: 23.6mm x 13.3mm

If the 11.59mp is based on 24.2 x 14.8 mm (as you've given), the spacing of photo sites is 5.559 micron. Thus, the number of photo sites used for s35 23.6 x 13.3 (16:9) would be 4246 x 2392 for 10.16 MP. That would represent oversampling of 8.3mp (16:9) by 22.4%, and the crop factor would be 1.58:1 so these numbers appear valid.

 

[Tom Roper]

These technical discussions can sound confusing, but if 8.3mp were fit into the area used for 16:9 s35: 23.6 x 13.3mm; the dimensions listed above by Ebrahim (thank you); then the sensor spacing (too big) would not allow 11.6mp to fit onto the physical chip dimension (too small) he gave. The 16:9 sensing area has to be super sampled.