Imager info from Service Manual

[Chris Medico]

I set up the F3 yesterday to do a demo and for the first time put it in 60i mode so we could use the HDMI and SDI outputs simultaneously. What popped up immediately on the monitors was a BRIGHT WHITE stuck pixel on the right third of the image near the horizontal midline. I switched the camera to 24p and it seemed to disappear. It was still there but very faint in comparison to the 60i mode. It was made more noticeable because we were shooting in room with very dark wood paneling and this was the first time I had put the camera in 60i.

I called Sony and they were nice enough to send me the info to fix it via the service menus (the black balance procedure in the user menu didn't take care of it).

What I found interesting was the info about the pixel addressing for the RPN procedure. There is a section where you can go in and manually select the location of the bad pixel if the "auto" RPN procedure in the service menu doesn't work. The limits for H and V pixel positions are as follows H = 1 to 2468 and V = 1 to 1398. This is directly out of the service manual.

Based upon a simple multiplication of the numbers that comes out to 3.45 megapixels. The service manual specifically uses the word "pixels" in the procedures.

Do you think this improves our understanding of the resolution of the imager in the F3?

[Nate Weaver]

Originally Posted by Chris Medico

What is interesting there is the limits for H and V pixel positions are as follows H = 1 to 2468 and V = 1 to 1398. T
Based upon a simple multiplication of the numbers that comes out to 3.45 megapixels. The service manual specifically uses the word "pixels" in the procedures.

Do you think this improves our understanding of the resolution of the imager in the F3?

Or 13.8 megapixels, if you count pure photosites, Red style. 4936x2796. 5K!

I don't know if that jives with the previous pixel sizes mentioned, which I can't recall.

Regardless, I just realized it doesn't matter to me! :-)

[David Heath]

Originally Posted by Chris Medico

Do you think this improves our understanding of the resolution of the imager in the F3?

Well spotted!

I think this should be the final word on the subject. Early on, Sony were saying the sensor was about 3.36MP effective, which works out to be about 2456x1372. I'd expect "effective" to be a little less than the actual count, so 2468 and 1398 are pretty well exactly what may be expected of the actual count.

[Chris Medico]

Relieved that the instructions from Sony worked. Hot pix now masked.

[Alister Chapman]

Nate: RPN will almost certainly be working at the photo site level as it will have an impact on the De-Bayer process. The camera most likely simply addresses an adjacent photo site as opposed to the defective one. So pixels almost certainly equals photosites on the F3.

[Nate Weaver]

Originally Posted by Alister Chapman

Nate: RPN will almost certainly be working at the photo site level as it will have an impact on the De-Bayer process. The camera most likely simply addresses an adjacent photo site as opposed to the defective one. So pixels almost certainly equals photosites on the F3.

I myself wouldn't make that bet. In addition, there's probably processing power issues that would make it easier to deal with dead pixels AFTER debayer, rather than before.

But most of all, that would imply that pixel res after debayer is 1200something by 650something, and the camera seems to look a lot better/more oversampled than that.

[Alister Chapman]

De-mosaicing interpolates diagonally, horizontally and vertically adjacent pixels. So if you don't eliminate the defective pixel prior to the de-mosaic it will affect the colour and luma output over an area much larger than a single pixel or colour/luma point. It's extremely simple just to read an adjacent pixel prior to processing as opposed to trying to compensate post de-mossaic for several incorrect results that may span a 3x3 interpolated block.

The resolution obtained from a bayer sensor is typically between 70 and 80 percent of the horizontal pixel count, so 2468 pixels or photosites should yield around 1800 - 1950 pixel (900 - 975 LP/PH) resolution which is pretty much what the F3 measures at. Don't forget this is not a dissimilar pixel count to Alexa know pixel count, which again has very similar resolution and aliasing performance.

Do you not think that Sony would have been shouting out loud if their new baby used a "5k" or 14MP sensor, just like they are making sure everyone knows that the F65 uses an "8k" sensor.

[Noah Yuan-Vogel]

Originally Posted by Nate Weaver

Or 13.8 megapixels, if you count pure photosites, Red style. 4936x2796. 5K!

I don't know if that jives with the previous pixel sizes mentioned, which I can't recall.

Regardless, I just realized it doesn't matter to me! :-)

No. It's still ~3.36MP (photosites) as the original Sony literature said months ago.