alexO

Does a7c still have 1.2X crop in 4K mode at 30fps compared to a7iii? Can't find it anywhere in the specs, not sure if they managed to do full sensor or just decided not to mention it.

Too bad 1.2x crop in 30fps. Most likely meaning that readout could not keep up in 30fps, as a result expecting high rolling shutter on the level of 35ms(similar to a6300/a6500). Quote from dpreview: "24p 4K video is taken from a 6K region of the sensor, while 30p shooting comes from a roughly 5K region with a 1.2x crop".

It seems NDA period is over. Weird review by videomaker. Rolling shutter looks the same to me, it seems they just picked the worse frame for a6300 than for a6500. Another one by cameralabs with samples: http://www.cameralabs.com/reviews/Sony_Alpha_A6500/

DPReview reports a6500 having the same crop for 30p in 4K mode. Looks like the sensor is identical to a6300, so I bet the rolling shutter is gonna be the same.

On this dpreview's video still comparison a6300 in HD looks worse than a6000 and significantly worse than let's say NX500 or GH4.

I was watching closeout sale of NX500(500$ US) and thinking if I should pick one up. Used were selling slightly north of 400. Now after sale is gone and NX500 sold out(aside from sellers who still sells it for 800 original retail price) used prices jumped to about 500 and I won't be surprised if they would go even slightly higher. I am sure eventually they would go down, but it probably not gonna happen until late this year if not later.

Actually never mind, RS is the parameter of the single frame, so there should be no correction for fps. So 25p RS should be identical to 24p, while 30p should be 1.225^2=1.5 times smaller(26ms). This makes sense, it should be kept under 1000ms/fps, otherwise the next frame starts before the current one finished processing. So for 30p RS should be under 33ms, while 24p under 42ms.

majoraxis, Andrew confirmed few pages earlier that there is no crop in 25p, therefore RS should be slightly worse. If we assume it scales linear with number of pixels needs to be read it should be worse by 1/24=4.2%. If we take baseline to be 39ms this would give 40.6ms. In case of 30p from sample videos I estimated relative crop to be 1.225, correcting by higher frame rate gives (30/24)/1.225^2=0.83 factor. With the same 39ms baseline it would give 32.5ms.

I was joking, I agree, in perfect world we should be able to control any post processing which is automatically done by the camera. "It's not a bug it's a feature" is usually referred to an excuse made by software developers when they try to convince the user that a flaw in their program is actually what it's supposed to be doing.

It's not a bug, it's a feature!

Is 4K 25p also cropped? a6300 manual on page 138 of pdf says(online: http://helpguide.sony.net/ilc/1540/v1/en/contents/TP0000818593.html) The angle of view will be narrower under the following conditions: When [ File Format] is set to [XAVC S HD] and [ Record Setting] is set to [120p]/[100p] When [ File Format] is set to [XAVC S 4K] and [ Record Setting] is set to [30p]/[25p]

Andrew, did you have a chance to test 4K 30p mode? Aside from the crop, how is it in terms of the sharpness/DR/noise?

Looks like 4K 30p crop is more severe than HD 120p - Jordan(camerastore) says it's around 2 relative to FF, which looks about correct from their review above around 14:10 mark. He also says it's about 1 stop worse.

KrisAK, I don't think HD straight out of camera qualifies as full readout, which assumes reading full 6K sensor(cutting only top and bottom section to get down to a proper aspect ratio) and then do a proper scaling to 2K/1080p. If it was the case, there should be no difference in sharpness between HD out of the camera and downscale to HD from 4K footage, also rolling shutter should be as bad as 4K(it takes roughly 30ms to read the whole sensor). I am not an expert and have limited experience with low level of hardware(I am an ex physicist), but if I have to guess there is some low level logic on a CMOS sensor level which combines all those pixel together to produces HD footage. So scaling is not done by the CPU, but by the sensor itself. Initially we had line skipping, then pixel binning, and now I think it might be a bit more sophisticated, but still not a proper downscale you could archive by computers in whatever editing software you use. So despite all pixels been used I don't think it qualifies as full readout(no one calls binning the full readout, this is just extension of it). Feel free to correct me if someone knows more technical background on sensors. This is just my guess.

When people talk about raw video I don't think they imply it at full native resolution of 28Mp, this would be insane. Look at 5Dm3 raw video for example, each frame is at HD resolution with 14bit of color information. So it's not true raw per say, but a downscaled raw instead. If we had video files with 14bit that would be perfect. No need for any kind of s-log tricks, which improve dynamic range but introduces banding and other issues after grading.