80D is here

sploo wrote in post #17943616
I do also wonder if a lot of the problems with 4K video on smaller devices is down to the quality of the compression algorithms. Broadcast bit rates for 4K seem to be around 40Mb/s (5MB/s), so well within the capability of even an SD card - but, you've obviously got to compress those 4K frames in real time. I suspect a lot of corners need to be cut in order to do that on a smaller device. Probably also the reason why the 1Dx II is using MJPEG (essentially, just a flick book of photos) - but then the bandwidth requirement is massive.

I assume compression and dynamic range would be the major downsides of the smaller cameras in addition to heating issues when designing a CPU to process all the compression without proper thermal design (the 1DXII seems to be using a dedicated heat sink for this purpose). Also, jello effect is horrendous on many of the smaller systems; not sure why that is compared to Cinema EOS.

Shooting wrote in post #17943167
May be old but true. You don't need video in a camera just to have live view shooting. You can take out the video and still have that technology. Take out the video especially the programming and there is lots of room for more still shooting programming. I mean a sensor/chip can only hold so much Like putting in the technology to shoot at different croppings, is that in a camera yet? If I want all my shots to be 8x10 crop then there needs to be a choice on the menu to select 8x10, 5x7, 11x14, etc, what crop you want your files to be. I know a place that makes a mirror with cropping lines but to actually shoot the crop you want would be nice. Lots of stuff to take the place of video programming.

Do you program cameras yourself? I'm trying to understand how you could make an assumption that because of video programming other features are potentially being left out.

Wilt wrote in post #17943444
Human eye is said to resolve 0.5 seconds of arc...It can detect a single dot ...

• at 10' distance on a 1080p screen which is 33" across
  
• at 20' distance on a 1080p screen which is 66" across
  
• at 30' distance on a 1080p screen which is 100" across
  
• at 40' distance on a 1080p screen which is 133" across
  

The stuff I found on this states at 20/20 vision you can resolve 60 pixels per degree or 1 seconds of arc which is half of what you claim and would double your screen sizes.

There is a degree of disagreement about HALF arc-second vs. FULL arc-second between the sources of information, but that is due to interpretation of the information presented.

Some aspects can easily be quantified...In one explanation https://www.nde-ed.org …oduction/visual​acuity.htm it says

• "the eye as a sensor has the maximum concentration of cones of about 180,000 per square mm in the fovea region.
  
• "The standard definition of normal visual acuity (20/20 vision) is the ability to resolve a spatial pattern separated by a visual angle of one minute of arc. Since one degree contains sixty minutes, a visual angle of one minute of arc is 1/60 of a degree.
  
• "The spatial resolution limit is derived from the fact that one degree of a scene is projected across 288µm of the retina by the eye's lens. "
  

• "At 12 inches, the normal visual acuity of the human eye is 0.00349 inch. What this means is that if you had alternating black and white lines that were all 0.00349 inch wide, it would appear to most people as a mass of solid gray." ...sounds like a line-pair, does it not? that takes three pixels to represent a line-pair. OK, to continue on...
  

Wilt wrote in post #17944261

"At 12 inches, the normal visual acuity of the human eye is 0.00349 inch. What this means is that if you had alternating black and white lines that were all 0.00349 inch wide, it would appear to most people as a mass of solid gray." ...sounds like a line-pair, does it not? that takes three pixels to represent a line-pair. OK, to continue on...

Wilt here you are incorrect, it only takes TWO pixels to represent a line pair, not three. One pixel of the white line, one pixel for the black line. Remember that you are not looking at a series of black lines on a white background for this test. You are looking at alternating black and white lines, with no background at all. So for example if you have 15 line pairs, you need 30 pixels to record them, or more generally 2n pixels, when n is the total number of line pairs. This is what gives us the Nyquist limit in the Nyquist Shannon Sampling Theorem, which states that for any continuous signal of frequency f the sampling rate must be 2f samples.

Alan

DPreview just posted DR data
seems that the 80D is much better than the 70D - but not as good as the Nikon D7200 - however the gap is now closing - i shall get the 80D in a couple of months !!!

link is here: http://www.dpreview.co​m …nd-for-canon-low-iso-dr/2

Canon always seems to do better with the 2nd or 3rd body with a new sensor, so I cannot wait to see what comes out for the 7D or 5D replacements. That is probably where my money will go. The 80D is good, but Canon will learn how to tweak hardware and software once the camera hits the masses, and the next body will probably sport more efficiency than the 80D.

coolshot wrote in post #17902840
Is 4k really important? I am a tech enthusiast, but I don't have a single 4k panel in the home. I have 3 LCD tvs in the house and 14 displays(including family tablets,laptops & phones). Short of someone buying midrange 4k tvs from Costco, how they watching 4k content?

My sentiments exactly!

Teamspeed
I wonder what will be the DR of the 1DX2
Will it lag behind the D5?

BigAl007 wrote in post #17944327
Wilt here you are incorrect, it only takes TWO pixels to represent a line pair, not three. One pixel of the white line, one pixel for the black line. Remember that you are not looking at a series of black lines on a white background for this test. You are looking at alternating black and white lines, with no background at all. So for example if you have 15 line pairs, you need 30 pixels to record them, or more generally 2n pixels, when n is the total number of line pairs. This is what gives us the Nyquist limit in the Nyquist Shannon Sampling Theorem, which states that for any continuous signal of frequency f the sampling rate must be 2f samples.

Alan

Are you guys talking about sensor pixels or the pixels of the final display image monitor? 'Cause monitor pixels is what the customer is going to see....

("Let it go, let it go, let it go....")

I would suspect so, as this is the first round of the Canon sensor, and the Sony has had plenty of years under its belt for any refinements.

omer wrote in post #17944561
Teamspeed
I wonder what will be the DR of the 1DX2
Will it lag behind the D5?

TeamSpeed wrote in post #17944645
I would suspect so, as this is the first round of the Canon sensor, and the Sony has had plenty of years under its belt for any refinements.

Well now here's a thing: it seems there's some consternation that the D5 has traded ultimate (i.e. low ISO) DR for improvements at higher ISOs: http://www.dpreview.co​m/forums/thread/398154​7

Makes sense, given the intended use of the camera, but interesting nonetheless.

sploo wrote in post #17944670
Well now here's a thing: it seems there's some consternation that the D5 has traded ultimate (i.e. low ISO) DR for improvements at higher ISOs: http://www.dpreview.co​m/forums/thread/398154​7

Makes sense, given the intended use of the camera, but interesting nonetheless.

I'm not getting too excited about the D5. It has horrible banding at ISO 3 million in the sample RAWs I downloaded, and that is a push from ISO 51K, so that means that it is there at 51K, too, in the shadows. DxOMark won't notice banding, though. If Canon keeps banding below visibility in the 1DxII, it might actually be better than the D5 at ISO 3 Million, even if DxOMark suggests otherwise.

John Sheehy wrote in post #17944943
I'm not getting too excited about the D5. It has horrible banding at ISO 3 million in the sample RAWs I downloaded, and that is a push from ISO 51K, so that means that it is there at 51K, too, in the shadows. DxOMark won't notice banding, though. If Canon keeps banding below visibility in the 1DxII, it might actually be better than the D5 at ISO 3 Million, even if DxOMark suggests otherwise.

A drop in low ISO DR. Banding in the shadows. Are we sure Canon and Nikon haven't accidentally swapped sensors for their new cameras

Being serious though: the ISO 3 mil is just a marketing number, so (for that type of body) surely it's the ISO 1600-12800 range that's really going to matter for their target market?

On a slightly different subject: I was having a play with some of the example images here http://www.dpreview.co​m …nd-for-canon-low-iso-dr/2 and it's definitely a case that the 80D is much better than the 7DII when pushing a low ISO image by several stops, but ISO 3200 looks significantly better to me on the 7DII. As such, there's a crossover, that to my eyes looks to be somewhere around ISO 400 + 3 stops; where the 7DII starts to be better than the 80D. Is that to do with the differences of pre vs post-gain noise that you mentioned earlier?

sploo wrote in post #17944984
A drop in low ISO DR. Banding in the shadows. Are we sure Canon and Nikon haven't accidentally swapped sensors for their new cameras

Being serious though: the ISO 3 mil is just a marketing number, so (for that type of body) surely it's the ISO 1600-12800 range that's really going to matter for their target market?

Of course, but Nikon's implementation of ISO 3 million is going to give such ISOs a worse reputation than what they are capable of. It's not just the banding; they use ISO 51K's analog gain and under-expose by 6 stops, and then multiply all the RAW values by 64. The blacks are clipped, and even blacks clip the highlights, because the noise histogram for zero signal is bigger than the 16384 steps of the RAW! So, you have non-linearity at both ends of the histogram. Not a very practical implementation, IMO.

On a slightly different subject: I was having a play with some of the example images here http://www.dpreview.co​m …nd-for-canon-low-iso-dr/2 and it's definitely a case that the 80D is much better than the 7DII when pushing a low ISO image by several stops, but ISO 3200 looks significantly better to me on the 7DII. As such, there's a crossover, that to my eyes looks to be somewhere around ISO 400 + 3 stops; where the 7DII starts to be better than the 80D. Is that to do with the differences of pre vs post-gain noise that you mentioned earlier?

Yes. The 80D has less post-gain noise, and it is more random-looking, to. At higher ISOs, the post-gain noise has a very reduced contribution to total read noise.