How to increase Canon 7D dynamic range
- Thread starter marco0782
- Start date
marco0782
Well-known member
Well everything gets confused when someone does extensive chart testing to determine that a camera only has 7-8 stops of latitude. Not to say I'm not grateful for the time spent, because I am, but an extensive test like that sort of cements results as fact, even though with a custom picture profile the test could have demonstrated closer to 11 stops of latitude.
Marco
Marco
Barry wasn't using the Standard Pictuer profile. He pretty much went by Stu's.
And 11 usable stops? Come on now...
Barry_Green
Moderator
Point me to the scene file you want tested and I'll retest. I've downloaded the "superflat" one -- is there any other you want it compared to, to see these 11 usable stops?
marco0782
Well-known member
Where are the results of your 'superflat' test? I will perform a test in the next few days. 11 stops is my guess but I could be off. What I do know is that it will definitely be more than 7-8 as you reported.
Marco
PS Not trying to be confrontational. I just don't believe 7-8 stops based on my experience.
Marco
PS Not trying to be confrontational. I just don't believe 7-8 stops based on my experience.
Last edited:
Barry_Green
Moderator
I reported exactly what I found, which is exactly what the camera does with its onboard scene files. I tested every one of them, against a calibrated test chart. There really is no margin for error there.
If the superflat adds dynamic range, I'll find that out too. I have not yet run that, may get to it tomorrow.
marco0782
Well-known member
Here's the thing. The 'superflat' profile doesn't add dynamic range. It's the built-in picture profiles that REDUCE dynamic range by increasing the contrast. The 7D has a ridiculous native dynamic range, similar to the Red camera, which is evident when you shoot stills and edit them RAW. By using the 'superflat' picture profile you keep the camera from crushing everything, and since this is done prior to compression, you end up with similar latitude to shooting RAW.
I would really appreciate you testing it, but if you don't have time I'll order myself one of the Stouffer charts. I predict at least 10 stops of latitude.
Marco
PS I wish that Canon was more up front about the potential of these custom picture profiles. With stills photography, at least we have RAW which will let us utilize the full latitude of the sensor. But with video, we need a picture profile such as 'superflat' to reap the same benefits as RAW prior to compression because compression is basically the point of no return.
Last edited:
It is the point of no return, right. And grading 7D footage after that point is certainly no RED affair. So if you shoot flat and try to grade back to where you want it to be, you suffer more image destruction as opposed to just getting it in camera.
We've done pretty extensive grading tests in house through several different methods and the result, for serious acquisition-- if not all-- you capture as close as you can in-camera.
But I'm interested in seeing a usable 10 stops out of this camera in video mode.
marco0782
Well-known member
When you transcode H.264 (native 7D .MOV files) into ProRes 422, the result is almost lossless. And ProRes can be re-compressed up to approximately ten times before you start to see degredation.
Adding contrast to a ProRes image will not increase compression artifacts, nor will it result in any other type of artifact. It is when you try to lift the gamma that you will see it start to break up because there was no information in the shadows, etc.
So when you shoot as flat as possible in camera, obviously you still end up with a compressed H.264 file, but the compression doesn't really hurt you in post because adding contrast does not produce artifacts.
Marco
Adding contrast to a ProRes image will not increase compression artifacts, nor will it result in any other type of artifact. It is when you try to lift the gamma that you will see it start to break up because there was no information in the shadows, etc.
So when you shoot as flat as possible in camera, obviously you still end up with a compressed H.264 file, but the compression doesn't really hurt you in post because adding contrast does not produce artifacts.
Marco
Lammy
Well-known member
I haven't got a 7d but this is not the experience I have with the d90 and it should also apply here.
When the image is flat, there is so much more highlight and shadow detail preserved. Then when you colour grade that to where you want it to be, you can still see the highlight and shadow detail and it still exists. Transcode that to whatever format you want after.
Whereas if you tried to get the look in camera, all you will see in the highlight and shadow areas is increased macro blocking and mushiness. You know it already dude, they're in your sticky, and Mattsand and Martin Lang have posted results from their own flat custum curves.
I shoot toward flat because I like flat images. Not because I'm shooting for a grade. At the very most it's a contrast boost or saturation boost but I am NOT a fan of heavy post grading unless for a hyper-stylized image.
Just for clarifacation as to why I choose to shoot flat(er).
I do see chroma noise and mid noise in images that've been caked in post. Converting to ProRes doesn't help the noise already captured that's obviously there. Just nudge the gamma a bit to see.
Barry_Green
Moderator
Okay, decided not to wait. I went to the cinema5d forum and downloaded the "superflat" scene file (.pf2) as recommended in that video. I installed the EOS utility. I downloaded the superflat scene file to the 7D. I went into the camera and confirmed that I was, indeed, shooting with the superflat scene file, on custom user file 1.
Then I re-shot the stouffer.
Here are the results.
Then I re-shot the stouffer.
Here are the results.
Duke M.
Veteran
This is the T4110 chart, which is in 1/3 f stop increments. You can read about the charts here, an excellent link:
http://www.imatest.com/docs/q13.html
My read of image was 24 increments usable DR, though I think I see hints of disernable increments below that. [EDIT: when magnified I can sample the increment edge to 27, which is 9 f stops.] An argument could be made for a usable maximum of 8.5 f stops. Also, there could be some loss looking at it on a monitor for an uploaded file like this.
However, for the sake of discussion 24 / 3 = 8 f stops of usable DR. Not bad at all. As for the Red being 11.3 stops, as I recall that was Red's measurement. Other tests have been as low as 8 for the Red (I assume as usable DR), but generally seem to be 9-ish range. (I prefer to try to figure usable DR rather than absolute max, but they are all indicitive.)
Red of course wins, but the 7D is showing good DR, and of course DOF control similar to the Red.
http://www.imatest.com/docs/q13.html
My read of image was 24 increments usable DR, though I think I see hints of disernable increments below that. [EDIT: when magnified I can sample the increment edge to 27, which is 9 f stops.] An argument could be made for a usable maximum of 8.5 f stops. Also, there could be some loss looking at it on a monitor for an uploaded file like this.
However, for the sake of discussion 24 / 3 = 8 f stops of usable DR. Not bad at all. As for the Red being 11.3 stops, as I recall that was Red's measurement. Other tests have been as low as 8 for the Red (I assume as usable DR), but generally seem to be 9-ish range. (I prefer to try to figure usable DR rather than absolute max, but they are all indicitive.)
Red of course wins, but the 7D is showing good DR, and of course DOF control similar to the Red.
Last edited:
Barry_Green
Moderator
It means that about 8.5 stops is the limit on a 7D, according to every picture profile I've run through it. Whatever "Luka" did on his video, I cannot replicate his results with the magical super-duper-wide contrast range that he demonstrates.
Thanks to Barry (as usual) and Duke for your insight on the testing. Ending speculation immediately about technicals that need clearing up, obviously.
There's NOTHING wrong with having 8 USABLE stops of range to work with. 8.5 is incredible at this price point.
Great information to have, although I don't like the superflat setting because it looks a little strange to my eyes. Almost like an odd gamma shift.
I'll have to DL it and play with it myself, but I think I'll stick to dialing neutral all the way down instead.
There's NOTHING wrong with having 8 USABLE stops of range to work with. 8.5 is incredible at this price point.
Great information to have, although I don't like the superflat setting because it looks a little strange to my eyes. Almost like an odd gamma shift.
I'll have to DL it and play with it myself, but I think I'll stick to dialing neutral all the way down instead.