For those of you that think the 7D is good with noise...

rhys216 wrote in post #12593775
I exposed the shot manually in a dim room, I basically got the exposure low enough until the image looked nearly black on my LCD.

The below image is how it looked when I imported it into lightroom.

You already posted that image. In order to be useful for TeamSpeed and anybody else who wants to give it a try, how many stops did you have to push it in post and how much underexposure was indicated by the camera's meter?

Tadaaa wrote in post #12593623
It is simple physics; With all other variables held constant, the sensor with the largest pixel pitch wins because that camera gets the best signal.

Of course a 1.6 crop today has less inherent noise being produced by it's self.. and is certainly capable of producing less visible noise than a full frame from 10 years ago. But it won't produce less visible noise than a full frame camera of the same generation, because it's simple physics, i.e. the 7D vs 5Dii

Well since all other variables are never held constant, we might as well move to fantasy land.

TeamSpeed wrote in post #12593721
This example seems to be in very good light actually. How far left did you underexpose to get 1/100th at ISO 100 at f2?

I would like to do the same with my 1D4 and my 30D. If I had a 7D, I would use that as well.

Ok I will be interested in seeing your results.

I have redone the test this time using Av mode with negative exposure dialled down to -5 stops, as it turned out, this is one stop less exposure than the previous example.
Also to get the exposure, the meter was set to evaluate the whole scene (basically the opposite of spot metre) to minimise the meter getting confused by dark colours.

Shadows were lifted in lightroom by increasing the fill light and exposure, and by increasing the curve slider for the 'darks' and 'shadows'.
This time however the lightroom dials limited me, and the picture was still slightly under exposed, so I exported the image as a 16 bit Tiff, saved the image, so that the dials went back to zero in lightroom, then I was able to increase the fill light some more to get the right exposure. I also moved the colour noise reduction slider all the way to 100, but the luminance slider remained at zero.
White balance was tricky to correct as I didn't use a grey card, so I used 'auto colour' in photoshop before saving the image for web.

Exposure settings were:
1/200
F2
ISO100

Below is what I started out with.

MIME changed to 'text/html' | Content warning: script

MIME changed to 'text/html' | Content warning: script

MIME changed to 'text/html' | Content warning: script

TeamSpeed wrote in post #12593645
The 1D4 is of the same generation, however it does a little better than the 5D2.

Generation is a vague term really, which is my point above. It is hard to determine what is in the "same generation". Only 6 months of R&D is enough to create enough of a change in a sensor design or software/firmware so what 2 bodies may seem to be in the same generation could noticeably vary from each other.

Absolutely... I agree with all of that. Ex: The 5Dii has less noise than the classic despite having a smaller pixel pitch. Sensor size isn't the only factor... but it certainly matters.

For whatever reason, DPP (which is all I use) only allows me to bring exposure back up 2 stops, my camera allows 3 stops negative EC unless I go manual. I wonder if I need to upgrade DPP to at least allow me to bring the exposure up +3. I may have to play with the curves then, which I don't know much about in DPP. I will play around a bit.

^^^
If you want I can do the test again, with 3 stops if you prefer?

That is correct. DPP allows you to choose brightness of plus or minus 2 stops ( four total ).
You can grab the edge of it in the curves box and slide it over for a total of 12 stops, which will of course turn your photo into a polar bear in a snowstorm.

george m w wrote in post #12594292
That is correct. DPP allows you to choose brightness of plus or minus 2 stops ( four total ).
You can grab the edge of it in the curves box and slide it over for a total of 12 stops, which will of course turn your photo into a polar bear in a snowstorm.

Yes, I was able to use the curves box and bring out the exposure accordingly.

So here is my test. I used evaluative metering (whole scene), and in Av mode, made a note of where it thought exposure should have been, then I went to manual and raised the shutter 5 stops. Then I used DPP to bring up the exposure 2 and the curve box the rest of the amount. Of course after I do this, I noticed your 2nd example was at 1/200th, and I somehow managed to mix your two examples. I took your first example but went 5 stops instead of 4.

So how it looked, how it looked after pulling it up, and a crop with just color noise removed. Its pretty difficult to mirror the same exact situation but this seems close enough. So now we need somebody to do the same with a 7D, since this topic is really about the 7D.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

The crop... (of course in this situation, I would have just shot this around ISO 1600), the noise level would be much better. I would never, ever shoot 5 stops below what would be considered a decent exposure. Plus it accentuates the noise.

HOSTED PHOTO
please log in to view hosted photos in full size.

I have LR3, I haven't taken the time to really learn it, and from most accounts, DPP still works better overall, since Canon wrote and maintains it to match their raw output on their latest offerings.

EDIT: Stop deleting posts!

Sorry Lol,

I think there is some inevitable variation in exposure, and subjectively I'd say you example was closer to my previous example of about -4 Stops, would you agree with that?

Crop looks pretty clean imo, although cross-hatching is more visible on the non-crop, although I'd say you increased the exposure about half a stop more than you needed to.

Below I arranged the images closer together for easier comparison?

(-4 Stops version)

MIME changed to 'text/html' | Content warning: script

MIME changed to 'text/html' | Content warning: script

MIME changed to 'text/html' | Content warning: script

MIME changed to 'text/html' | Content warning: script

Perhaps, I don't know, I am not at the point where I can visually determine how close two items are with exposure. The 1D4 is pretty clean, but I wouldn't emulate ISO 3200 this way, I would just use ISO 3200 which is very nice on this camera. I very seldom have had the need to shoot to the left. Even when I was in Hawaii, shooting the surf and mountains, etc, I never had such a situation.

Here is the 30D pushed 5 stops, it's not even worth discussing.... Again ISO 3200 out of the 30D is a bit cleaner than this.

HOSTED PHOTO
please log in to view hosted photos in full size.

For shiggles, I tried replicating the test. Set the camera to -5 EC. f/2.0 50mm iso100. I found I had to throw quite a bit of light in order to get it to 1/200s shutter speed rhys216 had used.

But started attempting the recovery via LR3. Adjusted exposure by +2 brighten the background more than enough. Then tried to toy w/ the shadow/dark curves and that's when the banding started to rear its ugly head on the dark black body of the 85L (my lens subject).

Naturally I don't ever see myself needing to do so but I can also accept/understand the argument for "if you could, wouldn't it be nice?" with regards to recovering that many stops in the shadows.

I like what Sony's cooked up, and/or the Sony/Nikon combo (between hardware/software engineers). It does up the ante and guess we'll all just have to see if Canon's CMOS boys can cook up something decent as well.

hmmm.... nice thing you've done rhys216. i'll try it with the 7d after work and school today..

Tried it out with the 7D and 85 1.8. Av mode F2. ISO 100. Set ISO increments to 1/2 stop and dialed down 8 times on the shutter speed and ended up at 1/350.

Processed it LR3 and in DPP. I am rather shocked by the difference in the software. I know Team has mentioned it but I just didn't realize what a difference it makes. Both allowed me to increase exposure by 4 stops. In LR I bumped the fill light and the shadows and highlights curves. In DPP I used the slider for shadows to adjust it to level 2. Sharpening is off in both. In DPP, NR is ;chrominance(color) set to 7 out of a possible 20, Luminance is at 0 and in LR the color noise is maxed at 100 with detail at 0 and Luminance is set at 0. No output sharpening in LR as well. I have the latest ACR update I believe as well.

Original: