7D ISO exposure test....

[Duke]

ISO TESTS ON 7D

Herm on DVXuser posted a very smart exposure test that you should look at here, but the results are below. ISO changes do NOT cause a linear progression of noise! There are some definite sweet spots.

http://www.dvxuser.com/V6/showthread.php?t=185471&highlight=noise


ISO/Noise tests. 1080/24p, lens cap on. In FCP, native format, maxed the mids, applied same filter to every clip.

Everything above 800 is noisy.
800 So, so. Use only if you have to.
640 Okay for a high value, better than 500

Virtually noise free:
320 Nearly as good as 200
200 Best
160 Best

100 Okay but 160 is better. Even 320 is better.

Do not use:
500
400
250
125

[Bif]

These tests are great in one way, you can really see how noise free the lower ISO's are.

However...

I have yet to shoot a project with the lens cap on.

My 7D is due to be delivered tomorrow so my comment below is based on what I experienced with the T1i. I have shot tests at night at ISO 1600 and yes, got a bit of noise, but the motion kind of blended it out to where you had to look for it to see it.

The segment that follows was shot out in a parking lot mainly to test some "concepts", some at ISO 400 most at higher. I don't see noise as a real problem with the 7D.

T1i test: YouTube - T1i Night Test Short Ver

[Duke]

Which is exactly why that test was so brilliant as a test. When shooting regular footage the images hide the noise. With the ISO test with the lens cap on, the only thing you can see is the noise. I wouldn't have thought of that, but I wish I had.

It makes sense though. I had noticed that some 7D footage looked noisier than 5D footage, and vice versa. Now we know why.

The most interesting thing is it isn't just the low ISO levels that are better. Certain ones are far better than other.

I'm waiting for my delivery too. :hv20-smilie77:

[SanVa]

Which is exactly why that test was so brilliant as a test. When shooting regular footage the images hide the noise. With the ISO test with the lens cap on, the only thing you can see is the noise. I wouldn't have thought of that, but I wish I had.

It makes sense though. I had noticed that some 7D footage looked noisier than 5D footage, and vice versa. Now we know why.

The most interesting thing is it isn't just the low ISO levels that are better. Certain ones are far better than other.

I'm waiting for my delivery too. :hv20-smilie77:

The test is brilliant and it provides valuable information, but I would still ask myself if I am really interested in noise that cannot be detected when shooting without the cap on the lens. Probably it does not answer the real question of which ISO can I really use.
Moreover, talking about film look, I remember when I used to put actual film in the camera and ISO 400 was at least as noisy as ISO 1600 in this 7D test. Probably I can live with that if necessary since we all already did.

[1]

Thanks for posting the tests, Duke.

I'll also follow with a HOWEVER:
a super clean image is not really what I'd call film-look, so I'd venture to guess that most will be putting back grain and noise either with Magic Bullet or some other software.
What I mean is: the strive for super clean images is only really occasionally important.
All these new DSLRs are WAY less noisy than what is desired for a "film look".

[zenio]

Bif your blacks in t1i video are very crushed ,you should update your codec it helps a lot.Especially if youre gonna get 7D ,you will get the same problem cause its the same codec.

http://prolost.com/blog/2009/1/19/5d-crushing-news.html

But if you crushed them yourself then OK,but i wouldnt do that on night shots.
It looks like footage before i installed 7.6 version of quicktime.Version X have the same problem with crushed blacks.

[Bif]

Zenio,

Thanks for your comments and observations. No attempt was made to crush blacks. Two things were being tested:

One - a bit of a "film noir" look so the exposure in the parking lot was deliberately reduced. The lens was a 50mm F1.8 Nikkor and at wide open aperture there was too much facial detail and clothing detail so the lens was stopped down to F2.8. This may or may not be the actual look used if we develop any story idea from this.

Two - what might be possible to use for some "dramatic" street sequences at night. Again with the lens wide open store signs, street lights, oncoming headlights, and even tail lights all "bloomed" too much for the effect being tested for. Again stopping down to F2.8 gave the "look" I wanted to "play" with.

For a much more "light hearted" look where some street surface detail, building detail, and vehicle detail would be in order I would opt for more exposure and even let light sources "bloom".

Again, I appreciate your comments and ideas.

[zenio]

Difference is visible when you watch low light video on canon and then watch the same video on your PC,on canon you see more details in shadows and on PC many things are covered with blacks.

[Ian-T]

I'll also follow with a HOWEVER:
a super clean image is not really what I'd call film-look, so I'd venture to guess that most will be putting back grain and noise either with Magic Bullet or some other software.
What I mean is: the strive for super clean images is only really occasionally important.
All these new DSLRs are WAY less noisy than what is desired for a "film look".Exactly.

But at least now we know what to expect when we are going for that film look. The noise in this camera is so low that it can give that subtle film grain look without being so blantantly obvious.

I'm still several weeks away from getting mine.

Thanks for posting those Duke.

[Kin Lau]

BTW, I ran the same tests on my 7D, but I did not process the footage in any way (didn't max the mids), and my footage is just plain black, not the snow storm you see in the above footage. That's also considering I ran the ISO all the way up to 12800.

[Bif]

Difference is visible when you watch low light video on canon and then watch the same video on your PC,on canon you see more details in shadows and on PC many things are covered with blacks.

I don't think you understand the goal and purpose of that test. One of the first things I do with a new cam these days is to try to relate what I see on the LCD to what the same imagery will likely look like on both PC (in the edit stage) and as a finished image on a reference TV.

I was after a "dark look" and got it. I also got sequences where the exposure showed more detail but "discarded" those in the edit bay as being "overlit" for what was being tested for. And the main purpose of posting that test here was to show no objectionable (or even really noticeable) noise at ISO's from 400 to 1600.

I can assure you that after a second career as a professional studio photographer and 13 more years as a civilian photographer for the air force that I am more than passing familiar with exposure and well accustomed to managing exposure for "specific look". I come from a generation of photographers who had to learn to do all this on film without the benefit of being able to "chimp" on a little LCD on the back of the camera.

[Bif]

I'm still several weeks away from getting mine.



Here's hoping several weeks turns out to be only a few. I could almost feel the way you "lusted" after the 5D MkII when it came out (dang, so did I!).

UPS tracking shows mine "Out For Delivery" this morn, the "brown truck" usually hits my neighborhood around 4:30 - 6:30PM.

Gonna be a long day...

[Bif]

BTW, I ran the same tests on my 7D, but I did not process the footage in any way (didn't max the mids), and my footage is just plain black, not the snow storm you see in the above footage. That's also considering I ran the ISO all the way up to 12800.

That is very encouraging.

[lordtangent]

All these new DSLRs are WAY less noisy than what is desired for a "film look".

Very good point Mal. the magnitude of film grain tends to be incredibly high compared to video noise. The only problem is that video noise looks nothing like film grain. video noise is totally concentrated in the dark areas where film grain is almost perfectly distributed along the entire exposure range.

So, if the desire is a "film look" type grain you still need to start with a pretty clean image and add film like grain to it.

Anyone who wants to see film grain from scans can check out the links in my sig. I'll be adding more as I acquire them.