7D noise

I've had 2x 7D bodies and whilst they were superb cameras, the LOWER ISO values, 100-400 definitely produced more visible noise than the 40D I had before it and the GF1, G1 and D3100 and 5D2 I have had since....

...But only if you sit in front of your monitor zoomed in to 100% :nono:

FFS get outside, take more pictures, enjoy your camera and PRINT some of your best shots to enjoy them on the wall! :thumbs:
 
jonneymendoza said:
Well my 400d diddnt come with DPP mate!!

It did.

Its boxed with ALL Canon dslr's.
 
Unless it's changed Adam I think you need to install from the disc and then run the updates from canon, although I believe there is a workaround to get the update to install on its own.
 
Davie said:
Unless it's changed Adam I think you need to install from the disc and then run the updates from canon, although I believe there is a workaround to get the update to install on its own.

No you can just download it. It has changed. Probably got lots of moaning of Canon users
 
I was also initially disappointed with noise apparent on my 7d images after I moved from my sony a700, especially at lower iso values such as 200 and 400, obviously moving from a 12mp sensor to a 18mp sensor would have some effect but still I was slightly disappointed.

However after some experimentation I also discovered that when shooting raw increasing the masking value of the sharpening section in ACR made a large difference (i typically increase the masking from zero to between 25 and 50 depending on the shot).

I am now very happy with the performance of the sensor overall:)
 
Sorry to drag this up again, but went out at the weekend and snapped a random shot of my friends daughter outdoors and it came out almost perfect straight out of the tin, ISO 100, F4, 1/800 using a 17-55mm.

So it goes to show that as people have said before if you get the shot exposed and focussed just right the noise isn't a problem and the images are great.

I think that its just a case of this camera being less forgiving if you don't get a good exposure.
 
I shoot with the 7D pretty well by default at 800 ISO with the 500mm (and also quite often still have to crop) with no noise issues, so I find the noise problems people have a bit bewildering. Certainly under-exposure will increase noise, but I think you're going to get the best noise performance with the 7D from Canon DPP.
 
Blasted said:
Sorry to drag this up again, but went out at the weekend and snapped a random shot of my friends daughter outdoors and it came out almost perfect straight out of the tin, ISO 100, F4, 1/800 using a 17-55mm.

So it goes to show that as people have said before if you get the shot exposed and focussed just right the noise isn't a problem and the images are great.

I think that its just a case of this camera being less forgiving if you don't get a good exposure.

At 100 iso I wouldn't expect any noise anyway?
 
I have to admit that at some points I've almost jumped backwards when a high ISO RAW's been loaded up in LR :lol:

Sometimes they can look REALLY bad to the point I've felt like hitting the delete button BUT with a little work you can still get a perfectly acceptable image, yes it may look a little iffy at 100%, yes some other photographers will tell you it's waaaaay too noisy and totally unacceptable but someone who's not so critical and has a lovely picture of their pet/kid etc will more than likely be completely chuffed with it.

I shot a rugby match in November last year on my 400D, only managed to take shots for the first 30 mins and ISO1600 wasn't cutting it any more. In Feb I was using my 7D after 30 mins I had to resort to ISO3200 and most of the second half was at ISO6400, they weren't great by any means but they were usable and definitely better than nothing ;)

Moving from a 400D to a 7D and it is a sharp learning curve. There's a lot more emphasis on getting the exposure right and setting the AF system to best meet your requirements. The 400D was quite forgiving in terms of PP, i.e I mainly shot in JPEG and didn't do any but to get the best out of the 7D you really need to shoot RAW and get PP'ing.
 
I don't understand why people keep talking about 'exposing to the right' when using high ISO. If it were that easy to expose to the right under dark circumstances, why not just lower the ISO instead, thereby getting less noise in the photo? Exposing to the right is fine in a normal situation when there is enough light to do so. Correct me if I'm wrong.

You first pick your aperture (creative choice for DOF control) and your shutter speed (creative choice for motion/shake control) and that determines how much light the sensor will receive.

Then you raise the ISO until you get an image that appears well exposed. If 100 ISO would leave you 3 stops underexposed then you would be better off to shoot at 800 ISO and get a "proper" exposure as a result. If you have no highlights in the scene and plenty of unused highlight headroom then you might well choose to raise the ISO another stop, to 1600, and brighten the image further. In theory that should lower the sensor read noise and give you an image that will be a little cleaner still when it comes to editing.

It's not shooting at high ISO that causes noise. It's failing to capture enough light that is the problem, and then having to raise the ISO to try to compensate. High ISO is your friend. It bails you out when the light is poor. If there was no value in it we might as well go around shooting at 100 ISO all day (and night) long, leaving masses of highlight headroom, and then simply bringing everything up in post. I haven't heard that technique recommended for a while.

So, "Why ETTR when shooting at high ISO?" Well, why not? It's better than underexposing the shot.

That said, there comes a point of diminishing returns when reaching the upper ISO limits. With the 7D I have read well reasoned arguments that there is no benefit in going beyond 3200 ISO, and above that point you are no worse off by leaving a shot underexposed and then working it in post. One advantage of that approach is that you do buy yourself extra highlight headroom and supposedly the resulting IQ is no worse. The thresholds will vary from camera to camera, but it's 3200 ISO for the 7D. Of course, if you're shooting to JPEG and need results ready to use SOOC then you'll just have to pick whatever ISO you need to get the results looking right. If you shoot raw then you have options to consider.
 
So, "Why ETTR when shooting at high ISO?" Well, why not? It's better than underexposing the shot.
Whilst I agree with what you've said, what I don't understand is why using a higher ISO gives you anything more than post processing a lower ISO.

A given shutter speed and aperture will give you the same photons on the sensor no matter what the ISO. The "exposure" is determined solely by the aperture and shutter. What the ISO setting does is amplify the signal before it gets digitised. Whether you ETTR or not, the fundamental data (i.e. photons on the chip) are the same. As I see it, all ISO is doing is some of the pre-processing for you.

Or am I missing something here?
 
The part you're missing is what happens within the camera between the analogue and digital processing elements, and for reasons which I don't understand, raising the ISO (analogue gain) within the camera reduces the read noise of the electronics. This means you get a better SNR out of the camera by raising the ISO, while leaving the EXPOSURE (shutter speed and aperture) unaltered than by leaving the ISO low and boosting everything digitally in post.

EDIT....

Consider an analogue audio recording made onto good old fashioned tape with a microphone. You would have a "recording level" or gain control which you can adjust to optimise the recording level, given the volume of the sound reaching the microphone. It would be prudent to turn up the gain in order to saturate the tape for the peaks of the sound, rather than leaving the sounds buried in the hiss of the tape itself. Turning up the recording level is like turning up the ISO within a camera. Once you have the best capture of the signal at the analogue stage you can then set about digitizing it. If you'd left the analogue signal at a low level prior to digital conversion then your digital signal would faithfully immortalise the noise too.
 
Last edited:
Whilst I agree with what you've said, what I don't understand is why using a higher ISO gives you anything more than post processing a lower ISO.

A given shutter speed and aperture will give you the same photons on the sensor no matter what the ISO. The "exposure" is determined solely by the aperture and shutter. What the ISO setting does is amplify the signal before it gets digitised. Whether you ETTR or not, the fundamental data (i.e. photons on the chip) are the same. As I see it, all ISO is doing is some of the pre-processing for you.

Or am I missing something here?

Briefly ....

Camera counts light photons in a linear fashion, take a typical camera with a 6 stop dynamic range recording a 12bit image, capable of recording 4096 tonal values (probably greater with newer technology but will do for this example)

Each of the 6 stops does not record the same number of tonal values, the brightest records half of the total, the next stop half of that and so on.

So from brightest to darkest would record the following tonal levels 2048, 1024, 512, 256, 128, 64.

Half of the information is in the brightest stop.

So if you under expose to prevent blown highlights you are wasting half of the levels available, then when you try to boost the darkest shadow areas you only have 64 levels to spread over a wider tonal range. This is what exaggerates the noise as there are fewer tonal values available to give a smooth transition from one to another.

Expose to the right and take care to avoid clipping ... simples :D

There is plenty of info on this on the web, that will include drawings graphs and other aids to understanding the principal, but the above is a summary of why it is best to expose to the right. :thumbs:
 
Interesting posts Tim and Martyn :) Both of which make sense, but am I seeing a bit of a contradiction.

Tim makes the point about raising ISO because that is the only way that we can benefit from the anlague-to-digital stage of the process, which happens in the production of the Raw file before it is converted to digital.

But Martyn's explanation skips that bit and only mentions over-exposing to increase the tonal capture, which can be done by adjusting either ISO, shutter speed or aperture.

I would put a further spin on that, which is that as I understand it the key driver of ETTR is photon capture, ie light, the basic raw material. And that is at odds with Tim's method.

Question: what would deliver the best result here? Assuming that the 'correct' exposure to put mid-grey in the middle of the histogram, is 1/250sec at f/8 with ISO100, and also that the subject is benign and there's two clear stops of headroom above the brightest highlight.

Shoot at 1/250sec at f/8 and raise to ISO400, or drop the shutter speed to 1/60sec at f/8 and stick with ISO100? For former allows the ADC to work it's magic, and doesn't compromise shutter speed, but the latter collects four times as much light to work with. The histogram in both instances should be identical.
 
There is no contradiction. Of course it is best to capture as many photons as you can (without unwanted saturation), so if you can fill the sensor at 100 ISO then that's the best way to go. But if shooting at 100 ISO leaves you 2 stops underexposed, and you have no wiggle room on aperture or shutter speed then it is better to raise the ISO by 2 stops than to leave the shot underexposed.

That is not the same thing as raising the ISO by 2 stops and then dropping the light collected by 2 stops by raising the shutter speed.

Two basic rules....

1. Capture as much light as you can by choice of aperture and shutter speed;
2. For that combination of aperture and shutter speed, raise the ISO as high as you can without unwanted highlight clipping.

That's how you ETTR.

Note, for point 2, that advice does not include using fake ISOs. Only use real ISOs that are effected through analogue gain rather than digital pushing and pulling.
 
Each of the 6 stops does not record the same number of tonal values, the brightest records half of the total, the next stop half of that and so on.

So from brightest to darkest would record the following tonal levels 2048, 1024, 512, 256, 128, 64.
Doh... Of course. If I'd thought about it a bit more, I'd have realised this,.

Thanks
 
Doh... Of course. If I'd thought about it a bit more, I'd have realised this,.

Thanks

You are welcome :thumbs:

Best not to over think these things, if it works use it and concentrate on the subject and composition.
 
There is no contradiction. Of course it is best to capture as many photons as you can (without unwanted saturation), so if you can fill the sensor at 100 ISO then that's the best way to go. But if shooting at 100 ISO leaves you 2 stops underexposed, and you have no wiggle room on aperture or shutter speed then it is better to raise the ISO by 2 stops than to leave the shot underexposed.

That is not the same thing as raising the ISO by 2 stops and then dropping the light collected by 2 stops by raising the shutter speed.

Two basic rules....

1. Capture as much light as you can by choice of aperture and shutter speed;
2. For that combination of aperture and shutter speed, raise the ISO as high as you can without unwanted highlight clipping.

That's how you ETTR.

Note, for point 2, that advice does not include using fake ISOs. Only use real ISOs that are effected through analogue gain rather than digital pushing and pulling.

Thanks Tim. It was just the way I read it :thumbs:
 
Back
Top