High ISO vs Underexposing

[mitchbooth]

Hello everyone,

Just returned from the Indoor State Track meet where I took about 1500 photos. As with most indoor venues the lighting was sub par. I met several photographers from MileSplit Ohio who ranged from shooting in sports mode to underexposing every photo by 3 or 4 stops in Raw only to bring up in post production. I was typically at 640 shutter speed and around 3.2-4 on aperature and my iso was set on auto ranging from 2000-10,000. I also used all large Raw files with no JPEG at all.

I'm new in photography and have the Canon 70d, Tamron 24-70 F2.8 and Tamron 70-200 F2.8 as my first ever camera. I've finally gotten out of auto modes and relatively comfortable with the camera and operating in manual mode. I use Adobe Lightroom for organizing and editing my photos.

I've been trying some night photography and typically leave Long Exposure Noise Reduction off. I have been leaving High ISO Noise Reduction on but have heard by doing so you sacrifice some definition in the photos...?

My question to you more experienced photographers is this. Should I shoot indoor events underexposed by 3 or 4 stops and then bring them up in Lightroom or should I use higher ISO for the Canon 70D and let it properly expose them in camera. Which way gets the better result? Is it poor practice to leave the High ISO Noise Reduction off in camera?



Mitch

 

[Overread]

Higher ISOs result in higher levels of noise for the same level of exposure in the overall scene.

However underexposing and then raising the brightness in editing will result in more noise than if you'd exposed correctly and used a higher ISO.

Thus its always preferable to expose correctly in camera, even if that means using a much higher ISO. You basically want the most light-data that you can get to ensure the best possible result. The "expose to the right" theory is one you should look up and make use of, where practical. Indoors it won't always be possible where you've got a limit on your shutter speed (to avoid blur) and a limit on your aperture (it can only go so wide) - so sometimes you will be underexposing even at your cameras top ISO value.



Also note that Sony and Nikon have more recently pushed what can be recovered from underexposure. Thus you're more likely to run into shooters of both with new end cameras who can be a bit more lenient on underexposing and still getting back good quality results after brightening. This is a more recent change and reflects advancing in sensor technology. Canon are still very good, just not quite the same; plus the expose to the right theory still stands - its just that the limits are a bit wider now than in the past.

 

[runnah]

3 to 4 stops seems a bit extreme.

Also noise is not a big deal. Unless it looks like static on an old TV most people won't notice unless they are pixel peeping internet photo warriors. 90% of people consume photos on their phones, tablet or small laptop so you can get away with a number of small technical issues.

Worry about composition and light before things like noise.

 

[mitchbooth]

3 to 4 stops seems a bit extreme.

Also noise is not a big deal. Unless it looks like static on an old TV most people won't notice unless they are pixel peeping internet photo warriors. 90% of people consume photos on their phones, tablet or small laptop so you can get away with a number of small technical issues.

Worry about composition and light before things like noise.

Yes, I agree. 99.9 percent of the photos are being viewed on smartphones or tablets. I was just trying to figure out what process produces less noise. I hear one person say it produces more and the other says post production produces less and I have no idea who is more correct. I'm just trying to learn at this stage of the game.

 

[runnah]

3 to 4 stops seems a bit extreme.

Also noise is not a big deal. Unless it looks like static on an old TV most people won't notice unless they are pixel peeping internet photo warriors. 90% of people consume photos on their phones, tablet or small laptop so you can get away with a number of small technical issues.

Worry about composition and light before things like noise.

Yes, I agree. 99.9 percent of the photos are being viewed on smartphones or tablets. I was just trying to figure out what process produces less noise. I hear one person say it produces more and the other says post production produces less and I have no idea who is more correct. I'm just trying to learn at this stage of the game.

A lot depends on the sensor. My canon I can pull the shadow a few spots before the noise starts to show up. With my Sony FS5 I actually overexpose by a stop and then darken to reduce noisy shadows. Best thing would be to practice in non critical situations and learn what your particular camera can do.

 

[Pedro_lopez]

Just to kind of reinstate the op's original question since I too am wondering. When using high ISO is it better to leave noise reduction off?


Sent from my iPhone using ThePhotoForum.com mobile app

 

[KmH]

That depends on if you can wait for the noise reduction to be applied in the camera.

See page 226 of your D300 Reference Manual.

 

[Derrel]

If you would like more scientific info on underexposing and then compensating in post software, look on the web for articles on the topic of "ISO invariance".

 

[Overread]

Just to kind of reinstate the op's original question since I too am wondering. When using high ISO is it better to leave noise reduction off?

That depends:
1) Most DSLRs have a menu function called "long exposure noise reduction". This form of ISO noise reduction is based on noise generated by sensor HEAT rather than simply a high ISO setting in the camera (or underexposure followed by brightening). It's typically used in astrophotography where exposures can last many minutes a time and the sensor heats up a lot.
It works by taking two exposures, one after the other. The first is normal, the second the shutter blades don't open so the only data that should be recorded is noise spots from the heat of the sensor. The camera then uses code to deduct the second from the first and outputs the final adjusted photo. As a result each shot takes twice as long (at least) because it has to take two for the same duration.
Also note that normal photography doesn't induce enough heat to make this mode worthwhile to use.

2) In camera noise reduction settings apply to the JPEG only. So if you're using in-camera generated JPEGs the noise settings will affect how harshly (or not) the camera applies noise reduction. Note that this is always global; it will apply to the whole photo equally. In editing in the computer you can specify where noise reduction takes place; applying more to low detail areas (eg backgrounds) that show noise more readily; whilst applying less to high detail areas (main in focus segments) where noise doesn't show as readily; and where you want to retail crispness of detail.

3) RAWs won't have any in-camera editing noise reduction applied, however some RAW processing software might use in camera settings to set a default starting position when you start editing.


NOTE - In this I've said that the camera applies no noise reduction to RAWs - this is a lie.
The camera does have built in noise removing features, however they are universally applied to every RAW and subsequent JPEG regardless of user settings. As a result its simpler to ignore as you can't get any other result from the camera.

 

[Ysarex]

My question to you more experienced photographers is this. Should I shoot indoor events underexposed by 3 or 4 stops and then bring them up in Lightroom or should I use higher ISO for the Canon 70D and let it properly expose them in camera. Which way gets the better result? Is it poor practice to leave the High ISO Noise Reduction off in camera?

Mitch

You got some pretty good info already from Overread. And as Derrel noted you may want to look into the topic ISO invariance. Your camera does not have an ISO invariant sensor and so there's no question about what you should do; use the camera ISO settings to get the best result. Raising ISO reduces noise.

A few more comments about your question as you mentioned you save raw files. When you raise the ISO on your camera two things occur: 1) The meter system in your camera is adjusted to calculate a reduced exposure. If you rely on the meter in the camera, raising ISO causes the camera meter to calculate an underexposure for the sensor. When you raise the ISO the camera does not properly expose your shots in camera as you suggested. The sensor is underexposed exactly as if you simply forced an underexposure at base ISO. You force a 3 stop underexposure the sensor is underexposed by 3 stops. You raise the ISO 3 stops and the meter calculates a 3 stop exposure reduction and the sensor is underexposed by 3 stops -- no difference at that point.

The second thing ISO does 2) is a post processing job. It occurs after the underexposure. The signal coming off your sensor is analog and is processed in the camera through circuitry we call the ADC (analog to digital converter). The ADC turns analog electrical signal into the numbers that are stored as you raw file. Ultimately this is an engineering issue and is addressed by different camera manufacturers in different ways. What they have to do is compensate for the underexposure and what most cameras do now is apply a signal boost to the sensor output as part of the ADC processing. That analog signal boost has as an advantage an improved noise profile: ISO reduces noise. What's been happening behind the scenes in recent years is that the sensors are becoming so good with output so clean that the manufacturers have started to skip the analog signal boost -- no benefit -- and replace it with digital scaling (just upscale the digital content) which is less expensive (always cheaper to do it in software). These are ISO invariant cameras. There are also hybrid implementations now where both a partial analog signal boost is combined with digital scaling.

Because what ISO does other than bias your camera meter is a post exposure process, that of course means it could be saved for later if it's digital scaling. And so the underexposed raw file can be digitally scaled in the raw converter. To do that you want a camera with an ISO invariant sensor, you don't have one so this is just informational.

From our standpoint as users this starts getting pretty esoteric -- raise the ISO on your camera as needed and get on with your life. My previous camera (Fuji X-E2) was ISO invariant. It took me awhile after some testing to really believe it, but eventually I just started to ignore the camera ISO setting entirely. The last year and a half that I owned the camera I never touched the ISO dial. It was nice to be able to ignore it, but no big deal. When I got my new Fuji X-T2 I tested it and it certainly is not ISO invariant. It's a new sensor and there's no way I can treat it like my old X-E2 so I'm back to changing ISO -- no problem.

There was one advantage that the ISO invariant sensor gave me. When the sensor capture is analog boosted or scaled up all the data is equally raised in value and so it's possible to clip highlights in the process. Something of a paradox but you can set a camera to ISO 1600 and underexpose the sensor by 4 stops and still get clipped highlights because of the signal processing through the ADC. Well, if you skip the signal boost/scaling and leave the raw file as-is underexposed you'll typically retain all highlight detail. Here's an example:

I got roped into bringing my camera along to a recent event and grabbed some snaps. This was the X-E2 and I just left the ISO at base and underexposed as needed. Given the indoor lighting I was underexposing 2 and 3 stops. It was an overcast day but nonetheless the light outside (through window) was much brighter than indoors. Had I raised the ISO much of the data out the window would have been blown out. It was easy enough processing to mask the windows.

Putting that in very technical terms when you raise the ISO on your camera the post processing that takes place reduces the dynamic range of your sensor. With an ISO invariant sensor you avoid that DR reduction.

Joe

 

[Soup du Jour]

I was just trying to figure out what process produces less noise.

I think neither. The gain (amp gain) penalty comes in the end with the image in both cases. With one method you have to remove noise relative to an already sufficient gain level, while with the other you need to first make the image brighter with software (increased gain on an insufficient signal) and then remove noise. If a question is "which method gives more signal information" I think the high ISO method, but with the higher noise.

If you try to lift an extreme underexposure too far, there is no data there to do it. More data is there in the high ISO capture but with higher noise.

 

[KmH]

NOTE - In this I've said that the camera applies no noise reduction to RAWs - this is a lie.
The camera does have built in noise removing features, however they are universally applied to every RAW and subsequent JPEG regardless of user settings. As a result its simpler to ignore as you can't get any other result from the camera.

The camera does not apply any noise reduction to a Raw file in the camera - unless the Raw file is to be further processed as a JPEG in the camera.

Raw converters external to the camera, like ACR/Lightroom Develop module, perform some combination of tone mapping, edge detection, anti-aliasing, noise reduction, and sharpening to a Raw file as part of the demosaicing process that makes the Raw file a photo that looks like the real world.

 

[Ysarex]

I was just trying to figure out what process produces less noise.

I think neither. The gain (amp gain) penalty comes in the end with the image in both cases. With one method you have to remove noise relative to an already sufficient gain level, while with the other you need to first make the image brighter with software (increased gain on an insufficient signal) and then remove noise. If a question is "which method gives more signal information" I think the high ISO method, but with the higher noise.

With an ISO invariant sensor, noise is the same with or without the ISO analog signal gain. With a non ISO invariant sensor noise is much worse without the analog signal boost compared with the result obtained from the gain applied. Raising ISO with a non ISO invariant sensor reduces noise.

If you try to lift an extreme underexposure too far, there is no data there to do it. More data is there in the high ISO capture but with higher noise.

The amount of data recorded is a function of exposure and only exposure. Raising the ISO does nothing to increase the amount of data captured. With a non ISO invariant sensor, the increased noise resulting from failure to apply the analog signal boost will result in the weakest data masked by the increased noise. Raising the ISO in this case reduces noise and prevents slight loss of the weakest data.

Joe

 

[petrochemist]

NOTE - In this I've said that the camera applies no noise reduction to RAWs - this is a lie.
The camera does have built in noise removing features, however they are universally applied to every RAW and subsequent JPEG regardless of user settings. As a result its simpler to ignore as you can't get any other result from the camera.

The camera does not apply any noise reduction to a Raw file in the camera - unless the Raw file is to be further processed as a JPEG in the camera.

Raw converters external to the camera, like ACR/Lightroom Develop module, perform some combination of tone mapping, edge detection, anti-aliasing, noise reduction, and sharpening to a Raw file as part of the demosaicing process that makes the Raw file a photo that looks like the real world.

Sorry Keith but Overread is quite correct. At least some cameras do apply noise reduction to the signal before storing the RAW file. It caused quite a shock when it was discovered Nikon did this, as RAW was always considered to be unprocessed data. I suspect Nikon are not the only ones who do it. The amount of processing of RAW data is minimal so more will be needed but some is done.

 

[DarkShadow]

Yep Some camera's do apply noise reduction of the raw file for example the Nikon D3300 but supposedly its less then it would be if turned on.