Iso

[Shadowman]

Spurred by this, I just did some controlled tests with my Canon 5DIII, shooting the exact same scene with ISOs from 100 to 6400. One thing that is relevant is that the lighting was ideal, extending almost the full range, and with the majority of mass in the top half of the histogram:



I deliberately chose a scene with a flat surface, to show noise, and some detail, to show the impact of noise on sharpness. For the latter, look at the lettering on the red and white White Mountains guide, which was the point of focus.

I shot these at f/7.1 using a macro lens. I did no processing at all other than reading them into lightroom and allowing lightroom to apply its default capture sharpening. I exported the jpegs at 92% quality with output sharpening turned off.

ISO 100:



ISO 800:



ISO 6400:

I did a similar test and the only noise that was visible was what presented itself on a darkened monitor. With in-camera noise reduction the noise was less noticeable. I usually turn off the NR on the camera, I may have to try it on my next action sequence shot.

[xpatUSA]

Hi Ted,

Understood about reason for posting, just wanted to get an idea of how much light was reaching the sensor.

Here's the raw histogram from RawDigger:



Scales are log; exposure noted in the header differs from EXIF - don't know why.

EV0 is set at 4095 which is just above 'sensor well full" in raw data for this camera.

The sensor exposure was very roughly 0.03 lux-sec for the green channel.

Hope it tells you all you need to know . . please ask if not.

[Manfred M]

John - your image definitely has a significant magenta cast and is a bit soft. You have shot quite wide open, but a bit of sharpening cleans up the skater in front of the others. Colour correction is really easy on this one; just use the colour correction eyedropper on the ice and all the other colour problems go away.





When it comes to noise, I don't find that that is an issue with this image.

[rtbaum]

Spurred by this, I just did some controlled tests with my Canon 5DIII (a fairly low-noise camera), shooting the exact same scene with ISOs from 100 to 6400. The results surprised me. That's one nice thing about empirical data.

I suggest looking at all of these in the lytebox.

One thing that is relevant is that the lighting was ideal, extending almost the full range, and with the majority of mass in the top half of the histogram:



I deliberately chose a scene with a flat surface, to show noise, and some detail, to show the impact of noise on sharpness. For the latter, look at the lettering on the red and white White Mountains guide, which was the point of focus.

I shot these at f/7.1 using a macro lens. I did no processing at all other than reading them into lightroom and allowing lightroom to apply its default capture sharpening and color noise reduction. I exported the jpegs at 92% quality with output sharpening turned off.

ISO 100:



ISO 800:



ISO 6400:



At 100%, the ISO 6400 image shows degradation, although less than I thought. You can see the full-size jpeg here:

https://dkoretz.smugmug.com/photos/i.../i-2V6mbxq.jpg

Some observations:

-- As expected, at this level of downsampling, they all look fine. The effects, when present, will be more apparent at larger sizes.

-- At 100%, the degradation of the image is considerably less than I had expected given past experience. For example, here is a 100% crop from an ISO 4000 image, with the ISO selected by the camera, after processing but without noise reduction other than LR's default:



This one looks fine when downsampled to a small online display, but the image degradation is substantial.

I think the reason is apparent from the histogram, which is very different from the first:



So, my take-aways:

1. Don't use small online displays to evaluate image quality, unless that is the only way you are going to present the image. They will obscure problems that may be apparent at larger sizes.

2. Modern lower-noise cameras can handle fairly high ISOs reasonable well, if the lighting is pretty good, meaning that the distribution of luminance has its mass fairly far toward the bright end.

3. If you have to use high ISOs with images that have a lot of the mass at the darker end, the results will not be as good.

Seem like reasonable conclusions?

Dan

pretty much what I have observed

[Shadowman]

John - your image definitely has a significant magenta cast and is a bit soft. You have shot quite wide open, but a bit of sharpening cleans up the skater in front of the others. Colour correction is really easy on this one; just use the colour correction eyedropper on the ice and all the other colour problems go away.



When it comes to noise, I don't find that that is an issue with this image.

Thanks Manfred, I'll try that on a few of my other images.

[Shadowman]

Here's the raw histogram from RawDigger:



Scales are log; exposure noted in the header differs from EXIF - don't know why.

EV0 is set at 4095 which is just above 'sensor well full" in raw data for this camera.

The sensor exposure was very roughly 0.03 lux-sec for the green channel.

Hope it tells you all you need to know . . please ask if not.

Did you use RawDigger's Dynamic Range Evaluation tool or the In Focus Areas and Noise in Shadow tool to evaluate your camera's limitations?

[xpatUSA]

Did you use RawDigger's Dynamic Range Evaluation tool or the In Focus Areas and Noise in Shadow tool to evaluate your camera's limitations?

Haven't seen those tools in my Research Edition's menus, John. Perhaps they're in the Pro Edition?

[Manfred M]

One thing that is relevant is that the lighting was ideal, extending almost the full range, and with the majority of mass in the top half of the histogram:
So, my take-aways:

1. Don't use small online displays to evaluate image quality, unless that is the only way you are going to present the image. They will obscure problems that may be apparent at larger sizes.

2. Modern lower-noise cameras can handle fairly high ISOs reasonable well, if the lighting is pretty good, meaning that the distribution of luminance has its mass fairly far toward the bright end.

3. If you have to use high ISOs with images that have a lot of the mass at the darker end, the results will not be as good.

Seem like reasonable conclusions?

Dan

Dan - I will have to agree with you and your conclusions. Modern cameras, especially higher end ones do a fairly good job with digital noise so long as the image is properly exposed. This is not always the case when one gets into mid-range cameras, especially those with a fairly small sensor pitch.

The place where noise becomes more of an issue is when the image is underexposed.

On the other hand, the images that you posted are not going to be particularly sensitive to noise. Digital noise tends to occur in dark areas of the image; shadows and night skies, rather than in the lighter areas of the image. There is less data in these dark areas (remember 100% black has a value of "0"), so having digital error (noise) is more noticeable in those parts of the image, even though the actual noise will be distributed throughout the shot. I have been told by some that the blue channel sees to be more problematic when it comes to noise.

All that being said, the dynamic range and colour depth issues are still there, even if the noise performance has been improved. I'm particularly interested in Robin's posting in the other thread that shows that Fujifilm has started to address the dynamic range with some of their newest sensors.

[Loose Canon]

Spurred by this, I just did some controlled tests with my Canon 5DIII (a fairly low-noise camera), shooting the exact same scene with ISOs from 100 to 6400. The results surprised me. That's one nice thing about empirical data.

I suggest looking at all of these in the lytebox.

One thing that is relevant is that the lighting was ideal, extending almost the full range, and with the majority of mass in the top half of the histogram:



I deliberately chose a scene with a flat surface, to show noise, and some detail, to show the impact of noise on sharpness. For the latter, look at the lettering on the red and white White Mountains guide, which was the point of focus.

I shot these at f/7.1 using a macro lens. I did no processing at all other than reading them into lightroom and allowing lightroom to apply its default capture sharpening and color noise reduction. I exported the jpegs at 92% quality with output sharpening turned off.

ISO 100:



ISO 800:



ISO 6400:



At 100%, the ISO 6400 image shows degradation, although less than I thought. You can see the full-size jpeg here:

https://dkoretz.smugmug.com/photos/i.../i-2V6mbxq.jpg

Some observations:

-- As expected, at this level of downsampling, they all look fine. The effects, when present, will be more apparent at larger sizes.

-- At 100%, the degradation of the image is considerably less than I had expected given past experience. For example, here is a 100% crop from an ISO 4000 image, with the ISO selected by the camera, after processing but without noise reduction other than LR's default:



This one looks fine when downsampled to a small online display, but the image degradation is substantial.

I think the reason is apparent from the histogram, which is very different from the first:



So, my take-aways:

1. Don't use small online displays to evaluate image quality, unless that is the only way you are going to present the image. They will obscure problems that may be apparent at larger sizes.

2. Modern lower-noise cameras can handle fairly high ISOs reasonable well, if the lighting is pretty good, meaning that the distribution of luminance has its mass fairly far toward the bright end.

3. If you have to use high ISOs with images that have a lot of the mass at the darker end, the results will not be as good.

Seem like reasonable conclusions?

Dan

Interesting Dan and about what I have found running my own experiments with a 5DIII.

But here is the thing. Its good to run these kinds of tests to see where your camera is and what to expect at elevated ISO values. But one thing one has to keep in mind that it’s not the same thing sitting in your office shooting a bookcase under ideal lighting as it is in the field.

There are many reasons for shooting at high ISO values. Low light is one of them. Also the low light scenario is where the most prevalent noise issues raise their head.

I contributed to the other thread just answering the OP’s posted curiosity. Interestingly, the OP there seemed to abandon the thread! And I’m seeing some great contributions on John’s here. And I know you guys are well aware of the thoughts I’m about to raise.

While it’s needed to see what a particular camera can do at high ISO under ideal circumstances, and charts, graphs, data, etc. The most important element is missing. I have seen not one post addressing the techniques that can make or break success shooting at high ISO. This absolutely has to be coupled with the camera sensor’s abilities regardless of what that might be. I consider this paramount and if you can get a grip on this it will give you a way better chance at getting a usable, or even great shot at high ISO.

You, of course, nailed it Dan in your second example with the histogram. It is underexposed. By a minimum of 1.5 stops and probably more considering the bumped ISO and to mitigate noise. This is way more underexposed than acceptable at ISO 4000. Just checking the shot in ACR and adding my own experience with this I would say maybe another half stop exposure on top of the 1.5 stops and maybe a tad more. Cardinal Sin with high ISO!

[Loose Canon]

Oops!

Sorry Manfred!

I was writing my contribution but there was a lot of goldbricking as well as skylarking distractions going on while I did!

Hey! It’s the weekend!

Sorry I didn’t see your post before I tapped go!

[rtbaum]

Looks good, how many light sources?

about a 60 watt bulb in the hanging and heavily shaded lamp, just bright enough to not trip over my shoes

[William W]

. . . I've posted a recent capture done at slightly high ISO.
f/2, ISO 3200, 1/100sec handheld
I'm not particularly happy with the quality, this is an experiment in capture and post-processing. . . Not happy with the degree of sharpness or the amount of visible noise.

The image is too low res to adequately interrogate.

However, some considerations:

1. You might not be happy with the sharpness, if you thought that you focussed on the Subject centre Frame. That Subject appears to be moving camera right to camera left. If at walking pace, then, 1/100th second is likely not a fast enough shutter speed to arrest SUBJECT MOVEMENT at that shooting distance. It is likely that you a very minor Subject Motion Blur, which is often misinterpreted for a Focusing error or too shallow Depth of Field.

2. I assume you used your Nikon 85/1.8 on your D750. 1/100th second safe for hand held, but not always 100% safe in all situations. You could have Camera Blur.

3. I assume you used Matrix Metering and Manual Mode. There is a lot of white (ice) in the frame my interrogation of the EXIF is not quite clear, but it seems to indicate that you pulled to shot 1 STOP OVER exposed to what the camera’s TTL meter indicated. If so then I assume that was your manual compensation for an assumed false meter reading. If you did do this, then perhaps it was not enough, perhaps 1½ or 1⅔ or even 2 Stops over the TTL Meter’s reading might have been a more accurate exposure.

The understanding of NOT under exposing at high ISO is critical, because even ⅓Stop underexposed can exacerbate to appearance of noise with some cameras when used at ISO3200 – I am ignorant of first hand use of the D750’s performance, so that is a general comment.


*

I'm not getting a mauve color cast on my monitor.

There is a universal colour cast.
It appears to be magenta (major)/cyan (minor).
Here is a quick A/B referencing the ice as (averaging) neutral grey tones throughout the image.
The original is on the top:



A bluish colour cast can be indicative of underexposure, in some lighting conditions.

I am not stating that the colour cast is an indication of underexposure in this particular image.

WW

[mastamak]

Just to keep a balance, here's how my Sigma SD1M handles ISO 3200 . .



. . which is why my camera setting never varies from ISO 100 . .

See image full-size and observe:

the infamous Foveon green/magenta blotching is well in evidence

weird discoloration on the front of the speaker

horizontal banding all over the place

de-saturated color (what there is of it)

that the pink lampshade is actually light tan . .

Definitely not a low-light action camera. Didn't buy it for that of course, Wife's DMC-LX7 works well enough for that king of thing.

Hello Ted,
I don't know that I like your high ISO image that much but I would love to get my hands on the Hydrographica book
Grant

[Shadowman]

Hi Bill,

You are correct on all counts, I was trying to pan with the skater who was at a pretty good pace and I did not compensate for the swaying movement. I did try a few shots at a slower shutter speed to let in more light and blur background but wasn't happy with the results.

Here is another with similar settings (aperture changed to 2.8, ISO 3200, 1/100s-85mm lens). I settled on this shutter speed assuming it was safe to handhold and get a reasonably exposed image with those settings, I wasn't happy with higher ISO, although I did go as high as 16000 on a few shots.

Thanks for comments and tips.

[xpatUSA]

Hello Ted,
I don't know that I like your high ISO image that much but I would love to get my hands on the Hydrographica book
Grant

Thanks, Grant,

Regrettably, it is faux. Actually it's a photo album, sorry.

[William W]

Hi John,

Thanks for the extra information.

. . . [in the first image in post #1] . . . pan with the skater who was at a pretty good pace and I did not compensate for the swaying movement. I did try a few shots at a slower shutter speed to let in more light and blur background but wasn't happy with the results.

In the first image, it would be very difficult to pan it at 1/100th Second and to get a SHARP Subject, especially if the skater was moving reasonably fast (i.e. faster than a slow walk).

There are two (interrelated) reasons for why panning that shot would be very difficult:

1. The Subject is NOT moving at 90° to the axis of the Lens (i.e. NOT transverse movement). So that means the Skater’s motion can be broken down into two DIRECTIONS of MOVEMENT: Moving across the Lens; and also Moving toward the Lens.

2. The Framing is reasonably tight on the Subject (i.e. Framing is the combination of Focal Length and Subject Distance.) Having a tighter Framing on the Subject means that a faster Shutter Speed is required to arrest any Subject Movement. So whilst you might have perfectly panned the TRANSVERSE movement of the skater, your Shutter Speed of 1/100th Second was not fast enough to arrest the movement of ‘Moving toward the Lens’

*

Here is another with similar settings (aperture changed to 2.8, ISO 3200, 1/100s-85mm lens). I settled on this shutter speed assuming it was safe to handhold and get a reasonably exposed image with those settings, I wasn't happy with higher ISO, although I did go as high as 16000 on a few shots

I agree with your assessment that 1/100th is safe for you to handhold. Note in the second image example the two Skaters in centre frame are definitely moving transverse to the Lens (90° to the Lens’s Axis).

It is quite clear that this shot was panned (evidenced by trailing edge blur on vertical railing posts in background); also the two Skaters in the centre of the frame are quite sharp indicates that the panning technique was accurate: the Skaters being sharp also indicates that 1/100th is a safe shutter speed for that lens at that Subject when you are hand-holding.

*

It occurs to me that the Rink was not evenly lit. This is a common lighting situation for Gymnasia and indoor swimming pools, both of which I am very familiar.

I recall we’ve discussed the 85/1.8: it’s a terrific lens to have (either Nikon or Canon), my EF 85/1.8 has been unquestionably the best value for money EOS Lens I have owned: I have always had a fast 85 Prime, Minolta; Nikon and now Canon.

Also note that noise is not as much noticed in a PRINT as on the screen – and when viewing on a screen (MOST) people don’t pixel-peep they just look at the picture, so, whilst it is always important to get the best image possible I think it is easy for the membership of Forums to place too much emphasis on the dissatisfaction about the presence of noise when we are picking over our pictures.

Really interesting conversation, thanks,

Bill

[Clactonian]

My trusty D70 at ISO 800 and that's high for me and it.

[Shadowman]

Hi John,

Thanks for the extra information.



In the first image, it would be very difficult to pan it at 1/100th Second and to get a SHARP Subject, especially if the skater was moving reasonably fast (i.e. faster than a slow walk).

There are two (interrelated) reasons for why panning that shot would be very difficult:

1. The Subject is NOT moving at 90° to the axis of the Lens (i.e. NOT transverse movement). So that means the Skater’s motion can be broken down into two DIRECTIONS of MOVEMENT: Moving across the Lens; and also Moving toward the Lens.

2. The Framing is reasonably tight on the Subject (i.e. Framing is the combination of Focal Length and Subject Distance.) Having a tighter Framing on the Subject means that a faster Shutter Speed is required to arrest any Subject Movement. So whilst you might have perfectly panned the TRANSVERSE movement of the skater, your Shutter Speed of 1/100th Second was not fast enough to arrest the movement of ‘Moving toward the Lens’

*



I agree with your assessment that 1/100th is safe for you to handhold. Note in the second image example the two Skaters in centre frame are definitely moving transverse to the Lens (90° to the Lens’s Axis).

It is quite clear that this shot was panned (evidenced by trailing edge blur on vertical railing posts in background); also the two Skaters in the centre of the frame are quite sharp indicates that the panning technique was accurate: the Skaters being sharp also indicates that 1/100th is a safe shutter speed for that lens at that Subject when you are hand-holding.

*

It occurs to me that the Rink was not evenly lit. This is a common lighting situation for Gymnasia and indoor swimming pools, both of which I am very familiar.

I recall we’ve discussed the 85/1.8: it’s a terrific lens to have (either Nikon or Canon), my EF 85/1.8 has been unquestionably the best value for money EOS Lens I have owned: I have always had a fast 85 Prime, Minolta; Nikon and now Canon.

Also note that noise is not as much noticed in a PRINT as on the screen – and when viewing on a screen (MOST) people don’t pixel-peep they just look at the picture, so, whilst it is always important to get the best image possible I think it is easy for the membership of Forums to place too much emphasis on the dissatisfaction about the presence of noise when we are picking over our pictures.

Really interesting conversation, thanks,

Bill

Bill,,

Thank you for the follow up conversation, I will include your suggestions during my next effort.

[Shadowman]

My trusty D70 at ISO 800 and that's high for me and it.

Nice shot and effort.

[William W]

. . . So, my take-aways:

1. Don't use small online displays to evaluate image quality, unless that is the only way you are going to present the image. They will obscure problems that may be apparent at larger sizes.

2. Modern lower-noise cameras can handle fairly high ISOs reasonable well, if the lighting is pretty good, meaning that the distribution of luminance has its mass fairly far toward the bright end.

3. If you have to use high ISOs with images that have a lot of the mass at the darker end, the results will not be as good.Seem like reasonable conclusions?

Yes. Ditto Terry’s and randy's comments. Your test of the MkIII correlates to my fiddling with one and doing similar shots across the compass of its ISO.

*

. . . I have seen not one post addressing the techniques that can make or break success shooting at high ISO. . .

Regarding Dan’s Kortez’s point #3. I learned this back in 2004/5 when I cut over to Digital. My general solution when there is a large mass of shadow detail and if it is at the limit, forget the detail and just make the dark mass “black-black”.

When any shot approaches the limits of the gear and/or the Photographer - the Photographer must begin to evaluate the compromises available to her/him. I think this is one very important reason why we should know what are the limits

This is an example using an EOS 20D at “H” (‘equivalent’ ISO3200) it was shot around 2005, there is an huge amount of noise in the detail of the Girl’s Racing Suits, my solution was to just make the suits - Black:


"Famous" 20D / 70 to 200L (at 200): F/2.8 @ 1/800s @ ISO3200

*

. . . There are many reasons for shooting at high ISO values. Low light is one of them.

Yes, not always low light.

Sometimes it is to get to a necessary Shutter Speed and Aperture in reasonable light: hard direct early morning sun side-light; reasonably fast action; tight framing – requires f/8 for a safe DoF and 1/1250th for a safe Shutter Speed . . . ISO is now already at ISO800, add a bit of cloud cover and I’d be at ISO1600 or thinking about compromising the DoF when two Players are in a tight shot:



*

. . . [Being UNDER-exposed is a] Cardinal Sin with high ISO!

I concur.

However referring to "limits" & "compromises" - here is a 5D and the shot is three stops underexposed - there were few other options and I don't like walking away from any shot: the conversation underneath the image explains all.

Obviously the EOS 5D is an old camera by today's standards and (by contrast) the marvelous shots we've seen made by the 5D MkIII I would not have needed to underexpose by three stops if I took that shot today . . . . but the point is the 5D MkIII has its limits also - and it is good to know what the limits are, whatever gera one is using.

WW