High ISO

[Rufus]

My 10 year old Canon 450D camera has ISO settings from 100 to 1600. Not very high compared to today's cameras.

I have taken on board the general advice to set the ISO as low as is compatible with a suitable shutter speed and aperture. Consequently I tend to shoot at ISO 100 or 200.

I gather also, that noise is more of an issue at the dark end rather than the highlights, and in the blues. But I take no notice of this because I try to shoot at ISO 100 or 200.

Is there a form of test that I could perform to see the extent to which increasing ISO degrades the image? I could just the same view at different ISOs, but that does not seem a very rigorous way to go about it.

I guess what I am looking for is to get a feel for when ISO 800 or 1600 might produce an acceptable image; and why/how a more recent and higher end model can go up to say ISO 64000, and whether that could ever have a practical use or always be too noisy (assuming that I don't want to have noise as part of the artistic intent). I suppose technology has marched on in 10 years and perhaps a today's cameras shooting at say ISO 32000 would provide the same level of noise as mine at 800. But how would I tell? Perhaps there is something in camera technical specifications that indicates this?

I shoot in RAW if that is relevant, and my camera is a cropped sensor which I gather is more likely to generate noise.

[Shadowman]

Rufus,

Your decision to use higher ISOs will be dependent upon the subject, any movement within composition, the dynamic range of the subject/scene. Consider shooting a lowlight cityscape, a scene with moving subjects, and perhaps an indoor scene such as a party.

[Andrew1]

For your camera your suggested test of running it up and down the ISO range is exactly how you should check the capabilities of your own camera. You will also see varying results if you do the same test for different scenarios such as outside in bright sunlight or duller subjects being shaded, evening or indoors. The noise will be different for each as the electronics process the information. The results from your camera will be different from even it's own previous or next generation.

It's because chip technology and electronics, whether switching or optical, is simply being developed too fast to keep up with. Your example of a 10 year gap in that world makes it ancient history. Wait until you see what's coming with Graphene and some other technologies being worked on now.

From a three year old article -
https://www.extremetech.com/extreme/...cmos-processes
https://www.extremetech.com/extreme/...ht-photography

For more advanced details on the cameras
https://www.dxomark.com/

[george013]

My 10 year old Canon 450D camera has ISO settings from 100 to 1600. Not very high compared to today's cameras.

I have taken on board the general advice to set the ISO as low as is compatible with a suitable shutter speed and aperture. Consequently I tend to shoot at ISO 100 or 200.

I gather also, that noise is more of an issue at the dark end rather than the highlights, and in the blues. But I take no notice of this because I try to shoot at ISO 100 or 200.

Is there a form of test that I could perform to see the extent to which increasing ISO degrades the image? I could just the same view at different ISOs, but that does not seem a very rigorous way to go about it.

I guess what I am looking for is to get a feel for when ISO 800 or 1600 might produce an acceptable image; and why/how a more recent and higher end model can go up to say ISO 64000, and whether that could ever have a practical use or always be too noisy (assuming that I don't want to have noise as part of the artistic intent). I suppose technology has marched on in 10 years and perhaps a today's cameras shooting at say ISO 32000 would provide the same level of noise as mine at 800. But how would I tell? Perhaps there is something in camera technical specifications that indicates this?

I shoot in RAW if that is relevant, and my camera is a cropped sensor which I gather is more likely to generate noise.

Once I played with a program called DeadPixelTest. I think its this one http://www.photo-freeware.net/downlo...ad-pixel-test/
It shows you the dead and hot pixels. You'll have to read the help file. It was a nice and simple utility. Noise is something as hot pixels.

George

[NorthernFocus]

Hi, Rufus. As Andrew said, the test you proposed in your OP is as good as any for testing a single camera. What one considers to be "too noisy" is dependent on the subject matter, the intended end use of the image, and is highly subjective to the user.

As for objective testing to compare one camera to another, the link that Andrew provided to dxomark is as good a source as there is available. They use a standardized test on every camera body and plot the results for comparison.

Yes much progress has been made in 10 years. But the real question is whether it matters to you and your needs.

[xpatUSA]

Is there a form of test that I could perform to see the extent to which increasing ISO degrades the image? I could just the same view at different ISOs, but that does not seem a very rigorous way to go about it.

Never owned a Canon, so this a general comment:

Perhaps your test should include a variation in shutter time.

ISO 1600, 1/8 sec:



ISO 800, 25 sec:




Sigma SD10, well over 90 deg F ambient.

Them Foveons aren't too fond of warm weather . .

[proseak]

What do you use to process? If High-iso is of interest, DxO wth it's PRIME NR is worth a look.

To compare your 450d with a modern camera (700d)

https://www.dpreview.com/reviews/canoneos450d/20

and

https://www.dpreview.com/reviews/can...d-rebel-t5i/17

[rpcrowe]

What is most important is at what ISO you can be confident of shooting in real-life use. That can be accomplished by shooting at a range of ISO levels starting at ISO 100 and progressing through the maximum ISO of your camera.

However, remember this:

Noise is dependent upon your exposure. Correct exposures will result in less noise than under-exposure, followed by a brightness increase in PP.

Since real-life use is what I expect that your intention is (rather than just an academic exercise in ISO); I would suggest that you include noise reduction in your testing. Noise reduction is available in-camera. Noise reduction is also available using Adobe Camera RAW (I'd would recommend shooting RAW for this testing) or using NIK Dfine (available in the FREEBIE NIK collection of programs). There are also many other noise reduction programs available in other PP programs.

I would not look down my nose at the 450D until I experimented with proper exposure and noise reduction...

I would then process your images to conform with your needs. Of course, if you use your imagery primarily on the Internet your noise problem (at any ISO) will be a lot less than if you are going to make large prints of your images.

Shooting night images (which is one venue in which some folks use high ISO) noise is often more apparent because most of the image in a night shot is often black...

[dje]

Hi Rufus

If you do look at DXOMark, the most informative measurement for assessing shadow noise is Dynamic Range. The higher the dynamic range, the lower the shadow noise (or read noise as it is called). The attached curves from DXOMark show a comparison between the 450D, the 760D and the 6D (full frame camera).






You'll notice that the 760D is marginally better than the 450D but bear in mind that the 760D is a 24MP camera compared to 12MP for the 450D and thus has smaller pixels. When you go to the larger pixels in the FF 6D, the noise gets better.

Dave

[Manfred M]

Rufus - if you head over to the previously recommended DxOMark website, you will see some graphs that you should think about - there is one that Dave has just mentioned that shows how the dynamic range drops off as you turn up the ISO, there is another that shows the impact of the signal-to-noise ratio (SNR 18%), tonal range and colour sensitivity. Notice how they all drop off as you increase ISO?

The bottom line is that the base ISO will give you the "best results", and your current practice of shooting there continues to make sense. I think you will also find that the subject matter itself / image composition as well as its final use will also determine how much noise is too much. In fact, a well exposed image is going to generally look reasonably clean, but if you have underexposed and have to bring up the brightness, you will start noticing a lot more noise, especially in the shadow detail.

As others have noted, there are noise reduction techniques out there in PP to help as well, but the quality of the final results may vary. Some algorithms simply smear the pixels while others do an amazing job in preserving the original image while doing a good job handling noise (I'm also a fan of DxO Optics Pro's Prime that Peter has mentioned). Nik (mentioned by Richard) is decent enough for some situations as per Richard's comments.

My mantra (even in the film days) was "shoot as low as you can go" when it comes to ISO. Sometimes that means shooting at some pretty high ISO numbers. I'd rather get the capture and try to do something about the noise in PP than not get the image at all.

[george013]

I did install the Deadpixel utility again and played with it.
The idea behind it is analyzing a dark picture, so one made with the lenscap on and the viewer covered. If evreything ok, then all the pixels will have a value of 0, theoretical. Dead pixels will have a high value, hot pixels will have a low value.
So put the cap on the lens, set it to M, both exposure and focusing and select tiff. Choice a shutter speed, i did 125, and start with the lowest iso. Then double the iso and just go on. I did rename them with shutter speed and iso, so 125 200 etc.
If you load them in the utility, you can see the amount of pixels that's above the lower threshold and above the higher threshold. Noise is in the lower part of the values.
If you want, you can see the influence of shutter time too. Choice the lowest iso and play with the shutter time.

It's just a utility to play with. It gives you an idea.

George

[Rufus]

Thank you everyone for your helpful replies. They have stimulated my thoughts beyond ISO, although that is still the main purpose of this thread.

I shall have a look at the tutorials area to learn more about dynamic range (mentioned by John and Dave) as I had not connected this to ISO and noise.

I had heard of Graphene, Andrew, but had not come across its potential application to sensors of the future, so that was an interesting browse.

Dan is spot on in questioning whether it matters to me and my needs. I generally get by at low ISOs and don't push my luck, but I am beginning to appreciate more that I could use ISO to allow me faster shutter speeds when I want greater depth of field via a narrow aperture.

In answer to Peter, I open RAW images with Adobe Camera Raw to make global changes, and then use Photoshop CS6.

My default approach is to set exposure to half a stop to the left/minus side of the exposure meter because detail can be brought out of the shadows. I should probably use the histogram more to ensure that there are pixels close to the far right.

Richard is also spot on; I am indeed interested in what ISO I should feel confident of shooting at in real life, so real life tests are definitely on the agenda for me and I will learn about NR in camera so as to include this. Then I can have a look at NR also in PP.

George mentioned the Dead Pixel Test, and Ted suggested different shutter speeds. I have followed this up taking images at various times from 1/30 to 30 seconds at all ISOs. The results showed:
ISO 100 - 1 hot pixel at 15 seconds
ISO 200 - All OK
ISO 400 - 1 dead pixel at 8 seconds and 1 dead pixel and 5 hot pixels at 30 seconds
Thus far very reasonable performance!

ISO 800 - 0 to 4 hot pixels up to 8 seconds; 6 to 7 hot pixels and 1 dead pixel at 15-30 seconds
ISO 1600 - Only 1 dead pixel (at 30 seconds), BUT 264 to 332 hot pixels at up to 4 seconds and then just 38 to 55 hot pixels at 8 to 30 seconds.

I am not sure what this means, but I infer that ISO 1600 is to be avoided! Hopefully the results have uploaded okay:

I used the default settings in the Dead Pixel Test software, but notice that reducing them led to more pixels being identified as dead or hot. I stuck with the default settings.

The next stage for me is to take some real life shots at different ISOs and visually compare them on screen and in print (A4) and see what I learn.

[Rufus]

Thank you John. I have just uploaded a reply to you and others who have replied to my OP.

[Rufus]

Thank you Andrew. I have just uploaded a reply to you and others who have replied to my OP.

[Rufus]

Thank you George. I have just uploaded a reply to you and others who have replied to my OP.

[Rufus]

Thank you Dan. I have just uploaded a reply to you and others who have replied to my OP.

[Rufus]

I wanted to respond to everyone individually, but did not appreciate the posts would all appear together, so thank you also those who replied after Dan. Please don't be offended that I did not reply specifically to you.

Perhaps there is a way on the forum to let several people know that I have appreciated and taken note of their replies? Merely acknowledging a "Helpful Post" doesn't cut the mustard when their contribution is be being responded to and developed.

As you can tell I am still a bit new to this.

[george013]

You've seen a big increase in pixel values shooting black at 1600 iso. You can redo the test with a lower threshold in the darks.
If it's important for you as photographer depends on the pictures and your taste. In general I don't dislike noise, though analogue noise is much nicer as digital noise.

George

[Manfred M]

Rufus - I personally don't care about hot pixels or dead pixels as this knowledge will not change how I shoot one iota. I just clean those issues up in post. Sometimes it does take a bit of time, but the results are worth it.

[Dave Humphries]

Hi Rufus,

Welcome to the forums.

I wanted to respond to everyone individually, but did not appreciate the posts would all appear together, ...

Perhaps there is a way on the forum to let several people know that I have appreciated and taken note of their replies? Merely acknowledging a "Helpful Post" doesn't cut the mustard when their contribution is be being responded to and developed.

You can click "Reply with Quote" (as I just did), which will quote the person's post and you can (as I have) edit that to remove all but the salient points. This is on an individual post basis though.

Clicking the little thumbs up is helpful to use, people will see that you have done so.

Other than that, for thanking several people in one post, you can just write it freehand and, if necessary, copy and paste quote little bits of text from each of their posts.

Hope that's helpful, Dave