High ISO noise vrs underexposed noise

<p>Sorry, I had some trouble posting images so I just made a flikr slideshow. <br>

<a href="http://www.flickr.com/photos/40076327@N06/sets/72157625600131022/show/">http://www.flickr.com/photos/40076327@N06/sets/72157625600131022/show/</a></p>

<blockquote>

<p><em>“Even with my flash I would still like a little more exposer then I'm getting with ISO 800. I might not need to go as high as 3200 but It would be nice to have a little more faith in those tools if I so need them. I do use a defuser so I lose a little light from my 580. <strong>I don't want to shoot too slow because I don't have the worlds steadiest hands. I'm mostly using my 28-105L F4 because it gives me so much perspective freedom.”</strong></em></p>

 

</blockquote>

<p>I am not sure I am following the problem here.<br>

Even with a diffuser (like the Fong?), the 580 is a quite powerful tool. If you are inside and provided it is not a coal mine a lined with black velvet curtains and provided you are within shouting distance you will pull ISO800 most of the time and get adequate Flash illumination on the Main Subject.<br>

The Ambient of a Reception Lounge with Incandescent Lights usually allows a range of Tv, from (approx) 1/8s to 1/60s - using F/4 @ ISO800.<br>

The lens you have (I assume a typ0) is 24 to 105F/4 L IS USM – <strong><em>it has IS</em></strong> and provided you set the Tv to shoot the ambient about 2 stops under the flash exposure, with the IS engaged on that lens, you should have very few times where the Tv will be so slow, that you will engage noticeable camera shake in the resultant image.<br>

If you find yourself in a black velvet-lined cave or outside then use a Bounce card and to hell with the little shadows - also do as Nadine suggested and get an off camera cord and use the bounce card that way – that’s what I do. <em> </em><br>

<em> </em><br>

<em>***</em><br>

<em> </em></p>

<blockquote>

<p><em>"Have some test prints made so you can see with your own eyes. That's what got me over high ISO concerns. Up to 8"x12" I'm amazed at how good the old 5D looks at ISO 3200."</em></p>

</blockquote>

<p>Yes, I agree. The proof is in the printing.</p>

<p> ***</p>

<p><strong>"Unity Gain"</strong></p>

<blockquote>

<p><em>"Since all camera manufacturers and almost all raw converters don't tell anyone about the inner worki1ngs of their product technical discussions are many times fruitless. Sometimes you need both photography skills, post processing skills and certain equipment or software to make something work. So I think it is better to test an idea by doing some test shots and then decide for yourself if it works for you and has any benefits."</em></p>

</blockquote>

<p>I agree.<br>

And I add that the Proof (result to set your standards by) is still in the Printing even if the final product to the Client is to be viewed on a computer - or even a TV screen - they will not be looking at 100% and even MORE likely is that the TV or Computer has the aspect ratio wrong and colour and contrast will be off, anyway.<br>

I had already read Clark's essay on unity gain (and also other of his works).<br>

I have come to believe the "unity gain" is somewhat analogous with the point at which Canon provides “Extension of ISO”.<br>

I am sure I read a Canon White Paper in which this is alluded to - certainly the photon bucket was discussed, as was Reciprocity at Longer Tv – which I don’t think Clark gets into.</p>

<p>For those interested, here are the Summary Results for me as a compulsive tester apropos <strong><em>Limit of Maximum ISO to use</em></strong>:</p>

<p><strong>P5IS (JPEG only)</strong> – fine for an 11 x 14 print at ISO 200: <a href="../photo/11248154&size=lg">http://www.photo.net/photo/11248154&size=lg</a><br>

and can squeeze a very good 5x7 print at ISO 800 if necessary: <a href="../photo/11248155&size=lg">http://www.photo.net/photo/11248155&size=lg</a><br>

Underexpose it - and you are history: but still “just OK” for a small print or a small newspaper picture: <a href="../photo/11248156&size=md">http://www.photo.net/photo/11248156&size=md</a><br>

<br>

<strong>400D & 450D </strong>Play at ISO800 only – ISO 1600 can be OK, but I would rather use ISO800 and squeeze the Tv or Av, if I can. (No samples from these cameras in my portfolio - but there are some in older threads: note to self - put some in a folder for reference).<br>

20D & 30D: ISO1600 is very good. Can be close to very good at ISO3200 in a 14 inch wide print: <a href="../photo/10738830&size=lg">http://www.photo.net/photo/10738830&size=lg</a> but the shadows suffer – the black Racing Skins don’t have a lot of detail - I argue that the TYPE of lighting, plays a part in that.</p>

<p><strong>5D </strong>Ditto the 20D and 30D: <a href="../photo/10442931&size=lg">http://www.photo.net/photo/10442931&size=lg</a><br>

And, for fun, shot at ISO3200 then “Exposure +2.7” in RAW: <a href="../photo/10442964&size=lg">http://www.photo.net/photo/10442964&size=lg</a> is fine at a 7 x 5 print<br>

<strong> </strong><br>

<strong>5DMkII – </strong>only used her for a few hours but what I found was ISO 3200 was as ISO1600 is, on the 5D. I didn’t get a chance to extensively use the extension ISO on the 5DMkII</p>

<p>***</p>

<p><em><strong>I have found that using ISO 1600 on a 20D, 30D and 5D and pumping the exposure a stop in RAW: - ON THE AVERAGE is not as good as shooting at ISO3200 (“H”).</strong></em><br>

Now that might be because of many elements – perhaps the TYPE of lighting I often encounter is one of those factors.<br>

So after using both methods for many shots for many shots and comparing the outcomes I now set ISO3200 and don’t fiddle as much later in PP or think about “under by one” at the time of shooting. </p>

<p>There are more samples of "ISO testing" and also testing other stuff "in the field" in my portfolio, here at Photo.net; most of the tests are obvious and self explanatory. </p>

<p>WW</p>

 

<blockquote>

<p>Ok, this is a series of images [using a 5DMkII] I took at different ISOs. 6400 was a normal exposer, 3200 was - 1 stop and brought up one stop in ACR, and 1600 was -2 stops brought up two stops in ACR I just made a flikr slideshow.</p>

</blockquote>

<p>I was writing as you were posting the results.</p>

<p>From the images you posted, viewed on my studio monitor: I would have the “ISO6400 N” - it has more punch in the shadows, to my eye.<br>

I suggest you make test comparisons shot in your “usual” shooting conditions and think about it.<br>

IMO your one off test, is in line with what I would expect from a 5DMkII and is commensurate with my above summary.</p>

<p>Maybe I am biased and I was looking for what I found – I am open to that thought also and I am interested in others’ views, in this regard.</p>

<p>WW </p>

<p>I am surprised no one has mentioned that the major reason, in practice, for bumping the ISO is to avoid motion blur.<br /><br /> It's obvious that flash is an alternative if getting a 'correct' exposure is of primary importance, but there's really not much you can do about motion blur if you prefer natural light and shoot at too slow a shutter speed.<br /> <br />At the same time, there are all kinds of things you can do about luminance and chrominance noise, as is evidenced and discussed here a lot.<br /> <br />I used to be hesitant about shooting at very high ISO until I got my 5D Mark II and at about the same time began using PS CS5, which has much improved RAW noise reduction capability.</p>

On my D700 at higher isos teh exposure is critical. An iso 800 image can look like an iso 6400 image is underexposed and pushed in post.

<p>Rocky,<br /> from your samples it sure looks like shooting at ISO6400 is the best alternative for you when needed. The only thing that looks strange is that your images have loads of contrast and saturation and the brightness/exposure after correction is not the same. It also looks like you have sharpening and noise reduction active but there is still a lot of noise in these images. All this could screw things up for an apples to apples comparison.</p>

<p>I'll post some samples from an older Nikon just for reference. They are processed in C1 without noise reduction and sharpening and are highest quality jpeg for best viewing. This camera only goes to iso1600 hence the selection of iso. The iso 1600 shot is exposed to the right and left that way.</p>

<p> </p><div></div>

<p>And the 100% crops from the black patch in the lower right.</p><div></div>

<p>To me, Rocky's test images show exactly what I expected and nothing not already widely known and accepted:</p>

<p>- way more noise on the ISO1600 image pulled up by 2 stops<br>

- shadow detail is a lot worse on the same ISO1600 image</p>

<p>The second point really becomes obvious when considering that any show detail is now being heavily underoxposed which makes it all the more difficult to extract equivalent (to a properly exposed image) detail from it when dragged up in post processing. The opposite of severely clipping highlighst and trying to recover them in post. Think of the dynamic range limitations......</p>

<p>Well David, it's not as simple as it may look like at a first glance.</p>

<p>You see, with the same aperture and shutter speed <strong>the sensor receives exactly the same amount of light regardless of the ISO</strong>. Then after the exposure the camera reads every pixel and uses the iso to determine how much to amplify the signal in each pixel. The sensor is a linear device so for each ISO increase the amplification is doubled. Then the signal is converted from an analog signal to a digital. That digital information is then written to the raw file.</p>

<p>At some point of ISO amplification there is simply not more information to gain from each pixel. At that point it wouldn't matter how much you increased the ISO - there is no more information in the shadows to gain. At that point it is the same if you amplify the raw values in the raw converter or if you increase the ISO further. I don't know where that point is on the 5D MkII though.</p>

<p>It is also a well known fact that camera manufacturers often change the raw values from the sensor before writing them to the raw file. The raw file is already half-cooked so to speak. This is identical to changing the exposure on the raw file. For instance all 1/3rd stop ISO increments on all Canons except 1D series are done this way. The same with some of the higher ISO settings. It used to be that the "extended" iso settings was the fake ones and it would be better to push the raw file but some say that it has changed since the iso race begun.</p>

<p>But as I said before it is better to try and see what works for you and your workflow than to generalize.</p>

 

<p>Well Pete, it is not as complicated as it may look at first glance....</p>

<p>Of course at the same shutter speed and aperture the exact same amount of light hits the sensor, but this is clearly NOT the point.The point is that Rocky's image at ISO1600 has <strong>1/4</strong> the amount of light hitting the sensor compared to the shot at ISO6400. Clearly not the same amount of light. The ISO1600 shot is severely underexposed and pushing the limits of the camera's dynamic range. Imagine if it was 3 or 4 stops under and you tried to pull it back in post. The blacks (which when properly exposed are not blacks) would remain blacks as they would be buried below the dynamic range limit and when attempting to pull them up in post you would have massive loss of detail and horrendous noise issues. The exact same applies to the ISO1600 shots at 2 stops under, pulled up in post, just to a lesser extent.</p>

<p>One just cannot expect to apply the principle of shooting at 2 stops under, for example, thinking you will get the exact same results when pulling it up in post processing compared to exposing the image properly in the first place. The effect of bumping up exposure (in post) on shadow noise is very well documented on this forum and elsewhere, and I am sure the dynamic range issue is clear.</p>

<p>Just my thoughts...</p>

<em>"Rocky's image at ISO1600 has <strong>1/4</strong> the amount of light hitting the sensor compared to the shot at ISO6400."</em>

I thought of and typed out my response way too hastily so I need to make a correction my my statement quoted above => The 2 stops underexposed ISO1600 shot has exactly the same amount of light hitting the sensor compared to the nominally exposed ISO6400 shot. Pretty obvious as all you would do after taking the shot at ISO6400 is change the ISO setting to 1600 and leave the aperture and shutter speeds as-is, so apologies for that!

All my other comments hold...

 

<p>I appreciate you taking the time to write down your thoughts. You do have to understand though that what I'm sharing in this thread is nothing new. It's in fact well known even I assume you have never heard or read about it. What you are saying is true but only up to a certain ISO. I leave it up to you to decide if you want to further your knowledge or dismiss it as you have been so far. My participation here is for the moment over.</p>

<p> </p>

<p>Anyway... I thought I would do a quick test myself. The image below shows two crops, both shot at F5.6, 1/100th. The top image is a nominal exposure at ISO6400 and the lower image shot at ISO1600 at 2 tops under and pulled up by 2 stops in RAW. No sharpening, no noise reduction.<br>

I stand by my earlier comments that the image shot as ISO6400 has better dynamic range and less noise.</p>

<p> </p><div></div>

<p>That's a good test shot David. You do have a slightly different exposure in your test though. It's because camera manufacturers lie about the true ISO settings and there may be some variation in shutter speed and aperture between shots as well.</p>

<p>According to dxomark the 5D Mk 2 ISO6400 is really ISO3990 and ISO1600 is 1093. So if I calculated right that means you need pull up the ISO1600 around 2.13 EV or so for the exposures and histograms to match.</p>

<p> </p>

<p>PS. I can't raise the brightness on the jpeg you posted but if I lower the brightess of the ISO6400 to match the ISO1600 I get this:</p>

<p> </p><div></div>