ISO & Noise Level

Someone please educate me!


I went out shooting for a short while tonight and with the sun setting earlier and earlier, I really had to push the ISO. When I got home, I pretty much deleted every one of them. Could someone please tell me why ISO of 3200, 4000, 5000, 6400 is even an available option on the 7D? I can understand the wide range availble for shutter speed & f/stops but high ISO? Could someone give me a good scenerio where a shot would be worthy of this? I don't have PS but I can't believe there is a post-processing application that would be able to decrease the noise level & maintain a decent sharpness to still get a quality print level, except for maybe at 3200. Since I am a post-processing flunky, I most definitely could be wrong but I just can't imagine good prints @ ISO 6400, at least not from my 7d. Is ISO 6400 usable on a FF camera? Should I be getting more tolerable high ISO settings out of my 7D than I am possibly??


Denise

Hi Denise


I too often wonder myself why we have such ISO's available on the 7D. I do realize that at times there are reasons for using fast glass and high ISO to take that magic shot due t light conditions. Even with the 5D /2 anything over 2000 is pretty average. I have spoken to some experts friends, but even they could not shine a good light on it. Pardon the pun.


Look forward to the expert opinions. [:)]


Wally

Quote:

---

Originally Posted by ddt0725

Could someone please tell me why ISO of 3200, 4000, 5000, 6400 is even an available option on the 7D?

---

Because everyone has different shooting conditions, processing software, display resolution, and taste.


For shooting conditions, the color temperature of the light has a significant affect. In "bad" light, the could be a very big imbalance between the color channels. For example, red and green might be OK, but blue could be two stops darker. That means evenyou might have the camera set to ISO 3200, the blue channel is *effectively* ISO 12,800. So unless you get rid of that blue channel noise somehow (such as by converting to black and white without using much blue), the result will be as noisy as if you had set it to 12,800 in "good" light.


Processing software makes a big difference as well. For example, Lightroom 2 made a terrible mess of noisy images -- all blotchy and clumpy. LR3 is much better. The style of the post processing makes a huge difference too. High contrast, low-saturation, low-dynamic range image development shows noise a lot less than low-contrast high dynamic range images.


Display resolution is another vital factor. Every time you divide resolution by four you can increase ISO 1 stop and still have the "same" noise level. For example, if you felt that ISO 1600 was the highest you could go and still print a full-resolution 18 MP 18x12 beauty, then you would find that printing a 4.5MP 9x6" would give you the same noise visibility at ISO 3200. And 1 MP 3x4.5" print would again be the same at ISO 6400. And a 200x200 facebook image would look fine at ISO 12,800.


Taste is the big and obvious factor. Some people dislike noise more than others. (Personally, I actually *add* noise to my images.)

Thank you, Daniel! That is exactly how I needed explained to me!! I figured there had to be some reasoning behind it or it wouldn

I for one dislike ISO and I guess what I found out is probably what Daniel is saying. I don't like to get higher than ISO1600 on my 5D. But mainly because I view my images big and like to have the full benefit of all the megapixels. However lately at a wedding I also shot ISO1600 and for some reason, the shots with extra light from a flash look well....pretty amazing for ISO1600 shots.


I have seen this before and I don't know what's up, but the lightsource definitely plays a big role in ISO-noise as far as I''m concerned.


The noise application that I have is called noise-ninja. A lot of people and companies use it, but it does make your images softer. At least I think my poster prints look better without it. For some reason with printing you also get rid of some of the noise.


Don't ask me how though [A] Probably the smooth transitions between the different colors of ink or somethink.


But honestly, I also don't get why they put ISO12800 and even 25600 on my 5D other than marketing purposes. ISO6400 is actually very nice with video. But the same story here again, noise depends a lot on lighting conditions.


Jan

Quote:

---

Originally Posted by ddt0725

Could someone give me a good scenerio where a shot would be worthy of this?

---

Snowstorm? [:P] (someone had to say it....ok, maybe not [:)])


Seriously, I agree. The part that drives me nuts isn't that we have the option to shoot at high ISOs, options are fine with me, but that Auto ISO goes all the way to 3200 on my 7D and that is clearly too noisy. My understanding is that some cameras have the ability for the user to set the Auto ISO range, butthe 7Ddoesn't seem to have that feature. As a result I've stopped using Auto ISO. It was giving me too many noisy pictures when I could get the picture with changes to the exposure settings, not ISO increases.

Quote:

---

Originally Posted by Kayaker72

My understanding is that some cameras have the ability for the user to set the Auto ISO range, butthe 7Ddoesn't seem to have that feature.

---

True. The T2i and the 60D allow you to specify a maximum ISO for the Auto mode. In fact, the PowerShot S95 also has that feature. It's something that Canon could easily incorporate in a firmware update, but they've got a poor history of adding features via firmware.


Personally, I'm not a fan of ISO 3200 on my 7D, so for me that's a problem with Auto ISO. That's one area where the 5DII will perform better - like the 7D, the 5DII's Auto ISO is 100-3200, but ISO 3200 is cleaner on the 5DII. On the flip side, in M mode on the 5DII, Auto ISO is locked at 400, whereas on the 7D the Auto ISO is functional in M mode which is very convenient sometimes.

ISO is the new megapixel in camera marketing. Just don

Quote:

---

Originally Posted by barba

There are times where it could be argued that getting a shot, any shot, is better than nothing I suppose.

---

Yep. The S95 actually has a 'low-light' mode where the ISO ranges from 320 to 12800 - it does so with 2x2 pixel binning meaning you get a 2.5 megapixel image from the 10 megapixel sensor. ISO 12800 on a 4x crop sensor sounds horrible, but in practice it's not as bad as it sounds, because binning increases the signal-to-noise ratio. If Canon were to implement that with a larger sensor, the results would be interesting...

Quote:

---

Originally Posted by ddt0725

I don't have PS
<div style="clear: both;"]</div>

---

What are you using for post processing?


Mark

I have elements but haven

Denise,


I think the difference in high ISO performance between cameras can be atrributed to a large part on pixel size.


The 7D having 18 megapixels on a 1.6 crop factor sensor has much smaller pixels than a 40D which has 10 megapixels on the same size sensor. Think of 18 small buckets on your front porch in the rain versus 10 bigger buckets....the smaller buckets catch less rain drops each.


Thus.... the smaller pixel is able to collect a smaller number of photons...so ech pixel will have a lower signal to noise ratio.....now to increase ISO the camera amplifies electronically what it is receivng from the sensor and so the noise becomes even more evident.


The smaller pixels though will collect more detail.... so if you shoot a bird picture you will have more pixels on the subject and with proper exposure, you can get more detail and you can maybe get away with a larger crop of your images without losing image quality. This (and the fast frame rate and video) is the strong point of the 7D.....it is well known not to be a great high ISO performer. In the Canon line the 5D MKII and the 1D series are the best at high ISO. Also it is well known that the upper end Nikons are better high ISO performers than Canon.....especially the D700 and D3s.


It

Quote:

---

Originally Posted by Joel Eade

I think the difference in high ISO performance between cameras can be atrributed to a large part on pixel size.


The 7D having 18 megapixels on a 1.6 crop factor sensor has much smaller pixels than a 40D which has 10 megapixels on the same size sensor. Think of 18 small buckets on your front porch in the rain versus 10 bigger buckets....the smaller buckets catch less rain drops each.


Thus.... the smaller pixel is able to collect a smaller number of photons...so ech pixel will have a lower signal to noise ratio.....now to increase ISO the camera amplifies electronically what it is receivng from the sensor and so the noise becomes even more evident.

---

Daniel may chine in here, but I'm pretty sure this hypothesis has been debunked and the size of the pixels does notdirectly affect ISO noise. What does affect ISO noise is the size of the sensor - in other words, it's the total light gathered that determines noise, and total light is directly proportional to sensor size. That's the main determinant, with secondary differences resulting from the quantum efficiency of the sensor's pixels, in-chip processing, etc.


I did a quick empirical test of that myself - comparing images from two P&amp;S cameras, a Canon S95 and an Olympus C-765UZ. The S95 is 10 megapixels in a 1/1.7" sensor, so the pixel density is 23MP/cm&sup2;, and the Olympus is 4 megapixels in a 1/2.5" sensor, and thus has a lower pixel density of 16MP/cm&sup2;. So, the S95 has an 80% larger sensor than the Olympus, but the Olympus has 44% larger pixels than the S95. If larger pixels mean less noise, the Olympus should have lower noise, and if a larger sensor means less noise, the S95 should have less noise. Comparing at the same ISO, the S95 is the clear winner with noticeably less noise.


You would see the same thing if you compared noise across two dSLR sensors with the same pixel density, e.g. the 5DII and the 20D (both with 2.4MP/cm&sup2;) - the 5DII will have significantly lower noise at a given ISO, even though the pixels are the same size.


So, the smaller pixels of the 7D mean higher resolution than the 5DII, whereas the larger sensor of the 5DII means less noise than the 7D.

Quote:

---

Originally Posted by neuroanatomist

the size of the pixels does notdirectly affect ISO noise

---

I think the big argument was that it doesn't increase the noise of the picture overall. If you were to resize your 18MP 7D image to 10MP, you'd get an image with the same noise as you would from a 10MP sensor. At a 100% view of the 18MP image, you will see more noise per pixel. There are also some finer picture details along with the noise. If you're feeling you would have been better off with a 40D 10MP sensor, simply resize your 18MP image to 10MP and you'll have the same noise level, and perhaps some extra edge smoothness due to the extra detail also picked up by the higher res sensor.

Here is what I had previously read about pixel size.....I don

Great posts, everyone.

Quote:

---

Originally Posted by Joel Eade

Here is what I had previously read about pixel size.....I don't know....but it seems to make sense to me :


www.clarkvision.com/.../does.pixel.size.matter

---

Roger Clark is pretty brilliant, but there are several things he gets wrong. In a lot of places he mixes up the difference between pixel size and sensor size (claiming that large pixel sensors are better in circumstances when it's only better because the entire sensor is larger). In other places he compares sensor performance at completely different spatial frequencies without accounting for scale. It's like comparing road noise for at 50 MPH in one car and 100 MPH in another car -- you have to compare them both at the same speed to really find out which one has more noise.

Quote:

---

Originally Posted by Joel Eade

Here is what I had previously read about pixel size.....I don't know....but it seems to make sense to me :


www.clarkvision.com/.../does.pixel.size.matter

---

Nice read - thanks! I don't think the conclusions are mutually incompatible, but if you buy into the author's model of an 'apparent image quality' (AIQ) and look at how he's plotted that, it seems that while both pixel size and sensor size affect that value, sensor size has a much bigger relative contribution to the metric. Until he gets to the section on "Different Sized Pixels in the Same Sized Sensor," he's basing all of his comparisons on the S70 (0.38 cm&sup2; sensor) vs. the 1DII (APS-H, 5.48cm&sup2; sensor) - the latter has a 3.6-fold larger pixel pitch (linear dimension), but also a 4-fold larger sensor (linear dimension). Looking at his AIQ plot, a full-frame sensor surpasses the best possible performance of any APS-C sensor, regardless of pixel pitch.


We can all agree that the dSLR will perform better, but since both pixel and sensor size are much larger, that doesn't address the relative contributions of the two, and the author does nothing to distinguish them (which seems to bias the conclusions).


So, if you're comparing sensors of the same size but different pixel densities (e.g. 50D vs. 7D), the larger pixels should be better in terms of ISO noise, i.e. the 50D should 'win' - and by some of the metrics in the article it does. But by other metrics and real-world performance tests (e.g. Bryan's reviews), the 7D is actually a little better than the 50D - as Bryan states, "As sensor density increases, so too does high ISO noise -unless improvements in the sensor and subsequent image processing are made." Likewise, if you're comparing sensors of different sizes, the larger sensor will perform better in terms of ISO noise. But I think the sensor size effects trump the pixel size effects

Guess that

Quote:

---

Originally Posted by Joel Eade

How do the medium format cameras stack up in terms of high iso performance? Are they better because of bigger sensors?

---

MF cameras have better ISO performance than FF dSLRs, yes. The two highest-scoring sensors on that AIQ plot are the KAF-18000CE and the KAF-31600, and those are Kodak ~46x35mm sensors used in medium format cameras. So, both are higher than any FF sensor, with only the peak theoretical FF AIQ near the level of the 18 MP MF sensor. The 32 MP MF sensor is the clear winner in this metric.


So, looking at that overall trend and the theoretical curves for f/8 diffraction limit, it appears that sensor size defines groups within the AIQ range. For all non-P&amp;S sensors, existing models fall in a range of ~4-9 &micro;m pixel pitch. Within that range in terms of AIQ scores, the theoretical and measured sensors perform as follows:

• 1.6x sensor - AIQ of 40-55
• 1.3x sensor - AIQ of 55-80
• FF sensor - AIQ of 75-130
• MF sensor - AIQ of 120-170

Those group are almost non-overlapping, indicating that pixel size impacts AIQ primarily within a group defined by sensor size. That's consistent with sensor size being the key determinant of performance - it implies no matter how optimal your pixel size is for a given sensor size, the best you can do is to almost equal the least optimal pixel size in the next incrementally larger sensor. Put simply, sensor performance scales primarily (but not exclusively) with sensor size, less so with pixel size.


Looking at the DxOMark sensor scores, I wonder how Nikon cameras would fit into the AIQ plot...

Here iwww.dxomark.com/.../579|0/(appareil2)/436|0/(appareil3)/628|0/(onglet)/0/(brand)/Phase%20One/(brand2)/Canon/(brand3)/Nikons what I was searching for.......

That last post didn't come out right when I pasted the link.....anyway if you look at Dxomarrk and compare the PhaseOne65 vs Canon 1DsMKiii vs Nikon D3s.....the nikon is the clear high iso winner even though it doesn't have the biggest sensor and has the least number of megapixels but it does have the largest pixels.....so I'm really confused


[:S]

The Medium Format sensors are very far behind Canon and Nikon in technology, especially when it comes to low light performance (AKA "read noise"). Their new cameras match the performance of what Canon was doing about 5 years ago. That said, even with lower technology, they do have a pretty big size advantage, so that helps to balance things out. Depending on what ISO you are looking at, the DLSRs are either "same" (e.g. ISO 400) or much, much better (ISO 12800).

Me too! But I think it comes down to the fact that there's far more involved here than just numerical specifications. Roger Clark's evaluations are based on the specs - with pixel sizes, read noise and well depth in terms of number of electrons, etc., used to drive mathematical models of sensor performance. The DxOMark score is a composite of ISO sensitivity, bit depth, and dynamic range. If you want to isolate ISO sensitivity, click the Sports tab. In that category, Nikon still comes out on top, but all of the top 10 for that list are FF dSLRs, and the Phase One MF cameras don't show up until #11. Even so, if you look at the top two sensors, there's a substantial difference (ISO 3250 vs. 2300, so ~1/2 stop) between the D3s and the D3 - and both have the same size sensor and the same size pixels! Clearly, there's more to the story...

Well ...even if the reasons aren

Quote:

---

Originally Posted by Joel Eade

The Canon 7D clearly is not a great high ISO performer,

---

I kindly disagree. Personally, I would characterize the 7D as "great". Sure, it's 1.3 stops worse than the 5D2 at the same f-number (e.g. ISO 1600 vs ISO 640), but I like to remind folks that it comes at the cost of thinner depth of field. When you shoot them both at the same depth of field, then they have the exact same low light performance.

I rarely use anything over ISO 800, andthereforemy 50D and 7D suit me just fine. I took this shot with my 50D at ISO 800 Monday evening. The section that I applied an unsharp mask to didn't fair as well (her cheek/eye area farthest from the camera), but overall, I'm satisfied.


http://farm5.static.flickr.com/4034/...64d20d8b_z.jpg

Daniel,


My apologies......Poor choice of words on my part, should have said maybe not the greatest or best.....but I agree that great images can be made with all of the current choices available. Each camera body/system has it

Great post, Joel.

If I could dumb this post down a bit (which I am very qualified to do),if it has alot of "texture", you can make a decent 8x10 picture out of an ISO3200 picture. The picture below was just something I pulled up last night out of my "junk pile". If I wouldn't have chopped off her head, and framed it a bit better overall, I'd a kept it. This is a jpeg straight from the camera. I can't tell what it really looks like on the work computer screen, but noise isn't really an issue so long as your not posting it three feet wide. The point is, if you can avoid high ISO, do it, but don't be scared of it.[img]/resized-image.ashx/__size/550x0/__key/CommunityServer-Discussions-Components-Files/24/5148.mini_2D00_iso3200.JPG[/img]