High ISO


petebfrance 8 2.9k France
15 Feb 2017 8:47PM
What, only 25400? No good to me. Now, the new Pentax KP goes up to 819,200, that's more like it!
Our local supermarket occasionally sells cameras (DSLRs and compacts), and the maximum ISO and number of megapixels figure prominently in the specs. Meanwhile, back on planet Earth I limit mine to 320 (occasional foray up to 400!) for the bridge and usually 800 for the DSLR (although occasionally 1600).
But, as my view is based on old film tech and using things for low light rather than high shutter speed I'm sure that I am missing something here. I'm sure that the more technical savvy can give a more lucid view on things.

Join ePHOTOzine for free and remove these adverts.

15 Feb 2017 8:51PM
Somewhat irrelevant comment...

Amongst our company's products is a special infra-red sensor for detecting/counting printed labels on a backing web. Competitors claim a "switching frequency" of 5,000 Hz...so ours does 8000 Hz (and the very latest does 10,000 Hz...). In reality, if a counting machine actually ran fast enough to need that count-rate, it would probably rip itself off its mountings and make a hole in the wall...
...but it makes for a "USP" (Unique Selling Point)... that no one's gonna need/be able to use....! Tongue
User_Removed 6 328 United Kingdom
15 Feb 2017 9:54PM
General analogy: what's the motorway/highway speed limit in your country? How fast can your car go? Do you want a slower car?

Our answer 70mph, 100mph+ and No.
StrayCat 16 19.1k 3 Canada
15 Feb 2017 10:20PM
We were doing recurrent ground school on the B737 one time and the instructor stated the rpms of the engine driven generators; he paused, and then continued; would you believe we have pilots who don't know that? There was a chorus of; NO!, NO WAY! THAT's UNBELIEVABLE!
Chris_L 6 5.5k United Kingdom
16 Feb 2017 12:30AM

Quote:Why not just wait till it's a bit lighter? If you can't see what you're photographing in the dark I don't see the point.


Clean high ISO allows much more than that, it allows you to shoot handheld instead of needing a tripod, it allows you to use any shutter speed and aperture combo your camera is capable of whatever the weather.

5 years ago I couldn't shoot indoor sports that needed f8 and 1/2000th second. The lighting was too low, the photos too noisy, With a modern sensor I can. It's not marketing hype, It's genuinely useful.
ChrisV Plus
13 2.3k 26 United Kingdom
16 Feb 2017 1:50PM
The really high ISOs [we'll have to see with the KP, but I expect it will be the same] is just marketing hype.

These days most of the best sensors [usually made by Sony] are pretty much ISO-invariant. What that means is that you could take a shot at ISO 200, underexposed by four stops and 'push it' in software up to the equivalent of ISO 3200 and pay no penalty in noise over a shot at the 'right' exposure but actually taken at ISO 3200 in camera. There would be one difference and that's you're less likely to have any highlight clipping from the 'underexposed' shot.

That would not have been the case in the past as some of the noise in any given image would have been nyquist/other machine noise. Essentially that's additional noise generated by the camera's circuitry through heat and electromagnetic interference.

This suggests a couple of things which I think are borne out by tests. 1 That camera circuitry is now so efficient almost all noise is shot noise - which is dependent on the physical property of photons and as such is a baseline which can't really be altered. 2 Given that, it means using the same paradigm presently employed, it's near impossible to get much more performance out of sensors than we have at present. I'd go further and suggest that it's way back since the appearance of the Nikon D3s that there's been any more than very small incremental performance gains.

Some cameras have done better [current Canon sensors vs last generation and the Sony A7RII vs the A7R for example], but that's only because the previous iterations of the sensors used were very sub-par and not ISO invariant.

The same pattern comes into play when looking at the performance of various sensor formats - MFT tends to have two stops poorer performance than 35 mm format because it's gathering that much less light [quarter sensor area equating to that as each time you double the area you're getting twice as much light].

That means ISO 1600 from a four thirds sensor tends to perform the same as ISO 6400 at 35mm all things being equal. If you look at comparison tests with up to date sensors that's what you get, give or take a fraction.

Ultimately it depends how averse you are to noise and what size you're going to print/view at. Personally I'd shoot [at a pinch] at up to ISO6400 on MFT so long as I didn't need to crop the image or view it particularly large. If an image is strong a bit of noise doesn't make it bad.

Top ISO ratings on the likes of the D5/D500 and [I suspect] the Pentax KP would only be useful if you were shooting postage stamps at macro and only needed to be able to identify it was a postage stamp [just!]

In other words if you get a camera that only goes to ISO 25600 and you really would have need of something really crap at ISO 224,800, just underexpose by 3 stops and push by that amount in software. Presto! You have a mushy, noisy piece of **** like you'd get from one of those expensive wonder-cams! [I wonder if I should have posted that in Viz Top Tips].

ChrisV Plus
13 2.3k 26 United Kingdom
16 Feb 2017 5:14PM
52130_1487264994.jpg


This is ISO 6400 on an EM1. I make no claims for the brilliance of the image [obviously]. It was shot as a test on a brewery tour in Bruges a few years ago. At the size you'll view it here there's little to complain about in terms of the overall amount of noise or the level of detail that's captured. If I were tasked with making a documentary shot in such dim conditions that only needed modest sizes, this sort of quality wouldn't be optimal, but it also wouldn't be unacceptable IMO.

Edit: The generously proportioned gentleman centre shot is my b-in-law. He shoots Nikon and I think that's a D200 in his mitts. The shots from that would be trounced by the EM1 [I owned one and there was a huge leap in quality between it and the D300].
StrayCat 16 19.1k 3 Canada
16 Feb 2017 7:49PM
So my point in my original post was that some reviewers, when they list pros and cons, like the one I read recently in Popular Photography magazine on the E-M1 II, will say in the cons list; only 25,600 ISO, in other words, don't buy it, it's a piece of crap.Wink I don't think it should ever be listed as a con, at least by an independent reviewer; let's face it, manufacturers will dream up anything they can for a bit of advantage. Good article there ChrisV, I'll take your word for it.Wink
Fma7 5 1.1k United Kingdom
16 Feb 2017 8:04PM

Quote:There would be one difference and that's you're less likely to have any highlight clipping from the 'underexposed' shot.


Two differences: any dark greys below an output value of 16 at normal exposure would be pushed down to black and they're never coming back.

And in this thread you said 5 stops, so which is it?
ChrisV Plus
13 2.3k 26 United Kingdom
17 Feb 2017 10:35AM

Quote:
Quote:There would be one difference and that's you're less likely to have any highlight clipping from the 'underexposed' shot.


Two differences: any dark greys below an output value of 16 at normal exposure would be pushed down to black and they're never coming back.

And in this thread you said 5 stops, so which is it?



I said 5 stops previously between ISO 100 and 3200. Above I said 4 stops between ISO 200 and 3200. Because both are just examples. Do you spot the difference? Probably not. Bad Fido.
RoyBoy 15 303 2 United Kingdom
17 Feb 2017 10:42AM
I am frequently amazed when reading reviews about camera ISO performance and comparisons between makes and systems that the inter-related benefit brought about by image stabisation in actual usage isn't mentioned.
My OMD has terrific 5 axis stabilisation in the body which means that in any given situation I can take a hand held shot at a much lower shutter speed and therefore lower ISO than would have been the case with another system not having such IS benefits. Assuming such equates to say four stops, this means whereas with the Olympus I might be able to get away with ISO 800, with another system say my Canon 5D mk3 plus 24-70 f2.8 I might have to be at ISO 12800.
Reference is made in this thread about to under exposing and then pushing the exposure back up in post capture. My understanding is that so doing gives rise to increased noise because a significantly greater amount of information is captured at the highlight end of the histogram than at the shadow end. Having said that ChrisV seems pretty knowledgeable and perhaps I have missed something in that which he states?
Chris_L 6 5.5k United Kingdom
17 Feb 2017 11:42AM
Image Stabilization is a game helper but not a game changer, for many types of photography (eg sports or wildlife with fast moving subjects) you still get a blurred image even if your camera doesn't need a tripod to shoot at 1/20s etc due to IBIS.

ChrisV is the ISO expert and I defer to him. I'm glad that manufacturers are allowing ever higher ISOs to be selected as they are confident that noise will be acceptable, whether it's the same as underexposing and pushing doesn't matter to me.

I know that on my early Canons above 800 unacceptable. On my Sonys I don't worry about 3200, 6400, 12800 and higher. I wouldn't have even bothered to shoot at such high ISOs with micro four thirds, I'd have put the camera away.

RoyBoy 15 303 2 United Kingdom
17 Feb 2017 1:08PM
Accept all that you say Chris re wildlife, sports, etc but you seem to have missed my point re benefits of IS when it comes to ISO selection, a fact not often mentioned in reviews when comparing one system with another.

ChrisV is clearly an expert, as I have acknowledged, but I do raise a query re the underexposing comment because it does seem to go against everything I understand re digital capture where you generally expose towards the highlight end of the histogram. As I have said, maybe in this case its me that has misunderstood ChrisV point and in which case I hope he will advise me on this point.

The only other comment I would make is that I am not saying that Olympus is the answer to everything. I looked at Sony when buying my first OMD. At the time the Oly electronic viewfinder was the best out there by a mile and I found the Sony lenses rather buly and heavy by comparion. Having said all that there is no doubt that Sony, Canon, Nikon and Fuji, etc are all great camera's and its a matter of personal choice re handling, weight, cost, preferred photographic subject matter, etc. I believe a good photographer would get great pictures with any of them.
ChrisV Plus
13 2.3k 26 United Kingdom
17 Feb 2017 3:10PM
I don't make any claims being an 'expert' by any means - I've just read up on the subject a lot so hopefully my comments are at least informed to a decent extent.

I did say on the subject the last time this came up:


Quote:Ultimately it depends on the image. If you want the cleanest file use the lowest ISO and claw back a bit by pulling highlight detail. If lighting is trickier and there's a lot of highlight detail and you can live with the penalty of a bit of additional noise, underexposing isn't a bad strategy.


To some degree your exposure does matter to the amount of detail and the remaining dynamic range you're going to get out of an image. If you underexpose by a lot more than the amount of highlight detail there is available in any given scene, you are to some degree going to be 'throwing away' a certain amount of your DR.

You also obviously can't get any more in terms of shadow detail than your sensor is capable of ultimately. So if you have the sort of sensor that 'falls off a cliff' at ISO 12800 [where noise, false colour, very limited DR, little detail etc] mean the image is unusable, pushing the image to those sort of limits in software is going to be equally as bad. So there is a fixed limit to what you can do and the amount you should push by and that would vary dependent on the nature of the scene being captured.

But if you look at what cameras are doing with the ISO boost, it's pretty much amplifying the signal in the same way, [ie it's not in any way magically increasing the overall amount of light captured] so naturally the levels of noise would correspond where what you're talking about is shot noise, which is just a non-negotiable law of physics.

What matters most is light captured. Three things determine that: The amount of time the substrate [in this case a sensor] is exposed to light; the quantity of available light getting through [determined by the size of aperture] and the size of the area that collects the light. In that regard ISO is really just a measurement of the rate of sample of a signal - lower sample means a lower signal to noise ratio.

At good to moderate light levels pixel pitch doesn't matter so much. Where pixel pitch does come into play is where there are low light levels so a lower S/N. At a certain point very small light wells [the individual sensels] will be increasingly unable to record a statistically reliable value, so doing things like adding and averaging adjacent values is going to become ever more unreliable. That's why the cameras that do perform the best at very high ISO levels tend to have lower pixel counts because the point at which they will 'fall off that cliff' will be a bit further on.
RoyBoy 15 303 2 United Kingdom
17 Feb 2017 3:25PM
Thanks Chris. It all sounds pretty "expert" to me. I appreciate your input.

Sign In

You must be a member to leave a comment.

ePHOTOzine, the web's friendliest photography community.

Join For Free

Upload photos, chat with photographers, win prizes and much more.