[philbaum]

I've always wondered about the Sports ISO specified in the Camera Sensor DXO Marks - what that meant. So today i looked it up on their site. And it seems to be remarkably useful.

http://www.dxomark.com/About/Sensor-scores/Use-Case-Scores

An SNR value of 30dB means excellent image quality. Thus low-light ISO is the highest ISO setting for a camera that allows it to achieve an SNR of 30dB while keeping a good dynamic range of 9 EVs and a color depth of 18bits.

So when comparing cameras across mfr lines and across sensor sizes, one can now use this Sports number to qualitatively compare low light performance.

I object slightly to labeling it a "sports" rating, however. Anyone shooting moody pictures in the evening or indoors, knows the need for better low light performance.

Here's the kind of low light picture i'm talking about - taken during an evening art walk in a small town. Goal was to capture the mood and feeling of crowds walking around on a winter's evening - sampling and gossiping about this months art displays. Taken on a handheld Nex 6 with a Pentax 50mm f1.7 manual lens. 1/160s, probably fully open, iso 800. It was a struggle to keep the noise level down when processing this RAW image. A lightweight monopod or tripod would have been useful. But one can also see the camera i was using, the Nex 6, altho very portable, does not have the highest sports rating out there. But is it usable - of course.

[IMG][/IMG]

Here's a listing of "sports ISO" for some typical cameras:
1" sensor
Nikon 1 V3 384
Sony RX100 Mk3 495
Micro 4/3
Panasonic GH4 791
Panasonic GX7 718
APS sensors
Nex 6: 1018
Pentax K5 1162
Pentax K3: 1216
Nikon 7100 1256
FF sensors
Sony FF A7 2248
Sony FF A7R 2746
Sony FF A7S 3702
Nikon FF 610 2925
Nikon FF 750 2956
Nikon FF 810 2853
Nikon FF 4S 3074

Now, instead of just comparing MP numbers, we can also compare a number that means a lot more to me: ISO capability.

I have friends that talk about how wonderful their RX100's are, but yet look at their ISO performance: 495. And the folks that talk about how wonderful their 1" sensors, like the Nikon 1: 384 - not that great for low light.

From reading camera ads, one would think that all these cameras are stellar in all ways. Look at the details and differences emerge. And the MP density don't help certain cameras. For example, the D810 with its 36mp actually rates worse on its ISO than the more humble D610 with 24mp.

I keep reading comments that it will be 5 to 10 years before mirrorless cameras compete with DSLRs. Well according to the above data, the top dog in FF low light performance is the unusual A7S, and frankly is the camera i'm currently lusting after

If i'm missing something here, let me know. But i think this ISO capability is an interesting yearstick for the cameras i'm interested in.

---

Last edited by philbaum; 12-08-2014 at 11:33 AM.

[jimr-pdx]

That is good information, thanks for the extra effort! I like the 1-inch sensor for compact gear that can still isolate a subject but give me aps.c for low lighting like you show here!

[philbaum]

That is good information, thanks for the extra effort! I like the 1-inch sensor for compact gear that can still isolate a subject but give me aps.c for low lighting like you show here!

Thanks for the comment. I don't mean to denigrate the 1" sensor cameras, wish i had one myself. But APS cameras sure have taken a bruising lately from the FF fans. But as one can see, there are legitimate niches for all these sensor sizes. At least now we can see what one of the trade-offs is - i.e. low light performance.

[tuco]

Don't forget about dynamic range. At night, high contrast scenes are typical with night lights. You may only be looking at noise levels but how much light you can capture at a given ISO is also a factor on how smooth your tonal scale will be. That's more important to me than the noise. Perhaps because I'm use to grain in a film picture and today people want - expect - perfect, flawless pictures under any conditions. Generally, a FF or CMOS MFD will get you more DR for the higher ISO for night time shooting.

[philbaum]

Don't forget about dynamic range. At night, high contrast scenes are typical with night lights. You may only be looking at noise levels but how much light you can capture at a given ISO is also a factor on how smooth your tonal scale will be. That's more important to me than the noise. Perhaps because I'm use to grain in a film picture and today people want - expect - perfect, flawless pictures under any conditions. Generally, a FF or CMOS MFD will get you more DR for the higher ISO for night time shooting.

Good point. The dynamic range is more useful when shooting on a tripod where one can maintain a low iso like 100 or 200. But when one goes higher for handheld shooting, the floor tends to drop out of the dynamic range and it falls rather fast. The sports rating that DXO uses, takes DR into account by requiring that the dynamic range be at least 9ev, for the ISO listed. So its not the maximum DR, but still pretty decent. We all have heard photographers talk about their ISO standards, like i don't shoot about 800, 1600iso, etc. Thats because its an easy thumb rule to remember for their kind of shooting, its really hard to remember that swooping DR curve. So its an important number, but not so useful when high iso's are used.

[Christine Tham]

Interesting observations (for me) - assuming a doubling of the number translates to "1 stop" in improvement

Micro 4/3 is roughly 1 stop better than 1"

APS-C is not quite 1 stop better than Micro 4/3

Full frame however is 1-2 stops better than APS-C

It shows that the larger format plus larger pixels really does make a difference.

And the K5 really was a good performer for it's day.

I still think APS-C is in the sweet spot - nice balance between quality and size (of cameras/lenses).

[tuco]

...
Thats because its an easy thumb rule to remember for their kind of shooting, its really hard to remember that swooping DR curve. So its an important number, but not so useful when high iso's are used.

That's what DXO's sports ISO is about. It pretty much tells you the max ISO you can shoot at with the expectation of an acceptable noise level. They picked a threshold for that noise level and if you agree with it or not is another matter.

[philbaum]

Interesting observations (for me) - assuming a doubling of the number translates to "1 stop" in improvement

Micro 4/3 is roughly 1 stop better than 1"

APS-C is not quite 1 stop better than Micro 4/3

Full frame however is 1-2 stops better than APS-C

It shows that the larger format plus larger pixels really does make a difference.

And the K5 really was a good performer for it's day.

I still think APS-C is in the sweet spot - nice balance between quality and size (of cameras/lenses).

You're right Christine, when one sees the different formats sports rating listed together like this thread, there is a story there about the value of FF and other sensor sizes as well. For me, that story is more important than all this talk about FF DOF. And this popularity of the D750, shows that the enthusiast or pro consumer has some smarts about what a good camera is. The D4, which is pretty expensive model, 5 grand or something like that, looks to me like its being undermined by newer cameras like the 750. Although i know nothing about how its tracking AF compares to the 750.

[tuco]

...
The D4, which is pretty expensive model, 5 grand or something like that, looks to me like its being undermined by newer cameras like the 750. Although i know nothing about how its tracking AF compares to the 750.

Nikon artificially cripples the D750 as to make sure something is not equal to the D4s. And I think that is where Nikon's old-school, product line placement is a bad thing. As consumers can see right through the artificial limitations and don't tend to like it. What they should do to distinguish product lines is on a pro version make the difference "tangible" things instead of arbitrary, artificial things. For instance, a pro body could be more rugged, tough and made out of machined alloys, have more shutter accusations, physical body differences and what-not. Things that cost more to make and thereby justify the higher cost. Things such as lower flash sync or top shutter speed to differentiate the product line is pure BS!

[philbaum]

Nikon artificially cripples the D750 as to make sure something is not equal to the D4s. And I think that is where Nikon's old-school, product line placement is a bad thing. As consumers can see right through the artificial limitations and don't tend to like it. What they should do to distinguish product lines is on a pro version make the difference "tangible" things instead of arbitrary, artificial things. For instance, a pro body could be more rugged, tough and made out of machined alloys, have more shutter accusations, physical body differences and what-not. Things that cost more to make and thereby justify the higher cost. Things such as lower flash sync or top shutter speed to differentiate the product line is pure BS!

Very good point Tuco. And its particularly irritating when one doesn't find the crippling item until after you've bought it. Then it becomes a trust issue. If you don't like or don't trust the vendor, how's that going to color your experience with the camera. I agree, make the differences in physical items that one is informed about.

Although from the manufacturer's point of view, its a lot cheaper to have commonality of hardware between different models to reduce their storage and ordering expenses :-) Firmware crippling carries no long-term storage cost, other than the hostility from the customers

[Emacs]

Full frame however is 1-2 stops better than APS-C

Take a look:

Color depth: 24.8 vs 24.2 bits, that is 0.6 stops difference, thus less than a stop
Dynamic range: 14.5 vs 13.7 i.e. 0.8 stops difference
Low light: 2956 vs 1256. Crop factor of APS-C is about equal to 1.534, thus enlarged to FF 24Mp Nikon APS-C sensor will have 1256*1.534^2 ISO metrics. This equal to 2955.563936.
Thus, the difference is log₂(1.534^2) = 1.234 stops ~ this should be in any aspect. More than a stop, but it was expected since the sensor area difference.
The question is rather why some modern FF cameras performs that poor (obviously, I mean crappy A7 at first place with its obvious IQ problems).

[dosdan]

These days, there's not that much of a difference in QE between sensors (everyone's using microlenses, fill-factors are high and the difference between colour filters is not that great), so the DxOMark Sports Scores of sensors in different formats should be reasonably proportional to the ratio of their areas:

https://www.pentaxforums.com/forums/172-pentax-k-3/246214-dxomark-k3-k5ii-k5.html#post2623307

BTW, since writing the message linked to above, I realised that the DxOMark Landmark Score is a proxy for the ENOB (Effective Number Of Bits) performance of the imaging system's ADC.

Dan.

[Rondec]

I think tuco is right. The big issue these days is not noise, which can be cleaned up pretty decently with noise reduction software, but dynamic range. This is where a camera like the A7s is amazing. It's SNR follows the same curve as all the other full frame cameras out there, but it manages to hold on to amazing dynamic range, even at super high isos.

The DXO Mark certainly doesn't tell you what is usable either. I have shot iso 12,800 on both K5 and K3 cameras and it does OK. I wouldn't print 8 by 10s, but it certainly will get you a shot.

This was iso 12,800, 1/30 second, f4.5 on a K5 II.

[BrianR]

..
Nikon FF 610 2925
Nikon FF 810 2853
..

From reading camera ads, one would think that all these cameras are stellar in all ways. Look at the details and differences emerge. And the MP density don't help certain cameras. For example, the D810 with its 36mp actually rates worse on its ISO than the more humble D610 with 24mp.

Don't forget that these are ISO values where the cameras hit a minimum (and arbitrary) quality standard. Strictly speaking 2853 is a poorer score than 2925 but in practical use these values will be indistinguishable. It's something like 1/25EV, so it's more like you can shoot at pretty much the same iso, score the same on the parameters they used for quality, and get 50% more megapickles.

Distilling it down to a single number like this can always be handy, but all the data dxo gives should be considered if it's anything important (and sample pictures, other sources etc)


Mark me down as another hater of firmware crippling..

[philbaum]

These days, there's not that much of a difference in QE between sensors (everyone's using microlenses, fill-factors are high and the difference between colour filters is not that great), so the DxOMark Sports Scores of sensors in different formats should be reasonably proportional to the ratio of their areas:

https://www.pentaxforums.com/forums/172-pentax-k-3/246214-dxomark-k3-k5ii-k5.html#post2623307

BTW, since writing the message linked to above, I realised that the DxOMark Landmark Score is a proxy for the ENOB (Effective Number Of Bits) performance of the imaging system's ADC.

Dan.

Well, you and emacs know more about this stuff than i do. But if there is not much difference in QE between sensors - how did Sony come out with the A7S sensor, and secondly, isn't it likely that 1 or 2 more companies will follow Sony's tack on this low light sensor?