High ISO noise - how long until m4/3 is 'good enough'

[dhazeghi]

http://www.clarkvision.com/articles/does.pixel.size.matter/

Try using a D4 for comparison and not a D800.

I'm not quite sure what you're getting at here, but I have in fact compared those cameras, and as DxO predicts, at high ISOs, the results are pretty close. The D4 has a slight advantage as one goes up the ISO scale, but it is slight and easily attributable to the use of a better ADC and other hardware.

That's not to say that I wouldn't prefer 'fewer better pixels' in most circumstances, but I don't think the current sensors provide much evidence that this approach leads to significantly better high ISO images. The advantages have more to do with workflow at a time when single-threaded CPU performance has basically stalled...

 

[meyerweb]

It's more than good enough for me already. I've got shots made with an EOS 10D, at ISO 1600, that with NR applied in Lightroom are satisfactory. Far better than film was at that speed. Current model m43 bodies are as good at 6400 as that 10D was at 1600. I don't need any better than that. A bit of noise doesn't ruin an image any more than film grain did. I think some of us have unrealistic expectations, and unrealistic ideas of what is needed to make good images.

If what you expect is completely noise-free at 6400, then shoot medium format.

 

[Amin]

I found a bunch of high ISO Canon 5D (mk1) samples and had a chance to compare them against similar E-M5 samples. Honestly, the E-M5 ones don't really look noticeably better. Interesting to think that it took until mid-2012 for m4/3 sensors to equal the low-light performance available in full-frame sensors as of late 2005. Of course, you can get it now in a camera that costs $3000 less and weighs only 1/3 as much!

It's a pretty good feat to approximately match the low light performance of a sensor that is 4x larger, and the GH1 was most of the way there in 2009. I think the quality of IBIS on the E-M5/E-P5/E-M1 is a real game changer though. When I had the original 5D, most of my photography was with unstabilized primes and still subjects. Now I can shoot those same still subjects often at a small fraction of the ISO I used to use. I'm sitting here in a completely dark room, lit only by a small computer display, and I can take reasonably clean images of stuff that is too dark for my eyes/brain to see at all, made possible by handholding steady for a full second with a 35mm equivalent prime. On the 5D I would have needed 4-5 stops more light have any chance of avoiding motion blur at the same focal length equivalent.

The big change that came about with sensors over the past few years wasn't so much about light collecting efficiency as it was about read noise, and that breakthrough affected small sensors (eg RX100) as it did large ones. It's been many years since I felt like the bright areas of high ISO pics were too noisy. Only in the past few years have we been able to dig detail of the dark shadows.

 

[Adubo]

the e-m5 is good enough for me. the e-m1 takes it a little bit better but whatever. improved gear is important as you go on with your photographic journey and you improve your skills, but in my opinion the right gear-depending on what is "right for YOU" + a solid skill set as a photographer can only make great results.

 

[meyerweb]

Granted, my technique can use improvement, but I too would like to see "better" results at higher ISO. Even with my 25mm f1.4 I find that shooting in low light results in blurry/out of focus pictures or very noisy with high ISO. Also, it would be nice to have lower ISO at least like the 100 of my old Nikon D60.... BTW, I shoot with the OM-D E-M5

Well, we'd probably all like "better" at almost everything. But the original question was "good enough." I'd like to get 400 hp and 40 mpg in my car. Heck, 500 hp and 50 mpg would be even better. Not going to happen, though. And 200 hp and 28-30 mpg will do me just fine. It's "good enough." And I'd certainly like to get ISO 6400 noise free. But I think the latest m43 sensors are "good enough." 2 to 3 stops better than an APS-C sensor from not that many years ago.

 

[Reflector]

I'm not quite sure what you're getting at here, but I have in fact compared those cameras, and as DxO predicts, at high ISOs, the results are pretty close.

...DxO has historically shown itself to be biased and inaccurate in their testing results. Not only that, but in practice you can say the D800E at the high ISO region, even with resizing, is only ahead of the latest m43 sensors by 1-2 steps in most situations, not the "massive" number you see displayed by their results. On the other side, the D200 should be "very close" but their numbers are neither exponential or linear in a way you can even compare.

Ignoring the "single threaded CPU" (...Whatever this means) I have been able to produce just as good results in ACR+Photoshop+NR programs as DxO, given more time.

The D4 and D3s are significantly ahead even with resizing. They are the few examples of cameras that I consider worthy in the "equivalence" argument. Otherwise sensor performance individually brings down stuff like the D800 and D600 down to the point you could almost compare ISO 12800 shots with m43s with aggressive NR. Not only is that shown in my previous post but I show this over and over with my own, 1:1 shots that I present. The only stuff I see from the supposed "FF equivalence beats tiny postage stamp sensors" crowd is worthless, sub 2k images that lead me to believe they're too inept to even handle their own camera.

The only case where I've seen very small pixels perform admirably is Nokia's clever approach of pixel binning where they are going from around 40 down to 8. And that comes from Nokia integrating the sensor with their approach and not just slapping it into the phone.

 

[Neftun]

The bucket analogy for sensors illustrates one aspect of light but light is a wave. The larger bucket on larger sensors does not equate to more light in the bucket, we aren't collecting rain really. Photosite size is the key. Something that a lot of reviewers seem to miss.

Not trying to be pedantic here, but I have to straighten one thing out here: In physics, light has a dual nature. Photons can be both described as particles and waves, depending on the phenomenon observed.
Do carry on


Patrick K

 

[flash]

There's a difference between the highest noise free ISO and the highest ISO. Lenses are both faster on m4/3 and the Olympus cameras have class leading IBIS. Especially with primes you can shoot in completely different settings with m4/3 versus a DSLR. You may be able to get a sharp image at several stops lower ISO on an EM5 compared to a 5D3 if shooting a standard 1.4 prime. Of course the othe system has advantages as well (handling perhaps).

The point is that I can't make a judgement just based upon what will happen on a tripod or a test bench. It's up to a system as a whole to determine how far it can be pushed.

Gordon

 

[jrsilva]

I would rather wish that my G5 had ISO 100

 

[nstelemark]

I 100% agree this is lot more complex than buckets. My point was at the sensor at given fab and generation photo site size has a lot to do with noise performance.

The rain in the bucket is an analogy that is taken too far.


Sent from my iPhone using Mu-43 mobile app

 

[nstelemark]

Not trying to be pedantic here, but I have to straighten one thing out here: In physics, light has a dual nature. Photons can be both described as particles and waves, depending on the phenomenon observed.
Do carry on


Patrick K

Yes it certainly does, but the image of rain falling through the lens aperture blade opening and hitting the sensor is all wrong. Until it hits the sensor it behaves for our purposes as a wave.

 

[Neftun]

Yes it certainly does, but the image of rain falling through the lens aperture blade opening and hitting the sensor is all wrong. Until it hits the sensor it behaves for our purposes as a wave.

You may be right there. I don't have the knowledge to agree or disagree, really, I dropped out of photonic engineering, apparently wasn't bright enough (pun not intended).

But I do agree the analogy seems overly simplified. Considering there is enough light emitted from bacteria and other microbes to view them clearly with a microscope, there shouldn't be too little light within a 17mm*13mm area to register efficiently. Technology, I'd argue, is what's holding back, not physics.


Patrick K

 

[Gary Ramey]

Not theory, it's fact. Bigger sensor = more area for collecting light = better low light performance. Even if a new type of material would be made, utilizing the same sensor technology, one in say m4/3 size and one at full frame, the full frame would still have the physical advantage.

Not necessarily. I thought it was based more on pixel density. Noise always will be related to the energy applied to the sensor. A 36MP full frame will struggle as much as a 12MP 4/3 sensor, if all other characteristics are equal. The higher the ISO the more power applied to the sensor and therefore more heat and noise. I would think that if you have the same speed of optics your light collection would be identical.

 

[powderbanks]

Not necessarily. I thought it was based more on pixel density. Noise always will be related to the energy applied to the sensor. A 36MP full frame will struggle as much as a 12MP 4/3 sensor, if all other characteristics are equal. The higher the ISO the more power applied to the sensor and therefore more heat and noise. I would think that if you have the same speed of optics your light collection would be identical.

I'm just saying on a simplistic level, a full frame sensor is physically bigger than a m4/3 sensor. All things equal, a full frame would, by default, collect more light. As to the technicalities of sensors and the black magic and witchcraft associated with turning photons into binary code, I think the physical size is becoming less important.

 

[psu_13]

What matters is the picture you get out the other side of the sensor/processing pipeline.

The only DSLR I've used that has better noise performance at ISO 3200 than an E-M5 is my D700. And it's not *that* much better. The E-M5 is much better than my D200 ever was, and probably comparable to the D300 back in the day.

I think this level of performance is more than good enough. You could not have dreamed that this would be possible when shooting film.

 

[RT_Panther]

Meh....
I could really give a rats @ss what the "others" think.....

I just shoot and have fun...