Sunday, October 13, 2013

Battle of the sensor sizes – full frame (FF) versus micro four thirds (m4/3) – is there a 2 stop difference?

Well, yes. In fact, if you look at low light JPEG output from say, the D600 (almost state of the art – the D800 holds that crown) and the E-M5 (pretty much state of the art for its body type – I actually prefer its output over the newer E-M1), you will find that there is more than a 2 stop difference.

Looking at the D600 at 6400 ISO and the E-M5 at 1600 ISO, it is striking how much extra noise reduction is required for the E-M5 to present a decent image. Remember … low light.

Now, if you are reducing to something tiny such as 720px, which happens to perfectly fit the center column of this blog, then you are not going to see much difference. Although, in this case, the D600 JPEGs are much more palatable to me as I am favorable to warm tones and open shadows. The E-M5 presents a colder look and the contrast and processing are overall fairly harsh. That has always been my opinion, though, so make up your own minds.

I chose a pair of images from the mannequin series on http://imaging-resource.com, so of course they hold the copyright to the original images and you will not see me reproduce those here except at tiny size and crops.

Note: I got a huge amount of information about how the images were shot from the excellent EXIF viewer that Jeffrey Friedl offers on his blog. It binds into Chrome and gives an amazing amount of insight into any posted JPEG image that still has its embedded EXIF.

image
image

The effective focal length for the Olympus is 30% higher, with the adjustment in distance being only about 8 inches. After I cropped the D600 to 4:3 aspect ratio to match the Olympus, I also upsized the Olympus very slightly to match the Nikon’s size so we could get a normalized comparison.

In keeping with my definition of fairness, I then applied Topaz Denoise to both to get the best balance of noise reduction and sharpening (if you know how to use Topaz Denoise, you can get amazing results with no detail destruction at all.) I also played with curves a wee bit to get the softer Nikon to come closer to the harsher Olympus in tone.

And here is what they look at in tiny sizes … clicking on either image will take you to the originals on Imaging Resource.


Nikon D600  Sigma 70mm 2.8 Macro  6400iso  f/4  -.3ev


Olympus E-M5  Olympus Zuiko 50mm 2.0 macro  1600iso  f/4  +.3ev

So we have very little visible difference at such small size, at least in detail. The Nikon image is warmer, with the reds being particularly warm. But otherwise, they are a pretty close match.

Now … what is under the hood? Here are crops at 100%:

Now, that is not even a contest. The smearing on the Olympus image is very evident in each of the crops. The hair is a true helmet when compared with the D600 image, which has some clumping but masks it well with retention of a lot of detail. And that is already at 2 stops advantage, so the difference in JPEG prowess is probably approaching 3 stops. This is a phenomenon I have mentioned in past posts (although it has been a while.) That being that larger sensors have more of an advantage than simple noise measurements would suggest. The quality of the details is just that much higher as less noise reduction appears to be necessary.

The flowers are a truly extreme example, where the Olympus is a painting while the Nikon has a very fine grain that would not be noticed on paper, even at gallery size. Remember that you are effectively viewing a 55” print from about 20 inches with these crops.

And the pocket is the same issue … the Olympus is ok, but the Nikon is excellent. Again, at 2 stops higher ISO.

Now, the differences really would be closer to 2 stops if I had processed these myself in Lightroom. That way, the heavy handed noise reduction can be avoided and you can sneak up on the best balance of noise reduction and sharpening. But that is not the nature of a JPEG engine. To satisfy the largest number of people – most people still appear to favour the sledge hammer that Olympus uses for a JPEG engine – they have to “turn it to eleven.”

I favour the Nikon and Panasonic approach to JPEGs … much more subtle. Of course, Panasonic JPEGs have only recently managed to get there, but Nikon has been putting great JPEGs out for a long time.

Now … remember when I said take this with a grain of salt? Well, the crops are no obvious that there is no salt involved in this equation at all. If you are a JPEG shooter and want to do, for example, weddings … then do yourself a favour and shoot the D610 and skip over the m4/3 bodies. Helmet hair is just not on.

If, on the other hand, you are a RAW shooter, you might get away with putting the Olympus cam up to 3200 and processing very carefully. But even then you will struggle as the light dims to get lots of keepers (contrast AF is not happy in low light) and you will struggle in Lightroom to come anywhere near the quality the D600 would deliver.

I think Montgomery Scott said it best: “ye canna change the laws of physics!”