Main weakness of m43 (according do me)

[Aloyz]

Hi all,

After using m43 for more than 2 years I feel like the main disadvantage of the system is the way that lenses (especially the wide ones) are designed. Almost (?) all of them (even the high-end ones) have a significant distortion (5-6%) which needs to be corrected by software. Do you know why such a compromise was taken? It influences image quality. Even though the lenses are a bit wider to compensate it, still the matrix size is the same and details are deteriorated on edges. In landscape photography, where wide angle lenses are mostly used, one wants to have sharpness across the frame.
Apart from distortion, I'm pretty satisfied with the system and hope it will live long...

Do you also see distortion as a serious weakness?

Cheers.

 

[cjoliprsf]

Well, this is certainly not unique to M43. I'd think all modern mirrorless mount system which permits transmitting lens distortion parameters from lens to camera body are like that. Certainly was also this way with a Sony NEX-6 that I had for a while.

The advantage: simpler lens design, lighter and less expensive.

And since the parameters are passed to the camera body, the correction becomes completely transparent, the SOOC jpg is automatically corrected, even the EVF view is corrected! The only way to see the distorded image is by shooting raw, opening in a raw developing software and turn automatic lens correction off - then you can see the distorded image.

PS welcome to forum!

 

[oldracer]

With respect, I think you should be out taking pictures instead of worrying about stuff like this.

First, image quality: Probably 99% of members' photos are viewed on computer monitors directly or via social media. Some relatively small subset of those are also printed at relatively small size, say up to 16x20" Another smaller fraction involve very aggressive cropping or printing to giant size. IMO it's only the last two categories, especially aggressive cropping, where even the most maligned M43 lens will not produce a good result. I was in a printing class once where one of the other students had a beautiful print, at lest 16x20". I asked him what lens he used. The answer: the original 14-42 Oly kit zoom. IOW, I think the only "image quality" issues most of us will ever see will be when pixel-peeping, which is a fine hobby but it is not photography.

Second, post processing: We do all kinds of things to our images in post. Is there something special about post processing for distortion that makes it morally objectionable? Not for me.

 

[stevedo]

For me, it's the end result that counts. I am perfectly happy with the IQ I get from my M43 kit. I really don't care if some of the distortion is corrected optically and some in software.

@Aloyz How often do you print or publish your photos without the corrections applied? If, as I suspect, the answer is never, then why does it matter how a lens is corrected?

This forum is full of 1,000's of great photos which would have been so much better if the lens hadn't been corrected in software - said nobody, ever.

I agree with @oldracer, you are worrying about this way too much.

 

[mawz]

Hi all,

After using m43 for more than 2 years I feel like the main disadvantage of the system is the way that lenses (especially the wide ones) are designed. Almost (?) all of them (even the high-end ones) have a significant distortion (5-6%) which needs to be corrected by software. Do you know why such a compromise was taken? It influences image quality. Even though the lenses are a bit wider to compensate it, still the matrix size is the same and details are deteriorated on edges. In landscape photography, where wide angle lenses are mostly used, one wants to have sharpness across the frame.
Apart from distortion, I'm pretty satisfied with the system and hope it will live long...

Do you also see distortion as a serious weakness?

Cheers.

This doesn't actually have any impact on end IQ when comparing systems, as you only see the results after correction unless you are defishing deliberately or using a RAW converter which doesn't fully support the RAW format (or allows deliberate defeat of correction)

This behaviour is pretty much standard across the board in all mirrorless systems today, as it allows for simpler and smaller lens designs with superior performance. It does cost some corner resolution compared to a well designed optically corrected lens of identical optical performance (which don't actually exist), but it actually can improve CA, LoCA, flare and SA performance in return.

Modern UWA lenses are significantly better in the corners on higher-resolution sensors than older designs that are purely optically corrected, even after accounting for the correction. The modern mirrorless UWA's are significantly better in these areas than the DSLR options, even after accounting for correction.

 

[PakkyT]

Do you know why such a compromise was taken?

Mostly because they can. As others pointed out, m43 lenses are smaller and lighter when compared to the lenses used in the older 4/3rds system even though they use the same sensor. And part of the reason for this is because in the 4/3rds days the processing power of the bodies simply wasn't there to do distortion corrections and so what ever distortion correction you desired has to be done with the optics alone. Cheaper lenses got a bit less optically corrected while the high end lenses got a lot better optics to get the best image they could out of the back end of the lens.

As m43 rolled around and processors get even better, it became the norm to get your lenses close optically to the quality you wanted, but once the last percentages of improvements started to become an exercise in diminishing returns, just doing to final tweaks in firmware got you the rest of the way there.

It influences image quality.

I suppose theoretically but do you see it? I don't.

Do you also see distortion as a serious weakness?

It hasn't bothered me but I am also not nearly as picky as other people so maybe it is just me (I also don't notice the purple fringing as much as other people unless it is extreme). I used to use the old ZD 12-60 SWD lens on 4/3rds which was excellent optically but still people would mention the characteristic "mustache distortion" of the lens on the wide end. I never really saw it unless I purposely looked for it and even then I would be like "sure, I guess so" but don't recall any photos I felt were ruined by distortion. Naturally the subject matter (shot of a building vs. lake scene) makes a big difference as well as your own tolerance to it.

But your question I think is more to do with potential quality loss due to the correction of distortion rather than the distortion itself which you don't see unless you use a RAW viewer/developer that can show the image without the distortion correction. Again, depending on the subject, I am not sure how much a theoretical "softening" of the details in the edges or corners of an UW photos would ruin a photo for most people. Typically this kind of shooting is for the over all "wow" factor of the scene rather than going in and trying to count leaves on a tree or pebbles on a beach.

 

[BDR-529]

Do you also see distortion as a serious weakness?

Nope. It's the very reason why modern digital ILC lenses are technically so much better than the old ones.

Sounds weird? It's quite simple actually.

Lens designers must always make a tradeoff between several optical properties and decide which to optimize at the expense of something else..

Digital photography has given them new freedoms which their predecessors in the analog film era just didn't have because now there are several optical defects which can be fully corrected (almost) with in-body FW.

This list includes defects like vigneting, pincussion/pillow distortion and even purple fringing which modern lenses can afford to have to some extent because lens-specific SW will automatically remove them. In the old world it was very difficult to correct these and it always required skilled manual post processing.

This gives lens designers room to perfect other things like sharpness (up to corners), contrast, colours, lens flares etc.

Unlike corrections mentioned above, SW just can't recover data which is permanently lost due to lens imperfections like missing sharpness, contrast and colour rendering. The best SW can do is to create an illusion that looks better to human eye but isn't the real thing.

These were lost to some extent in even the best analog era lenses because designers had to optimize design for distortions and vigneting instead. These were the first things that were tested in every analog-era review but nowadays most reviews don't even bother to measure them since SW will correct those anyway.

 

[PakkyT]

The best SW can do is to create an illusion that looks better to human eye but isn't the same thing.

Actually I would argue that the human eye is a rather poor "camera" in that most of the scene you see is actually soft and out of focus. So even without SW, cameras take way better pictures of a scene than we see with our eyes. Our brains are the SW that gives us the illusion that are eyes are better than they really are.

 

[WhidbeyLVR]

What I have noticed about M43 is that many of the native lenses are are very sharp, across the frame, even wide-open. That is in comparison to my previous camera systems and compared to several adapted lenses. Does correcting m43 lens distortion in software soften the areas of correction (usually the corners)? Most certainly, but only very slightly. And the sharpness of the lenses is good enough to make that loss unimportant most of the time. I also shoot in raw, and have the option to disable the correction if I find the loss in sharpness objectionable.

 

[BDR-529]

Actually I would argue that the human eye is a rather poor "camera" in that most of the scene you see is actually soft and out of focus.

Human eye has a very small cone of very sharp vision but images we see are really created in the brain, not in the eye.

Eye is constantly scanning with this very shap cone everything we really look at and brain parses in real-time sort of "full vision" from these tiny sharp patches.

This is where you get this ridiculous number of human vision = 576MB. It assumes that your head is clamped between a vice so that only eyes are able to move. If you now scan everything you can see with this shapest cone and add all pixels up, you'll end up with the number above.

 

[BDR-529]

Does correcting m43 lens distortion in software soften the areas of correction (usually the corners)? Most certainly, but only very slightly.

If SW is correcting any sort of minor distortion it's essentially moving accurately recorded pixels from point A to point B nearby.

If done correctly, image might be slightly cropped but doesn't sort of lose data. SW must obviously calculate new pixels but since it's done with data that was sharp to start with, this is more of a mechanical calculation than educated guessing.

SW is not trying to re-create details which were lost to lens imperfections (like missing sharpness) but is just moving them to a new location.

Same goes for vigneting, as long as it's so minor that sensor has enough dynamic range to allow correcting the luminosity and SW can use lens profile which tells exactly where and how much to adjust.

 

[doady]

The optical correction has a cost too. It can still result in softness, even before digital correction is applied. If a little bit less optical correction allows for a sharper lens that is also more compact or has a wider zoom range, I think that is not a bad tradeoff.

Back when I was using 7 megapixel camera, I tried to avoid digital corrections or cropping as much as possible, but maybe now with 20 megapixels it is not as big a deal. I'm still stuck with the old mindset, but camera makers moving forward and taking more advantage of digital technology is probably not a bad thing.

I think that is the advantage of m4/3, that balance between digital and physical, being in a position to be able take advantage of both digital and physical technology, to not only better compete with smartphones but also better coexist with smartphones at the same time. Full frame and obsessing more and more over the physical attributes of cameras seems the wrong direction to take.

 

[WhidbeyLVR]

If SW is correcting any sort of minor distortion it's essentially moving accurately recorded pixels from point A to point B nearby.

If done correctly, image might be slightly cropped but doesn't sort of lose data. SW must obviously calculate new pixels but since it's done with data that was sharp to start with, this is more of a mechanical calculation than educated guessing.

SW is not trying to re-create details which were lost to lens imperfections (like missing sharpness) but is just moving them to a new location.

Same goes for vigneting, as long as it's so minor that sensor has enough dynamic range to allow correcting the luminosity and SW can use lens profile which tells exactly where and how much to adjust.

Let’s take a concrete example of the 12-40mm Pro. It is my favorite normal zoom and is very sharp, but at the shorter end it has significant barrel distortion. The correction for that is to stretch it into a pincushion to straighten things out again, and this stretch will make the image a little less sharp in the corners. This is similar to the correction involved in de-fishing a fish-eye image, but of much lower magnitude. Most of the time, I can’t tell the 12mm image from the 12-40mm Pro lens was stretched for correction. However, I have run into cases where I chose to disable the correction because the uncorrected barrel distortion was preferable to the slight reduction in sharpness.

So I think we are in agreement that the digital correction is a very acceptable compromise in many cases, but it isn’t a lossless operation of just moving details around. It stretches (i.e., scales) them, too, which reduces their spatial spectrum. With the de-fish correction for my fish-eye lens, this reduced corner sharpness is quite obvious at full resolution.

 

[Bytesmiths]

You should try the E-System 7-14mm ƒ/4. As good, or perhaps slightly better than, the µ4/3rds version after correction, albeit a stop slower.

And it's all optical.

I'll trade you even for the µ4/3rds version.

 

[Armoured]

The correction for that is to stretch it into a pincushion to straighten things out again, and this stretch will make the image a little less sharp in the corners. ... However, I have run into cases where I chose to disable the correction because the uncorrected barrel distortion was preferable to the slight reduction in sharpness.

So I think we are in agreement that the digital correction is a very acceptable compromise in many cases, but it isn’t a lossless operation of just moving details around.

I agree with your overall description (and solution). I think the point for most practical uses is that the trade-off of software correction is usually more optimal, for more users, given the constraints.

One of which is that certain types of corrections are quite tractable mathematically, and others are not, especially with lens profiles with known characteristics - and lens designers are optimising to leave the easily-corrected types and do this automatically in software, while working on the 'hard stuff' in the glass.

But - and what you hinted at - it can depend on subject matter and photographer preferences. Our brains find it a bit disturbing to look at images where 'straight' lines change shape in unnatural ways and we notice it, so correct for that distortion; and we're more forgiving about sharpness - usually. But in a subject that doesn't have those straight lines to begin with, preserving the sharpness may be preferable, say, a field of leaves or other highly contrasting but irregular small objects.

I'm using the straight lines showing as curved as an example here, similar for other issues - I think most of us can imagine in our minds - however vaguely - how a curved line would be corrected to make it straight, and how weird it looks to see straight lines depicted as curved or wavy.

 

[John King]

If anyone is wedded to optical correction, then use the FTs high grade lenses.

Otherwise, the 'problems' with software correction are mostly in peoples' minds. The lenses typically throw a greater diameter image circle than necessary, and most of the extra area is used for the correction and then cropped off in camera or by the PP s/w.

And yes, I prefer my FTs f/4 7-14 too, but it's bigger, heavier and was far more expensive at release.

 

[RS86]

So, after reading the replies, it seems the complaint is something like:

"I love the size and quality of M43 lenses, but I hate the way they get those results for me."?

Btw. I think Laowa has some wide-angle M43 primes with almost no distortion.

 

[JensM]

To this "problem", I just subscribe to "Out of sight, out of mind"...

 

[scb]

It's just not a problem for me. When correction is needed, it can usually be done in post processing. DxO Viewpoint 3 makes it easy to correct distortion, especially in architectural photos, but other post processing software can make the needed corrections. Seems to me, distortion with wide angle lenses is across the board. Using Adobe Lightroom or Corel AfterShot Pro allow for auto correction based upon lens used.

That one issue isn't one that would make me change to a different camera system as the pros and cons of each camera system makes changing systems rather challenging, or at least it would for me.

 

[pdk42]

OP - whilst your analysis is correct, I think the loss of sharpness at the edges is really too small to be considered as anything other than a very minor consideration. I do a lot of wide shooting and it's really not an issue. I am pretty picky too - so it's not that my standards are not high! Also, as others have commented it's not in the least unique to m43. I ran a Nikon Z7 for a few months last year and their 14-30 f4 zoom is exactly the same.