17/2,8 Lens tests

[HKRZ]

For those interested a test of this new pancake for m4/3 from Olympus:

http://lenstip.com/1739-news-Olympus_M.Zuiko_Digital_17_mm_f_2.8_lens_review.ht ml

[Kyle Jones]

Thanks for posting this.

The cask-like barrel distortion numbers were very surprising to me. I had thought the CA was the worst of the trouble. The JPEG distortion correction works well, though I wonder how much sharpness is sacrificed for it.

[Alan Baxter]

Thanks for posting this.

The cask-like barrel distortion numbers were very surprising to me. I had thought the CA was the worst of the trouble. The JPEG distortion correction works well, though I wonder how much sharpness is sacrificed for it.

The loss of resolution would be considerable at the frame edges. I think it is sad that Olympus have given up on their optimizing optics approach.

Regards,

Alan

[bg2b]

I think it is sad that Olympus have given up on their optimizing optics approach.
They have to optimize size as well as optics, and you can't do both to the ultimate degree at the same time. In this case, they obviously decided that the main point of the camera/lens is small size, not ultimate optical quality.

[er1kksen]

The Pentax DA 40mm f2.8 pancake has absolutely superb optics in an even thinner package, and while it is a longer focal length, it's also for a larger format with a longer register difference. I don't really buy this whole "these are the compromises involved in making such a small lens" idea. I think Oly could do better. I mean, if they're not going to bother to make the optics as good as possible in reality, they could at least do a little digital magic like the recent Panny cameras and make them as good as possible in practice.

[Alan Baxter]

The Pentax DA 40mm f2.8 pancake has absolutely superb optics in an even thinner package, and while it is a longer focal length, it's also for a larger format with a longer register difference. I don't really buy this whole "these are the compromises involved in making such a small lens" idea. I think Oly could do better. I mean, if they're not going to bother to make the optics as good as possible in reality, they could at least do a little digital magic like the recent Panny cameras and make them as good as possible in practice.

They do a lot of "digital magic" - there is no obvious barrel distortion on the out-of-camera Jpegs, it is only seen in the raw files.

Regards,

Alan

[dnas]

The Pentax DA 40mm f2.8 pancake has absolutely superb optics in an even thinner package, and while it is a longer focal length, it's also for a larger format with a longer register difference. I don't really buy this whole "these are the compromises involved in making such a small lens" idea. I think Oly could do better. I mean, if they're not going to bother to make the optics as good as possible in reality, they could at least do a little digital magic like the recent Panny cameras and make them as good as possible in practice.


I don't see what the problem is. They've optimised the lens performance for resolution, sacrificing distortion & CA..... JUST THE SAME as Panasonic did with the G1.
They DID specify in those tests, that they had taken RAW output for the ACTUAL optical performance of the lenses, not the lens/camera composite result. The actual lens performance looks very much like the optical output from the Panasonic 14-45mm lens on the G1 at 14mm.

[Alan Baxter]

I don't see what the problem is. They've optimised the lens performance for resolution, sacrificing distortion & CA..... JUST THE SAME as Panasonic did with the G1.
They DID specify in those tests, that they had taken RAW output for the ACTUAL optical performance of the lenses, not the lens/camera composite result. The actual lens performance looks very much like the optical output from the Panasonic 14-45mm lens on the G1 at 14mm.

The problem is that Panasonic said that their strength lay in electronics and that they would use image manipulation to optimize images from lesser optics, while Olympus declared that their strength lay in optics and they would try to optimize their images by good optical design. That was their entire rationale for the 4/3 standard - it was designed around the demands of top class optics. If they aren't going to bother with that, what's the big deal about sticking with a small sensor? Or a low volume manufacturer?

It is a pretty big issue here - CA can only be partially fixed in software (its generation occurs by wavelength, but its correction is only trichromic). Similarly, the resolution of a barrel distorted image after correction is actually a lot lower than the "nominal" resolution of the sensor. This limits what you can photograph and where you can put the subject in the frame. Not a problem if you think the circle in the view finder is a target; but it is an issue if you like to compose your images.

Regards,

Alan

[dnas]

The problem is that Panasonic said that their strength lay in electronics and that they would use image manipulation to optimize images from lesser optics, while Olympus declared that their strength lay in optics and they would try to optimize their images by good optical design. That was their entire rationale for the 4/3 standard - it was designed around the demands of top class optics. If they aren't going to bother with that, what's the big deal about sticking with a small sensor? Or a low volume manufacturer?

Ok, fair enough if that's what Olympus said.... did they say that about micro 4/3 as well as 4/3???


It is a pretty big issue here - CA can only be partially fixed in software (its generation occurs by wavelength, but its correction is only trichromic). Similarly, the resolution of a barrel distorted image after correction is actually a lot lower than the "nominal" resolution of the sensor. This limits what you can photograph and where you can put the subject in the frame. Not a problem if you think the circle in the view finder is a target; but it is an issue if you like to compose your images.

The partial software fix in CA that is rendered, is back down to what you would get with a good quality kit lens.

The "undistorting" and cropping you get in the final jpeg is what you actually see in the viewfinder/LCD on the G1, so the composed image is also what is captured on the jpeg. e.g. On the G1, the 14mm lens is actually more like 12mm, but what you see in the viewfinder/LCD for composing is the "corrected" view as would be saved as a jpeg.

The sensor on the G1 is 13.1M pixel, while the final image is 12.1M pixel. This is what is advertised.

I assume (though this may not be correct) that the Olympus E-P1 does the same, since people are saying that the result shows not too much barrel distortion.

[Kyle Jones]

For me, Olympus' strength is in their optics. If they've given up on that, then they've eliminated the last reason I had to buy Olympus gear going forward. But I'm clinging to some hope that serious primes for µ4/3rds are coming, either from them or Panasonic/Leica. Particularly Leica.

[e_dawg]

Let's not get too worried... this is not only a pancake lens, it's also a kit lens. So it not only has to be tiny, it has to be cheap enough to still be profitable as a kit lens. If you give Olympus lens designers more size and product cost to work with, the image quality will be just fine.

[Alan Baxter]

Ok, fair enough if that's what Olympus said.... did they say that about micro 4/3 as well as 4/3???

Not to my knowledge, but they did say "SLR quality".



The partial software fix in CA that is rendered, is back down to what you would get with a good quality kit lens.

Actually, I think it is a little worse than that.


The "undistorting" and cropping you get in the final jpeg is what you actually see in the viewfinder/LCD on the G1, so the composed image is also what is captured on the jpeg. e.g. On the G1, the 14mm lens is actually more like 12mm, but what you see in the viewfinder/LCD for composing is the "corrected" view as would be saved as a jpeg.

The sensor on the G1 is 13.1M pixel, while the final image is 12.1M pixel. This is what is advertised.

I assume (though this may not be correct) that the Olympus E-P1 does the same, since people are saying that the result shows not too much barrel distortion.

I agree that the final image does not show much distortion, but it's not as simple as "hey, they still used 12.1Mpx", because yes, they might have, but the information density at the edges is way lower than that average resolution (and a little higher in the middle). And then the edges need extra sharpening, and the middle needs averaging, so the colour acuity goes to pot, and before you know it people will say they can't see what's supposed to be so good about Olympus optics - it looks no better than my old Canon Eos.

Regards,

Alan

[jebir]

Olympus has all the time said that the µ4/3 format is for those who want better image quality and flexibility than P&S but don't want the bulk of a dSLR. That is exactly what they have delivered.

I think they would be foolish if they didn't take advantage of all kinds of technology - like electronic corrections of aberrations - to realize this if it means smaller and more compact units with minor IQ reduction (instead of stubbornly adhering to solving everything with optics). The µ4/3 system is definitely capable of dSLR quality images , which also the reviewers of the 17/2.8 noted but it is not eccesarily the task of the µ4/3 kit lenses to enable it. If those lenses aren't delivering what you need, then there is still the 4/3 lenses (and bodies) available to us.

IMHO, there are too many in this thread seing the glass half empty rather than half full.

Cheers, Jens.

[tri-x]

The E-P1 ORF files I opened in dcraw produced TIFFs of 4096x3084 ( 12.63Mpixels ) compared to the out of camera jpgs and Master2.2 RAW size of 4032x3024 ( 12.19Mpixels ).

[acme]

The E-P1 ORF files I opened in dcraw produced TIFFs of 4096x3084 ( 12.63Mpixels ) compared to the out of camera jpgs and Master2.2 RAW size of 4032x3024 ( 12.19Mpixels ).

This is normal for many cameras. dcraw interpolates all the data the author could find whereas "supported" software packages usually mask the sensor's output slightly (to achieve a precise aspect ratio or fit a specified surface area).

[tri-x]

This is normal for many cameras. dcraw interpolates all the data the author could find whereas "supported" software packages usually mask the sensor's output slightly (to achieve a precise aspect ratio or fit a specified surface area).

To allow for less obvious cropping after barrel distortion correction :(

I've got a Zeiss ZM 18mm ready and waiting for the E-P1 to arrive.
If there is any corner smearing with it - then software corrected 17mm might still be the better lens at the end of the day.

[jebir]

To allow for less obvious cropping after barrel distortion correction :(
No, looking at the amount of distortion, those few extra rows of pixels are not going to make any appreciable difference.

Such extra borders of pixels are there for various calibrations for the camera hardware as well as the RAW converters, and have been present in all .ORF-files since the E-1, and also in the files from the Camedia cameras before that.

Cheers, Jens.

[e_dawg]

I've got a Zeiss ZM 18mm ready and waiting for the E-P1 to arrive.

That is a nice lens. A tad big for the E-P1, but you're going to love the way it operates (much nicer than using a focus by wire lens) and the colour of the Zeiss images. I love shooting with the ZF 25/2.8 on my G1. The only thing I don't like about it is the size and weight. (but the ZM 18 is half the length and 2/3 of the weight, so you should be okay)

[Knight Palm]

If Panasonic and Olympus can write software correction algorithms for their m4/3 lenses, what stops good software engineers to write similar algorithms for Leica M, Zeiss, Voigtländer, old OM lenses, and even further refine the Zuiko Digital (ZD SG, HG, SHG) lenses to further heights?

What is needed is a chip on these manual lenses to store the corrective parameters. As an idea, good russian programmers are more than welcome to explore this route, based on their current dandelion adapter knowledge.

[Kyle Jones]

The problem is that "correction" is an oversimplification. The reason you don't want lens distortion is that it destroys information. Once destroyed, you can't get that information back. For instance, correcting barrel distortion will fix the large scale problem (the bowed lines) but it won't put back the detail that the distortion squeezed away. In fact the correction will remove more detail during the averaging and interpolation pixels required to flatten the bowed image.

So with software correction you improve one aspect of the image at the expense of another. You're better off starting with good lenses.

[olddigiman]

Yah, I'm with Kyle on this one. Debarreling via pixel interpolation may be fine for the casual shooter, but I for one don't want to lose any more IQ than I have to. I am definitely less interested in the DigiPen for serious photography than I was before I saw that test. Probably just as well since I can't really afford it right now...:D

Having said that, I would be interested in seeing some comparisons between the 17mm pancake and something like the 12-60 or 14-54 at 17mm and around f4 and f5.6 to see how much IQ REALLY does get lost. Volunteers?

[Knight Palm]

I agree with jebirs opinion, that there are too many in this thread seing the glass half empty rather than half full.

There are a vast amount of examples in other areas of processing control, where you add a correcting signal in order to perfect the total output signal. Every mobil phone is using this technique, our audio power amplifiers as well, and more relevant our cameras already.

Even Olympus Studio are using these techniques today with their set of i.e. Aberration Compensation, Distortion Correction and Shading Compensation. Having these functions in the camera, as a vital part of the image processing pipeline will be transparent to and benefit to the user.

Image processing and video processing are already present and will steadily grow in importance.

[Alan Baxter]

I agree with jebirs opinion, that there are too many in this thread seing the glass half empty rather than half full.

There are a vast amount of examples in other areas of processing control, where you add a correcting signal in order to perfect the total output signal. Every mobil phone is using this technique, our audio power amplifiers as well, and more relevant our cameras already.

Even Olympus Studio are using these techniques today with their set of i.e. Aberration Compensation, Distortion Correction and Shading Compensation. Having these functions in the camera, as a vital part of the image processing pipeline will be transparent to and benefit to the user.

Image processing and video processing are already present and will steadily grow in importance.

Okay. But the camera in my mobile phone cost nothing, and has nothing expected of it. The camera and lenses I use to take photographs cost a lot and I expect a lot. SO: I don't use vignetting control, I use good lenses; I don't use distortion correction, I use good lenses; I don't use CA correction, I use good lenses; and I don't use Olympus studio.

Using the best lenses I can afford, I still run up against problems of resolution and image depth. These corrections worsen the image in a way that I can see, and that I don't like.

I used to use a Canon DSLR and swapped to Olympus because the poor quality of the lenses was compensated for by smoothing and then sharpening, resulting in a "paint by numbers" looking image that I don't like.

All sorts of things grow in importance. McDonalds has grown in importance. That doesn't mean they make a good burger.

Regards,

Alan

[Alan Baxter]

Yah, I'm with Kyle on this one. Debarreling via pixel interpolation may be fine for the casual shooter, but I for one don't want to lose any more IQ than I have to. I am definitely less interested in the DigiPen for serious photography than I was before I saw that test. Probably just as well since I can't really afford it right now...:D

Having said that, I would be interested in seeing some comparisons between the 17mm pancake and something like the 12-60 or 14-54 at 17mm and around f4 and f5.6 to see how much IQ REALLY does get lost. Volunteers?

Probably not a fair comparison (prime vs zoom), but I'd be interested too. Although I don't think it should be limited to just one aperture, or just centre frame. I quite like the DPR reviews, that plot resolution at each aperture against that of the sensor - their plots seem to relate well to my practical experience.

Regards,

Alan

[Swandy]

Having said that, I would be interested in seeing some comparisons between the 17mm pancake and something like the 12-60 or 14-54 at 17mm and around f4 and f5.6 to see how much IQ REALLY does get lost. Volunteers?

If I remember this thread when I finally get my hands on the 17mm I will try to do it. (I have the 14-54 MKII, not the 12-60.)
Steve

[Swandy]

Okay. But the camera in my mobile phone cost nothing, and has nothing expected of it. The camera and lenses I use to take photographs cost a lot and I expect a lot. SO: I don't use vignetting control, I use good lenses; I don't use distortion correction, I use good lenses; I don't use CA correction, I use good lenses; and I don't use Olympus studio.Alan

By definition "kit lenses" by just about every manufacturer are generally not considered "good lens" with regards to perfect IQ. I think we have been spoiled (I know I have) with the quality of the Olympus 4/3 kit lenses to this point and that is making some people (most of whom have not even used in real life situations the EP1) look negatively at the lenses Olympus has released to this point. I agree that better lenses will probably follow and that Olympus original intention was to release "kit" lenses of reasonable quality, SIZE and COST. But how much bigger/heavier/costlier they will be has yet to be seen.
Steve

[finemom]

Mine has shipped and now I don't know what to do. I was never even sure if I wanted the lens but I got a good price on Amazon ($229) so I kept the order in. I'm not even happy about switching lenses on the EP-1 so I have the panny 14-140 coming sometime this month. I love the camera and leave the flash attached at all times so for me it's like a built in flash.

[olddigiman]

Yah, I'm with Kyle on this one. Debarreling via pixel interpolation may be fine for the casual shooter, but I for one don't want to lose any more IQ than I have to. I am definitely less interested in the DigiPen for serious photography than I was before I saw that test. Probably just as well since I can't really afford it right now...:D

Having said that, I would be interested in seeing some comparisons between the 17mm pancake and something like the 12-60 or 14-54 at 17mm and around f4 and f5.6 to see how much IQ REALLY does get lost. Volunteers?

Here is are some examples from the DPR test of the Canon 24mmTS that show how much IQ can be lost by software correction of distortion.

http://www.dpreview.com/lensreviews/canon_24_3p5_tse_c10/page5.asp

[e_dawg]

Just got the E-P1 + 17/2.8. First impressions of the 17/2.8?

1. It's a fairly average lens, optically speaking... pretty much as expected given all the reviews out there. I'm not even going to bother testing it head-to-head with the 14-54 or 12-60, as it's pretty clear it's not in the same league (especially wide-open). The usual Olympus 4/3 lens characteristic of remaining sharp in the corners and with good contrast wide-open does not apply to this lens, and I doubt it will with any m4/3 pancake in the future.

2. Having said that, this lens is quite enjoyable to use. It becomes fairly good when you stop it down to f/4.5+, and the CDAF is reasonably quick on it (have not noticed any AF speed or locking problems that have been reported in the media). Best of all, it is small and light -- far smaller and lighter than almost any other option for m4/3.

3. The bokeh is surprisingly nice at wide apertures. This is a nice lens to shoot B&W portraits with -- not only for the bokeh, but the B&W effectively masks the CA from this lens, which can be noticeable in colour shots. The lens remains nice and sharp near its minimum focusing distance, but due to its wide-angle effects on perspective and volume distortion, you will have to be careful about how your subjects' features look.

4. Adapted choices to the 17/2.8 in terms of size & weight include the Voigtlander pancakes when used with a Leica M to m4/3 adapter. But the CV Nokton 35/1.4 and 40/1.4 have rough and edgy bokeh compared to the 17/2.8, and the 15/4.5, 21/4, and 25/4 are slower than the 17/2.8.

5. Other adapted lenses include the Nikon 20/3.5 AiS and CV Color-Skopar 20/3.5 SLII in Nikon mounts. The Nikon 20/3.5 is similar to the 17/2.8 in that they both need to be stopped down to restore corner sharpness and contrast across the frame, are sharp at close-range, and have nice bokeh that's smoother than their competitors. The Color-Skopar 20/3.5 is a bit sharper in the corners wide-open, but produces rougher and edgier bokeh in some situations.