RAW Truth

[ChrisJ]

I was interested to see just how much distortion there is in the kit 14 - 44mm lens (I don't have the 17mm f2.8).

All were shot in Raw with a tripod the tripod was in exactly the same position during the shots.

First up is a shot with my usual workflow Photoshop CS4 Raw Converter (with op codes). Reduced to 800px and saved as a Jpeg.

The the same file opened in Photomatix Pro (no op codes) the downside is that Photomatix treats it as a single shot HDR, Reduced to 800px and saved as a Jpeg.

Finally a Pentax Auto 110 18mm lens opened in Photomatix Pro exactly the same procedure as the Panasonic lens shot above.

This, to me begs the question - if Pentax in the late 1970's can make an 18mm lens without distortion then why can't Panasonic produce a 17mm lens without distortion for a reasonable price in 2009?

Chris

 

[pellicle]

Chris

two thoughts ... the pentax 110 wasn't cheap then and its not a zoom

 

[pellicle]

Hi

If they come across a disk full of GF-1 raw files it's almost inevitable that they won't be able to.

then include a copy of dcraw.exe in there and you'll be right... With so many platforms moving towards hypervisor type environments and downloadable versions of DOS or even Win 3.1 I think you'll be able to execute an x86 bit of code for several decades to come.

The solution for the little micro 4/3 cameras (I'll reiterate that it doesn't affect ordinary 4/3) is for the corrections to be baked into the RAW files (just like they are in the jpgs) NOT added as instructions to the raw processor. Either that or to use proper DNG files

firstly I agree with the preference for storage of the opcodes as metadata not as changes to your RAW ... surely it would cease to become RAW then?

does anyone else except Leica mung (and essentially irrevocably alter) their RAW by doing this? As far as I know DNG is just a 'container' file not unlike TIFF and that the RAW data is essentially stored there anyway ... only the headers and arrangement of tagged fields is altered.

no?

Perhaps I'm speaking too much like an archival librarian here but we go to great lengths here for making sure that our data and meta data are preserved, and we're thinking in hundreds of years here in our planning at least.

 

[ChrisJ]

Chris

two thoughts ... the pentax 110 wasn't cheap then and its not a zoom

The 1983 price of the 18mm Pentax Auto 110 lens was £20.00, this was gleaned from a reprint of an old review in Camera magazine a copy for download here.

http://www.pentax110.co.uk/html/brochures.html

its the 1983 review which has the prices.

using this inflation calculator

http://www.thisismoney.co.uk/historic-inflation-calculator

Makes it £52.00 in todays terms, not what I would call expensive.

True the kit lens is a zoom, but people are reporting the same with the new 17mm f2.8 a directly comparitive lens(Unfortunately I don't have one to test). I realise the 17mm f2.8 has auto focus and an aperture system so lets triple it to £156.00 that's still just over half of what they are selling the new lens at.

The point remains if Pentax could produce an 18mm lens in 1983 with no distortion for a cheap price, why can't panasonic do it now without having to resort to deforming the image in software? Or are they just using technology to sell cheap lenses at expensive prices?

The older I get the more cynical LOL.



Chris

 

[kwalsh]

I didn't think it did. the older version produced the huge, horrible, demosaiced linear dng files - I thought the next version produced ordinary DNG files, but without the correction?

Maybe I'm wrong - I'd be interested to know, as it definitely does change things.

Well it certainly does include the correction metadata for my G1 files, I'd be surprised if it didn't for the EP-1. Of course the next question is does your RAW converter actually then apply the corrections specified in the DNG metadata. LR/ACR certainly do but perhaps others don't. You might search for threads on DPR by Andy Westlake, he's done a lot of tests with m43 corrections and you might find an answer in his posts.

Why? Leica are sorting out the vignetting and colour shift with IR for the M8 and M9, and these are written into the RAW file without doing the demosaicing. Or am I missing something here?

So the difference here is that color shifts and vignette don't require demosaicing to apply. They don't use neighboring pixels or change the shape of the image, just the values of individual pixels. Lens distortion and CA correction do change the image shape and neighboring pixels so you have to demosaic to apply the correction.

I understand why the manufacturers don't like it, but it can't be necessary to make a new file format for every camera (surely).

Yeah, that's why I said "more" and not "all". It seems like 90% or more could do DNG straight from the camera and DNG already supports maker-notes and all sorts of ways for manufacturers to innovate. Heck, if they really came up with something new that DNG didn't support they could use proprietary in that case until DNG caught up and then do a firmware update. Or modify DNG themselves as a propsal. The proprietary non-sense is getting old!

So, the question is:
Do the lens correction opcodes go into DNG files converted from raw?

As far as I can tell the answer is yes. Whether your RAW converter of choice actually uses them yet is another question!

Happy shooting!

Ken

 

[kwalsh]

The point remains if Pentax could produce an 18mm lens in 1983 with no distortion for a cheap price, why can't panasonic do it now without having to resort to deforming the image in software? Or are they just using technology to sell cheap lenses at expensive prices?

Well, to play devil's advocate I wonder what the CA and field distortion of that lens are. The theory behind software corrections is that you remove constraints from the lens that are easy to correct in software (distortion, vignetting and certain kinds of CA) and thus get more freedom to optimize other parameters than can only be done optically (sharpness, field distortion, bokeh, etc.). It isn't necessarily always about making it cheaper. Now, whether that is actually what Panasonic is doing is a whole different question!

The older I get the more cynical LOL.

Some call this "experience".

Ken

 

[Terry]

The point remains if Pentax could produce an 18mm lens in 1983 with no distortion for a cheap price, why can't panasonic do it now without having to resort to deforming the image in software? Or are they just using technology to sell cheap lenses at expensive prices?

The older I get the more cynical LOL.



Chris

But isn't this part of the problem that Leica had in making a rangefinder. Light hitting film is different than light hitting a sensor and m4/3 similarly to a rangefinder has a very short distance from the back of the lens to the sensor. Leica partially fixed it with offset micro lenses and a crop sensor. It was another few years before they we able to get from crop sensor to full frame.
I am not a physicist but I would guess you have two choices have a sensor based solution or a combined lens/software solution.

 

[ChrisJ]

Well, to play devil's advocate I wonder what the CA and field distortion of that lens are.

The theory behind software corrections is that you remove constraints from the lens that are easy to correct in software (distortion, vignetting and certain kinds of CA) and thus get more freedom to optimize other parameters than can only be done optically (sharpness, field distortion, bokeh, etc.). It isn't necessarily always about making it cheaper. Now, whether that is actually what Panasonic is doing is a whole different question!quote

Some call this "experience".

Ken

The idea behind showing the Pentax lens picture against the Panasonic lens picture was to demonstrate the Pentax lens has no (or very little) distortion, both lenses were set at 18mm and I've yet to see any CA from any of the Pentax 110 lenses and I've tried to force it by shooting against dark trees a bright sky for instance and even with pixel peeping I can see none. Though I can see some vignetting in the 18mm.

I'm not against software correction, how can I be I use Photoshop every day, but my main point was my last one, we are getting optically poor quality lenses that need all these corrections, but we are being charged top dollar for them.

Chris

 

[Terry]

The idea behind showing the Pentax lens picture against the Panasonic lens picture was to demonstrate the Pentax lens has no (or very little) distortion, both lenses were set at 18mm and I've yet to see any CA from any of the Pentax 110 lenses and I've tried to force it by shooting against dark trees a bright sky for instance and even with pixel peeping I can see none. Though I can see some vignetting in the 18mm.

I'm not against software correction, how can I be I use Photoshop every day, but my main point was my last one, we are getting optically poor quality lenses that need all these corrections, but we are being charged top dollar for them.

Chris

When you look at the shot bigger what's happening in the corners? That has generally been the problem with these wider lenses. For instance my 28 cron and 12mm voigtlander were both problematic in the corners, something I didn't have an issue with on the M8.

 

[ChrisJ]

But isn't this part of the problem that Leica had in making a rangefinder. Light hitting film is different than light hitting a sensor and m4/3 similarly to a rangefinder has a very short distance from the back of the lens to the sensor. Leica partially fixed it with offset micro lenses and a crop sensor. It was another few years before they we able to get from crop sensor to full frame.
I am not a physicist but I would guess you have two choices have a sensor based solution or a combined lens/software solution.

It's true Terry, the Pentax 110 lenses would not work well on a full frame sensor, but here we're talking about the G1 with a half frame sized sensor which is easier to design lenses for than a full frame.

I've lived through this debate before in the 1950's when the magazines were saying how poor IQ is with 35mm compared to 120 film. In the 1950's it was probably true, by the 1980's less so.

Panasonic/Olympus and even Pentax have put their faith in smaller sensors ultimately performing as well as full frame, just as in the 1960's and '70's it was easier to design lenses for 35mm than for the 645 (120 film) format which was (is) reflected in the price of lenses for the two formats.

Why Leica is going the full frame route only Leica can answer, as the trial between a G1 and a Canon full frame sensor in Amateur Photographer shows with every generation small sensors are getting better and better, and now that Fuji is joining the m43 bandwagon with their excellent Super CCD SR sensor (which apparently increases dynamic range) we may not need a full frame sensor at all.

Sensors are developing at a much faster rate than film ever did, there are some technologies that have'nt even been tried yet, it won't take 30 years this time for major improvements - you heard it here first .

Chris

 

[Terry]

It's true Terry, the Pentax 110 lenses would not work well on a full frame sensor, but here we're talking about the G1 with a half frame sized sensor which is easier to design lenses for than a full frame.

Chris

But, I'm not really talking about the lens working on a full frame sensor. If we just take the 28 summicron as an example.

It works just fine on the M9 and M8 but it doesn't work well on the smaller 4/3 sensor (where you would think that it should be the best because you are using the sweet center spot of the lens).

The M9/M8 sensors are specially designed to deal with wide angles and how the light hits the sensor. Not sure what Panny/Oly have done to attack the problem.

 

[Robert Campbell]

...but here we're talking about the G1 with a half frame sized sensor..

While the linear measurements for the 4/3rds sensors are about half those of 35mm, the sensor area of 4/3rds is about a quarter of 35mm.

 

[kevinparis]

I was oddly surprised to realise that the 4/3 sensor was pretty much the same as the old 110 Kodak format.

the whole need to get the light hitting the sensor straight on is the biggest bugbear of the smaller sensor formats. That coupled with the marketing promise of the 4/3 format of smaller has meant that the manufacturers have had to make decisions to go small on lenses and correct later rather than make lenses that would give the numbers in terms of focal length and aperture but be large and unwieldy... and not 'on message' marketing wise

Just look at the panleica 25/1.4 and the zuiko 25/2.8.... which one fulfills the promise of smaller?

Olympus know their optics... the OM lenses were great quality and very petite. Do you think if they could do the same for 4/3 they wouldn't?

There are big rules of physics here that most of us dont understand... a digital sensor is very different from film. Getting more light at a wider angle and still having a useable image is a conundrum.

maybe digital correction is the future... hell it was good enough for the Hubble telescope.... and that focuses closer to infinity in the dark better that we ever will




K

 

[Vivek]

Panasonic/Olympus and even Pentax have put their faith in smaller sensors ultimately performing as well as full frame, ..

Why Leica is going the full frame route only Leica can answer, as the trial between a G1 and a Canon full frame sensor in Amateur Photographer shows with every generation small sensors are getting better and better, and now that Fuji is joining the m43 bandwagon with their excellent Super CCD SR sensor (which apparently increases dynamic range) we may not need a full frame sensor at all.

Sensors are developing at a much faster rate than film ever did, there are some technologies that have'nt even been tried yet, it won't take 30 years this time for major improvements - you heard it here first .

Chris

I have little faith in the smaller 4/3rds sensors even catching up with the slightly larger APS-C sized sensors in terms of performance (Noise and DR, in particular).

IIRC, there has been only one major upgrade in the m4/3rds NMOS sensors (I will leave the GH-1's sensor out). There has been tremendous progress in the AA/UV/IR cut filters that are in front of the sensor, both in size and in performance. Astounding piece of engineering there.

The very reason that Oly and Pana are sticking with the m4/3rds NMOS sensors (after Kodak terminated sensors to Oly) is the cost. NMOS sensors are cheap. The ones in the camera are on a flex board unlike ceramic cased CCD or CMOS sensors (of any size), for example. I still can't believe that this is the piece that does the image recording.

When a Sony EXMOR-R sensor (~ APS-C size) camera (I hope they will not waste it on any of their mirror reflex cameras) hits the market, things will be shaken up quite a bit. I am keenly looking forward to that. Even with the current technology, if Sony were to offer a mirrorless camera instead of the A500, it would be simply so much better than any of the m4/3rds cameras for still photography.

On the Pentax-110 lenses, yes it is true that the 18mm has much less distortion or CA than the Oly-D 17mm lens, I reported on this a while back.

OTOH, the Pentax 18/2.8 lens has no changeable iris, is a pain when it comes to handling and shows much larger light fall off. The corner sharpness of the Pentax-110 18mm is much less desirable compared to the Oly-D 17mm.

Cost wise, a simple m4/3rds to Pentax-110 adapter would set one back by ~ $110. This can not be taken for granted while comparing the lenses.

TTL flash- sucks with manual focus lenses. Unreliable.

With the 17/2.8, this works seamlessly and reliably.

 

[pellicle]

Hi

I have little faith in the smaller 4/3rds sensors even catching up with the slightly larger APS-C sized sensors in terms of performance (Noise and DR, in particular).

based on what I see with my G1 and comparing with files from my 10D and 20D (both sold now) I would be interested to see what it is you are talking about here.

On DPReview I see little to distinguish them, save the difference expected in the smaller area (some 15%). It is interesting to compare 'cross camera' reviews and take cameras which are contrasted against the G1 in one review (such as the 450D in both the G1 and 500D tests)

right up to 800 iso everything is pretty comparable.

however if you compare to the 5d MkII then it sucks ... but then that's not APS-C as you mentioned

IIRC, there has been only one major upgrade in the m4/3rds NMOS sensors (I will leave the GH-1's sensor out). There has been tremendous progress in the AA/UV/IR cut filters that are in front of the sensor, both in size and in performance. Astounding piece of engineering there.

The very reason that Oly and Pana are sticking with the m4/3rds NMOS sensors (after Kodak terminated sensors to Oly) is the cost. NMOS sensors are cheap. The ones in the camera are on a flex board unlike ceramic cased CCD or CMOS sensors (of any size), for example. I still can't believe that this is the piece that does the image recording.

if the cost is lower and that is passed on to me then I don't have a problem with that ... as to the flexibility 1) there isn't really such a large area as to promote much movement at the focal plane (like bulge) and 2) is it any more flexible than film? Seems like a red-herring point but if you can justify it I'm interested to learn.

TTL flash- sucks with manual focus lenses. Unreliable.

that has not been my experience ... it has been flawless. However IF you exceed the ability of the compensation you'll get problems. For instance if you are too close and using a too high ISO you'll get over exposure as it seems that the flash's minimum power setting is not low enough. conversely if you are too far away you may exceed the flashes max output.

I have no way to measure what the EV range of the flash is for its min or max outputs (be interesting to know) as my flash meter is in another part of the world.

anyone got one?

 

[Vivek]

that has not been my experience ... it has been flawless. However IF you exceed the ability of the compensation you'll get problems. For instance if you are too close and using a too high ISO you'll get over exposure as it seems that the flash's minimum power setting is not low enough. conversely if you are too far away you may exceed the flashes max output.

Are you assuming that I use the pop-up and said that? I do know how a flash works, what a GN is and such tech details.

Your comparison's from Canon and Dp revs shows everything is comparable up to ISO800, well and good.

On the film plane- A lot of people paid a premium price (Leica M bodies) for a good reason. Film flatness is taken care of there.

Those who were very aware of the flatness issue also allowed time for the film to uncurl before the next snap.

OTOH, there were gazillions who used film without knowing that at all and may continue to use it in the very same fashion.

 

[pellicle]

Hi

Are you assuming that I use the pop-up and said that?

no ... but its the only TTL flash I've ever used on my G1 ... my Metz is not TTL

I do know how a flash works, what a GN is and such tech details.

I expected you did ... I didn't question that you did ... I only asked about your experience with its unreliablity and raised what I thought could be related parameters which may have effected the results.

the only assumption I made was that you were honestly reporting a problem and that you hadn't found a solution to it

On the film plane- A lot of people paid a premium price (Leica M bodies) for a good reason. Film flatness is taken care of there.

so too for almost all 35mm cameras I can think of ... espeically given how loosely wound the film is in the canister, how straight the film path is on every 35mm camera I've owned and how well the guide rails and pressure plate work. I've only ever seen 'curl' reported in post development. 120 is a different problem.

But (since you've pulled one apart) how do you see flex being an issue on 4/3 cameras? Since you raised it...

I use quite a bit of film ... 120 and 4x5 ... flatness is occasionally an issue for me in my 6x12 back but not in sheets. I've never seen anything on my G1 to suggest that sensor flatness is any issue ... have you?

 

[Vivek]

Pellicle,

Yes, it is a problem. Especially in very low light (within a given flash's capabilities). One frame is totally blown while the next is very dark.

The i-TTL flash works taking into account the exit pupil of the lens, the focal length, the distance and such factors. I am yet to make it work reliably for any one of my manual focus lenses.

On the flex board- No, I do not find that to be an issue affecting the image quality!

Full credit to Panasonic (the ones who make the NMOS sensors for m4/3rds and 4/3rds) to have achieved that! If they could get the flatness required with cheaper fabrication, when the QC is so high, I can only applaud them!

Yeah, the film curl issue is different for 120 which is different than 220 and is yet again different when it comes to larger format.

Since Pentax-110 is mentioned. Despite the tiny frame and use of a cartridge, that lousy camera (and the 110s) could barley hold a frame in the correct position. Just because something is old does not always mean it has be good. It was a crappy camera although, arguably, one of the cutest ever made.

 

[ChrisJ]

But, I'm not really talking about the lens working on a full frame sensor. If we just take the 28 summicron as an example.

It works just fine on the M9 and M8 but it doesn't work well on the smaller 4/3 sensor (where you would think that it should be the best because you are using the sweet center spot of the lens).

The M9/M8 sensors are specially designed to deal with wide angles and how the light hits the sensor. Not sure what Panny/Oly have done to attack the problem.

Sorry for being late to reply Terry.

That's the reason for my excitement at what new sensors can offer, the new Fuji sensors have two light sensitive areas for every pixel just to address this problem, it also has the advantage of widening the dynamic range more than a tad. It's just a question of upscaling it to a m4/3 sized sensor.

Chris

 

[ChrisJ]

Since Pentax-110 is mentioned. Despite the tiny frame and use of a cartridge, that lousy camera (and the 110s) could barley hold a frame in the correct position. Just because something is old does not always mean it has be good. It was a crappy camera although, arguably, one of the cutest ever made.

True Vivek the cameras were not good, only full program mode available etc. etc., but most of it's problems were with the 110 cartridge, definately not the lenses which were reported at the time to being as good as Leica or Zeiss.

Not for the first time Pentax backed a looser, they expected the 110 format to improve - it didn't.

Chris