Jack Flesher wrote:
Sigh....... I think they could have easily put the thicker IR cut filter in front of their sensor in the first place! On the off chance it actually did cause sensor CA (which I doubt it would), that could be easily fixed with firmware and/or coding for raw conversion. But obviously others disagree with me
Cheers,
The M8 is already thicker than the typical M Series film body but I really would have not mind Leica adding something like 1 or 2 mm extra to eliminate or at least minimize the problem. If push comes to shove they could have extended the sensor area backwards by that distance, meaning let the LCD display stick out some more....
ClubShooter wrote:
Have you actually tried it? The overlapping portion looks the same because they both cut the same portion of the spectrum. It's a great way to illustrate the difference between CC and spectral cut filters, with CC the density doubles, with cut filters there's no difference how many times the same portion of the spectrum is removed.
OK, Try it with a lightmeter, does the exposure increase with two #25 filters as opposed to just one? The eye may be too accommodating to see the difference.
Tom
The original Leitz finders with wide coverage are collectable and, therefore, expensive ($300 and up). They are also limited to 35, 28 and 21 FL. The metal construction is very nice - no broken feet - and fit snugly in a Leica shoe. The 50, 90 and 135 finders are very affordable, and give a 1x view (wow).
More recent Leica finders are plastic and not well regarded: manufacture was clearly farmed out. Better to get the Cosina/Voigtlaeder versions, which are available in metal build for 28, 35, 50, 75, and 90 FL. The 28, in particular, gets rave reviews.
Question about using IR filters on lenses for the M8 to correct for the Magenta problem -
On BetterLight scanbacks a glass IR filter is required, they come in thin (1/16") and thick (1/8") versions for different types of lighting. Several years ago when I was converting Opti-Copy process cameras to digital by mounting a BetterLight scanback in place of the film holder we used to mount the IR filter on the rear of the lens. As these cameras are programable to move the lensboard in increments of .0001", I noted the focus was different when the filter was in place as opposed to when the filter was not installed (aerial image, as the scanback is unusable without the filter). I can see the same effect when focusing on the groundglass of my Plaubel as I also mount the filter on the rear element of the lens.
I seem to recall reading somewhere that a glass filter mounted on a lens will change the focus by approximately the thickness of the filter on the side of the lens the filter is mounted on, and my tests on the Opti-Copy cameras seemed to confirm this. The lens we used on the conversion was a 240mm Schneider Componon-S, and reduction ratios were from 4x to 16x. The lens I mostly use on my Plaubel is a 210mm Apo-El-Nikkor, reproduction ratios are similar.
The only other filter I use is a polarizing filter, and I always focus with the filter in place on both the BetterLight and my 5D. Of course the polarizing filter is mounted on the front of the lens in both cases.
The point and question I am getting at is that if Leica had decided to retrofit the M8 with an IR filter mounted between the lens and sensor, this would have changed the backfocus distance and therefore required the sensor to be mounted in a different plane (slightly longer distance from the flange) to account for this. Big $$$ and Big engineering changes to accomplish this. Wisely (for a temporary fix) Leica decided on IR filters on the front of the lens, which when focused at infinity or moderate distances should have no effect on focus.
Just curious if anyone has noticed if critical focus is changed on close-ups shot with the M8 when using the IR filter as opposed to having no filter? Being a Rangefinder there is no way to check critical focus, you depend on the cam on the lens and the rangefinder. I know the Leica M is not designed for close-up work, I always used a Visoflex and 105mm Apo-El-Nikkor for close-up and copy work with my M4p. I assume at focus distances used for 99.9% of work done with the Leica M8 the filter on the front of the lens will have no effect, but wanted to see if anyone had noticed any effect when shooting the M8 close up without a Visoflex.
By the way, I'm downsizing my M8 lens collection. Already parted with the Nocti. The hardest one really to step away from. Now it's the turn of the 21 Elmarit, 3885... serial number. The shot is of the lens. It is in like new condition. I'm going to put the lens up on the for sale page sometime this week, but thought I'd give people here a heads up -- in the Guy tradition.
The 39 mp backs have about the same pixel pitch as the Canon 20D/30D and 350D.
See http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm for a discussion of how diffraction affects sensors of various pixel pitch. The assertions made at that web site are corroborated by lens tests at http://www.slrgear.com/reviews/index.php where a number of Canon lenses are tested on both the 20D and the full frame 5D. The lenses all show a distinct hit in performance in going from f/11 to f/16 on the 20D while the same lens does not take the same hit on the 5D until going fromf/16 to f/22 or sometimes beyond. So the 20D is diffraction limited at or about f11 and by implication so are the 39 mp backs.
A front mounted filter on a lens focussed on infinity won't affect focus distance because light from infinity is effectively collimated, so no lateral refraction occurs on the two filter surfaces (front and back). That situation is different with a filter mounted behind the lens. Collimated rays from infinity become convergent rays (ideally converging at the film plane). As such, self-cancelling refraction in the rear mounted filter will alter the path length, not necessarily by the thickness of the filter, but by some small amount deriving from the refractive index of the filter material. Thus a rear mounted filter will slightly alter correct focus. Infinity will be correctly focussed when the lens is set to focus a bit nearer.
In brief, the effect that makes rear mounted filters shift focus is the same effect that limits sensor filter thickness in the M8, since reflection within the filter also increases tangentially with angle of incidence. Distance of rear filter from lens makes no difference.
Edited as OT
Edited by brainiac on Dec 19, 2006 at 11:56 AM GMT
Can't remember if it was Jack or Mike who posted those two DNG files, but thanks a lot, it is appreciated. Have downloaded them but haven't had a chance to convert them yet. That's not my expertise.
Of course, they don't tell us much about what a 5D with top glass would have recorded in the same situation, and as we all know, shooting conditions make it tricky to compare two cameras by looking at different shots taken on different occasions.
To whoever said prints are the only way to assess images, I say we have been over this before. Fair comparison on a good monitor can tell you as much as a print, and numerical pixel values can help in that assessment.
Jack Flesher wrote:
Sigh....... I think they could have easily put the thicker IR cut filter in front of their sensor in the first place! On the off chance it actually did cause sensor CA (which I doubt it would), that could be easily fixed with firmware and/or coding for raw conversion. But obviously others disagree with me
Cheers,
Jack, I'm surprised you don't think they spent a lot of time on these options. These are smart guys. They knew there was a problem - and this was a problem without a solution in-camera (in their shipping time frame). Very easy to bench test this stuff. They took a calculated risk. They probably hoped the IR issue would be received as an occasional problem, similar to no aa filter. But the internet has caused most people with a new M8 to go looking for the issue.
The solution doesn't seem so bad. And the cost to Leica isn't much for two filters in volume.
In hindsight they should have talked about the IR issue and their solution before shipping.
Have looked at those M8 DNG files now. They look pretty much like a well exposed fine JPEG from a 5D with unsharp mask (100%, .5 pixel, 0 threshold), except that they are free from the slight ghost lining effect that JPEG/sharpening introduces around high contrast edges. I still suspect that the M8 has some kind of built in noise reduction. Either that or moiré is introducing a slightly artificial texture inside the resolution threshold:
Cyan vignetting in both as expected. It's impossible to know whether the 5D raw files with a good lens and appropriate sharpening would have resolved more detail here, but at a guess I suspect it would. I see nothing here that contradicts my resolution test. Obviously if we had these pictures taken using raw on both cameras with the same angle of view we could have a better idea of whether the M8 can make up for its shortfall in pixels.
I should add that legibility of small diagonal text at the focal distance is a very good measure of actual resolution (by which I mean resolution of information within a particular scene/frame boundary, not Pondria's meaning). The reason I say this is that acutance is quite a misleading characteristic: it can easily be generated with sharpening. The unsharp mask filter means that the acutance of a camera is more or less irrelevant. The Leica's have good acutance, which is seductive, but it is no substitute for superior resolving power, which, as far as we can see so far, the Canons have.
It is pretty clear from looking at my test and these files that over most of the frame, in common situations, our lenses are still outresolving our sensors, even non-Leica ones. That's good news for the future as it means there is room for better higher megapixel cameras using these lenses.
How many percent magnification was that crop you posted? I still don't understand why you show crops into the 400-600 percent magnification range.
BTW, on the tree shot, I focused on the tree, on the graveyard shot I focused on the tombstone in the center of the frame. Even with a 28mm stopped down, the Leica digitals are always sharper in the area you focus. In the case of the tombstone the focus was just shy of infinity.
Don't forget that these were also shot with a 20 year old 28mm lens which has had two generations of lenses made after it. I wan't using a tripod either.
http://www.robsteve.com/FM/L1040403.jpg
I think this is the image Richards crop was out of.
http://www.robsteve.com/FM/L1040412.jpg
Richard is correct that the sharpness can cause Moire, but in this case, turning up the color noise slider in lightroom dealt with most of it.
With some color artifacting on the fine blades of grass
http://www.robsteve.com/FM/L1040412-3.jpg
Color Nise slider moved up to do more correction.
http://www.robsteve.com/FM/L1040412-2.jpg
Edited by robsteve on Dec 19, 2006 at 11:05 AM GMT
brainiac wrote:
A front mounted filter on a lens focussed on infinity won't affect focus distance because light from infinity is effectively collimated, so no lateral refraction occurs on the two filter surfaces (front and back). That situation is different with a filter mounted behind the lens. Collimated rays from infinity become convergent rays (ideally converging at the film plane). As such, self-cancelling refraction in the rear mounted filter will alter the path length, not necessarily by the thickness of the filter, but by some small amount deriving from the refractive index of the filter material. Thus a rear mounted filter will slightly alter correct focus. Infinity will be correctly focussed when the lens is set to focus a bit nearer.
In brief, the effect that makes rear mounted filters shift focus is the same effect that limits sensor filter thickness in the M8, since reflection within the filter also increases tangentially with angle of incidence. Distance of rear filter from lens makes no difference.
Edited as OT
Edited by brainiac on Dec 19, 2006 at 11:56 AM GMT...Show more →
I interpret this to mean that a thick uniform absorption filter on the sensor would introduce not just reflections but also focus shifts with angle of incidence, so all else being equal the image would look progressively out of focus towards the edges even though the focus point was correct - am I right with this?
Guy - please don't be impatient with me. These things are just not obvious to me. I did read the DMR thread at the time. I didn't find it as conclusive as you did. One thing that bothered me was that I never felt that the issue of resolution was satisfactorily addressed. But here we are talking about two different cameras, mainly. I am not sure that the induction from 1DsII v DMR to 5D v M8 necessarily holds. The 5D may be better than the 1DsII in some respects, perhaps noise and shadow retrieval. AA filter strength is a variable that may have changed, for instance. That's why this comparison still interests me.
If I get a chance I will try to do a raw version of my previous test soon. I'll use f11 and bracket in tiny increments from one end of the focus range to the other with the M8. I'll post the DNG files so that anyone is welcome to do their magic on them.
To be fair, Richard so far is the only person to present an actual comparison on these pages -- regardless of whether it was flawed in some way or not consistent with the way some people work. Beyond the claims and assertions from "many" along these 200+ pages, no one has produced an actual comparison that provides hard evidence, not to mention one that refutes the conclusions of Richard's test. So far (putting my analytical cap on), all we've seen is anedotal information that indicates that the M8 is indeed "very good", but really doesn't indicate that it is indeed "better". (To throw a useless opinion into this based on all the anecdotal information and tests I've seen (here and elsewhere), it seems to me that the M8 and 5D are "comparable" with each having different strengths. It would be ideal to have both.
Guy Mancuso wrote:
Richard obviously the DMR thread means nothing to you as you keep finding a way that the Canons are better when clearly many disagree. i give up with this not worth 200 more pages to give you accurate information
nik chmiel wrote:
I interpret this to mean that a thick uniform absorption filter on the sensor would introduce not just reflections but also focus shifts with angle of incidence, so all else being equal the image would look progressively out of focus towards the edges even though the focus point was correct - am I right with this?
I don't think that is right Nik, no. The parallel faces of the filter introduce no variable focus shift across the frame, they simply lengthen the path by a small but exactly matching amount across the whole frame. That's because the refraction as the light exits the filter exactly undoes the bend that happened as the light came in.
