Um, Large Format? Advice please.

I love my Canon 5D. I really do. It shoots weddings as well as landscape and gives me a huge amount of resolution while being a SLR and all the convenience that is included.

Thing is though that although the 5D will give me a perfect 18X12" @ 250DPI, for landscape work (portrait can go as big as I like) where there is endless detail that needs to be resolved, any uprezzing just serves to show the lack of fine detail. Now for some landscapes it isn't a problem, I keep that in mind when shooting, the image below is a great example of an image that makes a beautiful 20X10" due to all the detail necessary being resolved.


However, IMO, for large prints, and I mean large, (40X50" at a very minimum) the detail just doesn't hold up. When you are selling prints that size, as I intend to, for people to hang in their homes, then 'viewing distance' just doesn't apply, people are looking at the print from right in front of it, not 2 meters back.

I want to shoot working on a specific genre of photo for sale as Fine Art prints to tourists and for exhibition. I want to shoot B&W (only) and print large. Large enough and with enough detail that you could almost 'walk into the picture', that you could stare at it time and time again seeing the smallest details.

I used to shoot 645 and top scans from the latest imacon at the highest resolution show me that it has about a 1.5 megapixel advantage over my 5D. No doubt 6X7 would be more so, however I'm looking for something a lot more than that, the 'wow' factor perhaps.

I've never been a snob about equipment or particularly married to any means to an end in photography. I have a concept in my head of what I want the end product to be and I'll use the appropriate tool to get there. Does it look like I'm barking up the LF tree?

Many Thanks.

Well Nicolas prints that large and larger with files from his 1DsII.

I have a 1DII and a 5D but now am looking a LF for detail-rich scenics and interiors.

So I have been working with Will Thompson. He's shoots 8x10 on a Calumet C1 with Polaroid B&W or Color film. I'm returning to 4x5 with my Crown Graffic with a 130mm and a 203mm lens.

I may end up with an 8x10 too. I don't need many images, just very good ones. My idea is to use the Eos camera as a lightmeter, check with a 4x5 film on a 4x5 reducing back then pop in my color 8x10 Polaroid or Portra and Bob's your uncle.

I really can't afford what Id like a 39MP digital back. That, anyway will be outdated. The film camera is inexpensive, but bulky and needs film that needs processing. If one does not require thousands of shots, film seems to be a great option for detail rich work.

Anyway, that is where I'm heading.

Asher

It's as you say Asher, I'll be working in B&W in a semi urban enviroment shooting a lot of semi architectural scenery. The scene won't move any more than it has in the past hundreds of years and the light is pretty much the same year in year out. I don't need to bracket hundreds of shots or wait for the perfect colour of lighting. What I do need is so much detail that you could almost walk into the picture.

I've been working these scenes at night until now for a 'different feel' but now I want them in B&W and big. Very big. And with every blade of grass and texture in the foliage and stone resolved, not a mush in the distance.

Ben Rubinstein said:

I'll be working in B&W in a semi urban enviroment shooting a lot of semi architectural scenery. The scene won't move any more than it has in the past hundreds of years and the light is pretty much the same year in year out. I don't need to bracket hundreds of shots or wait for the perfect colour of lighting. What I do need is so much detail that you could almost walk into the picture.

Click to expand...

Sounds like the perfect scenario for stitching, to me. You can choose a focal length that gives you the desired output PPI for the viewing distance. Normal or moderately tele focal lengths gives very much detail. Full image size and aspect ratio are only limited by the final size of the output, you just keep adding partial images till the scene is covered.

A tool like PTAssembler is made by a photographer, specifically with large output in mind (rather than 360 degree VR). Photoshop CS3 may turn out to be a good tool as well, but I have no experience yet how well CS3 handles gigapixel images.

Bart

Testing to see if a 1DsmkII will do for 40"x50" prints!

Ben Rubinstein said:

Does it look like I'm barking up the LF tree?

Click to expand...

Hi Ben, You might just be right.

Can you post an example of the type of composition and crop and I will try to shoot one with My 1DsmkII for you to try out to see if at best processing it will do for what you need. I think it might. If You were printing B&W, Sepia, or other monochrome type I would say not a problem but color is another thing completely.

Hi Will,

I have a 5D and have worked with a 1Ds mkII. Yes you can print big with certain types of composition and subject. I've found that you cannot invent detail that is not there and with highly detailed landscapes with lots of small details uprezzing just doesn't compensate for the mushed or simply empty 'space' around pixel level.

Take for example this picture. Shot with a 5D and the new 70-200 f4L on a MF3 with manfrotto MG head and RRS QR system. MLU and cable release of course. Focus on the wall at f16. Shot in RAW, sharpened carefully with PK.


The image is incredibly sharp, huge amounts of detail. But I had to rez it up slightly for print and looking close up to the print you can see that there are areas of wall that just don't show detail. At native resolution the print shows all the detail that the eye could want. Once you rez it up the detail just isn't there, the eye wants more when looking closely.

Of course you will say that we don't look at prints with our eyeball touching the picture and you are correct. From a meter back the print looks incredible. But large prints hung for display are not viewed at 'correct viewing distances'. Not by a large stretch. I'm looking for prints a full meter in width at a minimum that can stand up to close scrutiny for resolution and details. I'm sorry but I just cannot accept that any 35mm DLSR can come close to that with detailed landscapes from a lot of experience with working and printing with the files.

Bart, I thought about stitching. How well does stitching work with architecture in the shots, is it easy to stitch in 2 directions rather than just in panoramic, can you work fast enough to gave even lighting across the frame when the light is changing, etc, etc? I am interested in it but to be honest from all I've read and seen, LF looked easy and relaxed in comparison!

I think that for my needs flat stitching would be the only real answer, with architecture and my experience with PT in general you will be working very very hard to normalise a large amount of frames when you have rotated the camera. The distortion might just be too much or necessitate a silly amount of overlap to ensure only the middle of the frame is used and the resulting horrible amount of frames needed.

Ben, I was speaking of a portrait on the mantle type of shot you had mentioned. I believe this type of shot can be made to have the appearance of detail with the 1DsmkII and DPP SW without a doubt in monochrome and very likely in color.

I would like to try My Voodoo on a 1DsmkII RAW file to see if I could meet your detail requirments.

Ben Rubinstein said:

Bart, I thought about stitching. How well does stitching work with architecture in the shots, is it easy to stitch in 2 directions rather than just in panoramic, can you work fast enough to gave even lighting across the frame when the light is changing, etc, etc? I am interested in it but to be honest from all I've read and seen, LF looked easy and relaxed in comparison!

Click to expand...

Ben, stitching is just one of several approaches possible, but one worth mentioning. With more recent software (e.g. automatic control-point generation and automatic blending) components, stitching is getting to be a serious contender versus LF, IMHO. It is not the end all cure all solution and it requires/offers a different workflow, and there are some issues around moving subjects.

To specifically answer some of your questions;
- Architecture, which is one of my fields of interest, is easy to do in a stitching approach. The software allows to e.g. de-skew the image composite AND remove lens distortion at the same time. It also allows 'crowd control' i.e. removal of people/cars etc. by shooting with a more generous overlap.
- Multi-row stitching is not more difficult* than single row, it just produces larger files. It does help, as always, to prepare by getting the nodal/entry-pupil (the point of rotation) as correct as possible, especially if you include foreground elements. On distant structures you can even shoot handheld and get perfect results, and the use of longer focus lenses allows to get huge amounts of detail, especially if you cannot physically get close enough to even see the detail by eye.
- Lighting/exposure differences between images over time are more of an issue than with a single shot, but blending components in the software can easily compensate for relatively modest differences in exposure, and are getting better at bridging larger differences. If the fluctuations are rapid and large (would also be challenging to correctly expose a single shot), you can choose a larger overlap (e.g. 50%) between images.
- Etc., etc. Well it'll get easier with practice .

* The use of a good 3D pivot tool as camera support does help a lot. I recently switched from my heavy Manfrotto unit to the much more portable and flexible RRS Ultimate-Pro Omni-Pivot package. It's a bit of an investment, but also a delight to work with, especially if you do a lot of stitching, and it allows to work quite fast.

I think that for my needs flat stitching would be the only real answer, with architecture and my experience with PT in general you will be working very very hard to normalise a large amount of frames when you have rotated the camera. The distortion might just be too much or necessitate a silly amount of overlap to ensure only the middle of the frame is used and the resulting horrible amount of frames needed.

Click to expand...

Yes, I almost exclusively shoot in 'rectilinear' projection mode. That means that you'll get exactly the same type of enlongation/projection distortion at the edges as a single flat field lens with large FOV would. If you restrict the composite angle of view to normal angles, that won't be any issue. Even if you choose an extreme FOV, when printed large, and viewed from up-close from the perspective point, it looks natural. You are flexible in choosing your perspective by selecting the viewpoint with your feet (as usual), and your angle of view by stitching more images (instead of using a wider-angle lens). In post processing it is also easy to either fully correct for keystoning, or only partially, and you can change till it's exactly as you like.

With a non-Tilt lens you can even simulate infinite DOF by shooting more images with a narrower FOV lens, and adjust focus for each sub-image. The software can automatically correct for the resulting differences in magnification that follow from the different focus positions of the lens.

This is perhaps a nice report from a relative newbie in stitching his first landscape, and comparing that to his prior MO of lugging of his LF gear through the rocky mountains (which poses some other constraints compared to most architecture). While he is an experienced photographer, his first attempts at stitching were easier than he anticipated with moving grass and clouds and exposure fluctuations and focus variation. In comparison, architecture is easy ...

Bart

Gosh but it still sounds complicated! Especially with the DOF issues, etc, etc.

Can I ask a really stupid first stiching question?

I had a look at all the pano gear on the RRS website. Seems to me that the vast majority of the rails mean that your camera will not be sitting on the center of gravity of the tripod. It's hard enough getting sharp exposures as it is with a solid tripod and a good head, seems madness to me to then mount the camera 'hanging in mid air' when you have moved the camera on the rail to obtain the correct nodal positioning.

Of course there are many many more questions to come though!

Just did a quick test with a 5D file. To get a 50X30" I would need 8 files in two layers (4 each) at a very bare minimum and I still only get 230 DPI. Then of course I need the processing power to deal with it given that I usually bracket and combine to maximise the DR. Looks like I would have to stitch both sets of file, light and dark, and then layer them. Ouch!

But will it give me more resolution, or enough advantages at that size over LF. I know LF is not easy and no childs game, but heck it looks simpler than this!

Ben Rubinstein said:

But will it give me more resolution, or enough advantages at that size over LF. I know LF is not easy and no childs game, but heck it looks simpler than this!

Click to expand...

It's actually not hard at all, especially if you have a DSLR with zoom handy to test exposure. 4x5 film holds up very nicely at 2400dpi - you're still getting detail, not just grain. Even 4800 dpi (limit of my scanner) seems to be eeking a little more detail out. At that resolution you should have plenty of pixels to make your 40x50 prints (2400dpi is just shy of it - but scanner pixels seem a little better than camera pixels, since there's no interpolation). The trick is to avoid stopping down so much that diffraction loses too much resolution. IIRC, f22 is where effects are supposed to start showing on a 2400dpi scan.

I'm shying away from it myself because of the amount of work post-exposure, but it sounds like you're in for even more than that with all that stitching. You can save on the post exposure work by having someone else develop the film. I haven't tried paying for scans (too expensive for a hobbyist like me), but that MIGHT cut the other work (which is a combination of scanning and dust spotting the scan).

If your main worry is the complexity of actually taking the pictures, you'll be pleasantly surprised. You have to figure your exposure, but other than that it's only as complex as you want it to be.

I would buy a simple scanner to view the images then send the keepers for scanning on an imacon 868 so as to work the file for print.

Getting some idea about it now, for flat stitching how do you do two layers?

Ben Rubinstein said:

Gosh but it still sounds complicated! Especially with the DOF issues, etc, etc.

Can I ask a really stupid first stiching question?

Click to expand...

While I'm reluctant to post stitching solutions in this LF forum, it is my feeling that adult behaviour would/should also allow the consideration of alternatives. That being said, there are of course no stupid questions, only unnecessary ones (and stupid answers ).

I had a look at all the pano gear on the RRS website. Seems to me that the vast majority of the rails mean that your camera will not be sitting on the center of gravity of the tripod. It's hard enough getting sharp exposures as it is with a solid tripod and a good head, seems madness to me to then mount the camera 'hanging in mid air' when you have moved the camera on the rail to obtain the correct nodal positioning.

Click to expand...

There you go, the first sensible question has already been flung back at my suggestion ..., and I like it.

To answer the question, you are quite right about the center of gravity being moved away from the center of rotation, it's an unavoidable requirement for reducing fore/background parallax between shots. However, the RRS (or Manfrotto or other solid solutions for that matter) solution has/acquires both some mass and, maybe more important, vibration absorbing properties (obviously depending on the camera/lens mass) derived from the tripod underneath. That's one of the reasons I have recently also chosen for a Carbon Fiber tripod, with better vibration absorbing properties than many materials (except perhaps wood), and with a lower weight for mobile use. In practice (with a 1DsMk2 and a variety of lenses) I have found it to be not a real problem (already with a heavy Alu tripod).

Of course there are many many more questions to come though!

Click to expand...

I'm happy you are open-minded enough to consider alternatives to LF.

Bart

Ben Rubinstein said:

Just did a quick test with a 5D file. To get a 50X30" I would need 8 files in two layers (4 each) at a very bare minimum and I still only get 230 DPI. Then of course I need the processing power to deal with it given that I usually bracket and combine to maximise the DR. Looks like I would have to stitch both sets of file, light and dark, and then layer them. Ouch!

Click to expand...

Okay, let's address the concerns. Is 230 PPI enough for your almost nose-to-print viewing of large output? I'd wager it is and if it isn't, just add some more Focal length and images. Memory is relatively cheap. In fact, I just did several 7-exposure per partial HDR stitching image captures of an interesting church, the exterior of which posed several additional challenges.
I'll try and post an example of that, probably in a separate thread.

But will it give me more resolution, or enough advantages at that size over LF. I know LF is not easy and no childs game, but heck it looks simpler than this!

Click to expand...

Only you can answer whether it is enough, or if it's simpler in the end. I'd suggest you try it, and a free progam like 'Hugin' will already offer a lot of functionality without the need to invest much up-front (although some investment will ultimately pay-off). Experience only evolves with trying.

Bart

Hi Ben,
Why not, as a quick and dirty comparison, look here, http://www.openphotographyforums.com/forums/showthread.php?t=2615
and ask Asher and Will to post a 100% crop, of a similar area perhaps the ballast area.

The software trial is free, and a few hand held grab shots of a building near you may give a good idea of the capabilities of stitching, and handheld, for architecture.

Best wishes,

Ray

That's an interesting point Ray, I wonder if anyone is really bored and feels like shooting a comparison between LF and stitching (of a scene with architecture in it) to the same file size (big enough for a 50X40 ish) and show the methodology used for the stitching shot.

As far as flat stitching is concerned I hadnt realised the cost of TS lenses, hell I could get a new LF field camera and lens for that amount!

Thanks a lot Bart for your time, would be very interested in your image as it would parallel exactly what I would be trying to do.

Ben Rubinstein said:

That's an interesting point Ray, I wonder if anyone is really bored and feels like shooting a comparison between LF and stitching (of a scene with architecture in it) to the same file size (big enough for a 50X40 ish) and show the methodology used for the stitching shot.

As far as flat stitching is concerned I hadnt realised the cost of TS lenses, hell I could get a new LF field camera and lens for that amount!

Thanks a lot Bart for your time, would be very interested in your image as it would parallel exactly what I would be trying to do.

Click to expand...

Hi Ben,

I believe one can approach the same end result with any of these approaches. The special thing about LF photography is its simplicty and dealing with information with less mathematical passes.

All one needs is a lens of sufficent image circle for the f stop you need. If the lens has no geometric defects then architecture is dependant on one's own skill.

For 39M Backs, the same thing, one just needs the right lenses.

For stitching, the workflow can be longer depending on whether or not extra work is needed to deal with imperfect junctions due to color, contrast, texture defects due to geometric changes coming from X,Y or Z axis rotations.

If one stitches in one direction only and with one row fast and in stable light, then the methodolgy will be likely be fine for many applications. The more rows one needs, the greater the risk of geometric distortions on stitching.

Asher

Well I would be needing at least 2 rows to match the output.

I went onto B&H and RRS to do some pricing...

A 45mm TS lens will cost ~$1090 and I can use my present system.
A RRS multi level panning system costs ~$800. I would also need a computer upgrade to handle that much stitching.
A new Tachihara LF camera with new Rodenstok 150mm lens, holders and board, etc will cost ~$1295 and would be cheaper if I sourced the lens 2nd hand.

The Techihara is a great camera. I think you would be very happy. It is one that will give you the results you want every time.

Also it is professional and you can arrive at any job or clients set up with that and your digital cameeras and feel first class.

If one only has 4 needed shots of a building, you shoot maybe 12 and digital as backup.

I think you'd be happy depending on your numbers. BTW a 5x7 camera with a 6x17 Canham back back (totally automatic!) for $1100 is attractive.



It does weigh 2lb and fits especially well Canhams 5x7 cameras and the Walker indestructable polymer camera here .

These I know are for sure double the price but still much cheaper than any digital back! Just requires the matching work philosophy.

So back to the Tachihara. Yes that's a bargain.

Asher

For those who regularly shoot LF, I've been trying to teach myself the 'world' of LF but am coming up with some questions. The value of tilt seems not to be of use with the kind of 'alleyway' photography that I will be doing (see www.bphotography.co.uk/fineart/fineart.htm for more examples in the 'Old City Jerusalem' section, all distortion correction was PS based!). Seems to me that working with so much vertical foreground kinda makes tilt redundant. Any thoughts and will I be stuck with using WA lenses only to get enough DOF without too much diffraction?

Looking at a couple of lenses which match the focal lengths I'd been using for that kind of shooting, namely the 90mm and 210mm (~30 & 70mm, not too wide for distortion at one end and just long enough for some compression at the other) would I be getting enough image circle for those shots to rise and tilt/swing or am I going to be running short? A monorail and super expensive lenses are out of the question.

Many thanks

Hi Ben,

What camera did you use for the Jerusalm work previously?

I really love the Mamiya 7II for that sort of work. Nothing to my knowledge beats the resolution and quality of these lenses. Phenominal!

I like the the pictures you took of the wall from below, did you correct the images in PS?

Asher

That was mainly a Canon 10D actually which is what I was shooting in the beginning of 2005, I changed to the 1Ds in time for the Iceland trip but then went for the 5D almost as soon as it came out as shooting weddings with the 1Ds was very difficult, everything had to be absolutely perfect! There is PS perspective correction on several of the shots though only the middle of the trio needed any serious correction as I don't mind standing back to improve the tilts by using a longer lens.

The Western Wall in the rain was on the 26th of Dec 2006, just a couple of months back with a 5D, It is almost exactly as it was out of the camera apart from straightening the horizon and a bit of dodging and burning on masks after coverting to B&W. That is what I love about the 5D, you have to do so much less to the file out of the RAW converter than the 1Ds ever needed.
I was out there looking for an apartment as we are moving back there in December of this year which is when my 3 big projects kick off in between flying back to the UK to shoot weddings. It is almost exactly as it was out of the camera apart from straightening the horizon and a bit of dodging and burning on masks after coverting to B&W. That is what I love about the 5D, you have to do so much less to the file out of the RAW converter than the 1Ds ever needed.

One project is documenting Jerusalem and specifically the old style Jerusalem for two books that I will be working on with a friend who is involved with book/script writing for some interesting projects being tossed around in Hollywood at present about a film based on the 'shtetl' (ghetto) life and yiddish culture pre 2nd world war.

The 2nd project is shooting touristy type scenes for printing large on canvas and selling through tourist channels, the 5D will be plenty for this as the canvas is stupidly forgiving, I also want to put this work up for stock.

The 3rd is to shoot B&W in big, documenting the iconic images of Jerusalem in a way that you could walk into the image and be there. The pictures will not necessarily be the 'iconic' of the tourist trade, but a new look on so much that becomes 'normal' after walking the streets for even a few days. That is what I'm looking at the large format for and it will hopefully be exhibited and sold here in the UK, no one in Israel has houses big enough!

Ben Rubinstein said:

Looking at a couple of lenses which match the focal lengths I'd been using for that kind of shooting, namely the 90mm and 210mm (~30 & 70mm, not too wide for distortion at one end and just long enough for some compression at the other) would I be getting enough image circle for those shots to rise and tilt/swing or am I going to be running short? A monorail and super expensive lenses are out of the question.

Many thanks

Click to expand...

I have a Tachihara and 90, 135 and 210mm lenses. I've got plenty of movements on all 3, without using recessed lensboards. I can't use the movements to the full extent of the camera on the 90mm, but that's because I have an antique 90mm lens with only limited coverage. With a more modern lens I'd be fine.

Also, realize that movement limitations on wide angle lenses tend to go away when you focus in closer than infinity (the closest your lens gets to the film plane). If you focus in a little closer you get a lot more breathing room. I used to use a monorail that had no room for movements on the 90mm when focused at infinity and this made a big difference. The Tachihara should be able to work even a 75mm just fine at infinity.

On the other hand, a 90mm is a 90mm, and whether it's wide angle or telephoto (like on an APS-C sensor) you get the same depth of field for your f-stop. The good news is that you can actually view what the depth of field will be right on the ground glass by stopping down and checking it out. You know what you're going to get, unlike an SLR where the viewfinder is so tiny.

Thanks Erik, would you mind having a look at those photos and telling me whether tilt would be of any use at all? Trying to work out if I'm letting myself in for some serious DOF problems because the buildings in the foreground won't allow tilt or so I understand.

Ben Rubinstein said:

Thanks Erik, would you mind having a look at those photos and telling me whether tilt would be of any use at all? Trying to work out if I'm letting myself in for some serious DOF problems because the buildings in the foreground won't allow tilt or so I understand.

Click to expand...

Tilt, swing, etc will be tradeoffs. You'll have to pick which parts to make sharp, or stop down a bit. In your night shot of the castle, tilts could make the rear tower and foreground grass very sharp regardless of aperture. The hedge back under the trees might be out of focus. Or you could put the plane of focus parallel to the wall on the right, so the trees in the upper left would go soft.

It looks like you're rather far away from the tower - at 90mm, hyperfocal distance is somewhere between 30 and 120 feet at f/22 (depending on print size). A lot of these seem like you're that far away, so depth of field wouldn't be an issue at all. The same with your picture of the person in front of the wall, especially if some of those reflections out at the very lower edge of the picture can go a little soft.

For reference, this picture was taken with a Tachihara and 90mm lens at f16. The back edge of the large rock in the foreground was 10-15 feet away, and the closest part at the lower left of the image is out of focus (probably 6-8 feet from the lens). Softness in upper right is actually blown highlights on the slide film I was using. The same shot with black and white film captures detail in those areas perfectly (but oddly, I like the effect of the conversion better than the "native" black and white).

"I really can't afford what Id like a 39MP digital back. That, anyway will be outdated. The film camera is inexpensive, but bulky and needs film that needs processing. If one does not require thousands of shots, film seems to be a great option for detail rich work."

This is an interesting point: as a longtime LF-User with 8x10" and 4x5" i can´t see a REAL advantage in a 39MPx back - especially with it´s cost in mind.
I´d like to show a direct comparing from a job i did last year - a 39MPx and an 8x10", both landscapes - how can i do that?

best, Klaus

P.S.:
the 8x10" is here:
www.klausesser.de/Template2.htm
the 39MPx is here:
www.klausesser.de/Template.htm
here is a 6x17cm:
www.klausesser.de/Stadt.htm

the 8x10" and the 6x17cm are drum-scanned withh 2000dpi

Bloody hell but the resolution of that 8X10" just goes on and on, just mind blowing. Probably far more information still in there at higher scan resolutions, the detail on the leaves certainly looks like there is more to come.

It's interesting how obvious it is at enlargement whether the resolution has given all it has to give or whether there is more detail there. That 8X10" looks like it could go on forever. The digital back has fantastic detail but at enlargement you can start seeing the smearing of fine detail which shows that the picture has given you all it has to give. Film grain actually hides the smearing which can be so apparent in fine detail with digital and makes it far less ugly. Even if leaves turn into clumps of grain, it looks better than smear!

If I can ask you further as I try to understand and learn, what movements did you use to ensure that much DOF in the 8X10" shot? Would tilt not screw up the focus on the bush?

Ben Rubinstein said:

If I can ask you further as I try to understand and learn, what movements did you use to ensure that much DOF in the 8X10" shot? Would tilt not screw up the focus on the bush?

Click to expand...

Hi Ben!

Yes - i could have done it sharp all over. But in my opinion it should be even more blurred in the foreground - gives more deep-effect . .

One of the advantages in this shot is that i have done it very early in the morning - so the air was cleaner as it is normally over the day.
It was a 240mm Symmar in a Linhof on Fuji Velvia - i used a polarisation-filter.

best, Klaus