Ektar vs Alpha 900/Velvia revisited: new scans

[keithwms]

Okay, I agree and you don't have to make drum scans to please yourself... or me I was merely curious to see a 35mm atp frame scanned to the grain.

 

[domaz]

Multisampling is useful for digital (in which case it's usually done to expand the dynamic range); why not consider it for film as well. I'm guessing that simply by stacking negs in an enlarger (with careful registry) it should be possible to get grainless enlargements from 35mm well past 24 inches or more. Of course, via drum scanning, this approach might be easier.

Hmm I don't think stacking negatives would eliminate grain. You would get twice the density though. Digital multi-sampling works by taking the pixels of the two or more images and sampling them to filter out the "noisy" pixels.

 

[Photo Engineer]

You would also introduce a sharpness problem in all probability.

PE

 

[keithwms]

Well, you can enlarge as much as you like and do the alignment that way. Of course it'd be easier to drum scan each neg and then electronically align them, but a fully analogue alignment should be possible. I don't see why it should be any harder than analogue unsharp masking. Anyway maybe I'll try it soon.

Domaz, I allege that multi-neg printing would reduce grain because grain alignment is random, whereas "real" image information is not. So the grain structure would be effectively averaged out if you expose through multiple negs, thus improving signal to noise, i.e. reducing grain.

 

[Photo Engineer]

Of course, you can only go the multineg route with still life photos. Any motion would cause blur and color fringing.

PE

 

[domaz]

Keith- I don't see how it would work. Just think about this scenario: you took two identical pictures of someone and you want to crop their eye to be really big for some reason. You enlarge the eye and realize it's all grainy- so then you stack the other negative and the grain is going to disappear? Doesn't seem likely. The grain structure is different between the two pictures but at the same magnification you still have the same amount of grain. If you think of pixels as grain you can do the same simulation with a digital image- you will find having two pictures of the same magnification does not eliminate pixelation. I think you are confusing noise which is something that CCDs have a bad habit of introducing in digital images with grain which is actually what a film photo consists of physically.

 

[Photo Engineer]

Domaz;

The pixels in a ccd are regular, but grain in film is random. Therefore it is theoretically possible to superimpose several identical negatives and get no grain. The problem is that this can only be done with still lifes (see my previous post) and at the sacrifice of some degree of sharpness. If you had 10 images then, each one would be printed for 1/10th the exposure time onto paper, in register and the superpositiion of random grains would cause them to appear to vanish, while sharp edges began to slowly blur. There would be a tradeoff point in the number of images possible before blur overcame grain as the most objectionable feature.

PE

 

[keithwms]

Yes, I agree, this technique would work with still subjects, and that's why I wrote "when shooting static subjects" in my post above.

Notwithstanding subject motion, I don't see why sharpness would need to be sacrificed, though, Ron. The registry could be established at a much higher magnification than possible to discern by the naked eye. Well I guess the proof is in the pudding, so I'll try it when I get time.

Anyway, like I said, if you are willing to scan the negs and do the alignment electronically, then this kind of multisampling could be an extremely powerful way to reduce grain. Image-stacking and registry software abounds, and this general line of thinking will be familiar to any astrophotographers... and their subjects sure aren't static But anyway, I still allege that this can be done in a totally analogue way.

 

[pentaxuser]

Wow! Based on my reading of the posters I think I will be in a minority of one but I just have to say this for my own honesty's sake. I thought that the Sony Alpha 900 was streets ahead in terms of the blown up scans. All the others looked remarkably grainy and by that factor much poorer by comparison. If this is what a 25MP does then certainly by the "man in the street" judgement and I'd like to think of myself as a "man in the street, the Sony wins hands down.

If these were the blow-ups of my daughter at her wedding then I know which camera I'd want the wedding photographer to be using. OK at the usual photo print size little of none of this advantage would show but is there any reason not to conclude that nearly 25MP beats 35mm film.

I have no intention of migrating as I like what I do with film and its processes but in terms of quality I can see only one winner here. OR am I missing something?

pentaxuser

 

[Photo Engineer]

Don't forget that all digital images are manipulated to some extent, even in the camera. Otherwise they would be pretty bad.

Just as all analog images are manipulated via chemical means during development.

PE

 

[domaz]

If these were the blow-ups of my daughter at her wedding then I know which camera I'd want the wedding photographer to be using. OK at the usual photo print size little of none of this advantage would show but is there any reason not to conclude that nearly 25MP beats 35mm film.

I think what your missing is most wedding photographers used medium format film before digital came along. I agree that it beats 35mm film though. Of course digital has other advantages- like being able to delete an entire wedding worth of pictures in an instant. Search on photo.net and see how many of those threads you find. Ouch.

 

[Photo Engineer]

Keith;

There are two reasons I said sharpness would (probably) suffer. One is that in your initial post you suggested stacking the negatives in the enlarger at one time in register. This would give sharpness problems due to internal reflections and positional focusing problems.

Then again, if you did it with multiple exposures in register, it is virtually impossible to keep the enlarger stable beyond about 2 - 4 exposures, and this was a problem with Dye Transfer in the hands of many users until they learned how. It would probably have to be a contact printing method.

In addition, the flare generated by such a system might be too severe.

And lastly, the perfect registration of these images will be just enough off that there will be a smear effect in the edges.

This does not includ any problems introduced by camera movement during the exposure of all of these negatives.

You can get sharpness or grain, usually but getting both at the same time is very very hard.

PE

 

[alan doyle]

i think these kind of tests are great...but why not use a proper drum scan machine..some companies do you a cheap test scans for less than 30 dollars...
even this using these amazing leica nasa scanner machines, some companies can do for 10 dollars..per picture..
http://apollo.sese.asu.edu/ABOUT_SCANS/index.html
if it is a proper shoot out test get a pro scan done.

 

[Marco B]

i think these kind of tests are great...but why not use a proper drum scan machine..some companies do you a cheap test scans for less than 30 dollars...
even this using these amazing leica nasa scanner machines, some companies can do for 10 dollars..per picture..
http://apollo.sese.asu.edu/ABOUT_SCANS/index.html
if it is a proper shoot out test get a pro scan done.

Wooow... thanks for that Apollo project link, that's amazing! Pity the Panoramic and Hasselblad color and BW scans are not yet done, I would love to see those!

I would also be interested by the way to know what companies specifically you are referring too regarding these drum scans.

 

[JanaM]

Together with some friends we have recently made a similar test like Marco. But we have avoided his main mistake, using a test chart with a very, very low resolution figure of only 40 Lp/mm.

We have used a test chart with resolution values up to 200 Lp/mm. Our test chart has an object contrast of 1:16 (four stops).

Our lenses: Nikon 1,8/50 and Zeiss ZF 2/50 and 1,4/85.

With the Nikon D3x (24 MP sensor) we almost reached it's physical resolution limit of 85 Lp/mm. More is not possible because of aliasing artefacts (Nyquist theory).

With Fuji Velvia 100F we got 120 Lp/mm, with Provia 100F and Astia 100F 115 Lp/mm. With TMX 130 Lp/mm.

With Provia 400X and TMY-2 110 Lp/mm is possible.

With Spur Orthopan UR we achieved 200 Lp/mm, but only with the Zeiss lens.

With all slide films grain is generally much finer than with color and BW negative films.
Exceptions: Spur Orthopan, Kodak Imagelink and Technical Pan, Rollei ATP. These films are almost grainless and deliver outstanding resolution and sharpness.

 

[ajuk]

I thought Velvia was supposed to be uber saturated not natural? I doubt such a high end scanner would loose 1mm round the edges.

 

[dr5chrome]

scan negs?

PE; I have never experienced these scanning 'toe & shoulder' issues you speak of, please explain further... My experience of over 12 years as a custom lab service and 17 as a photographer and since '91 in custom positive processing; I have come to realize Negatives were for analog printing and chromes were for; then magazine work, then later for scanning.
[There is also an "original art" value to the positive image relevant for another conversation].

I have never in all my days met a negative, color or B&W, that liked the scanner better than a chrome. Maybe this is just me? I highly doubt it. Every one of my 3000 clients prefer chromes over negs to scan.

I hear this, but not often that negs are better in scan. I am just baffled by it. Its fine in explanation, in practice and result its awful. I can only commend those able to achieve a good scan from a negative.:confused:
In my personal work if I ever find time these days, If I am not making fiber prints, I would always shoot positives, color or B&W. I don't even offer C41 at my shop. Most of the Pros that send us film don't even use C41.

regards
dw

Marko;

Very impressive results. Good work.

I have one comment on your analysis at the end, of the reversal and negative systems.

You must remember that color negative is purposely built to have a contrast of 0.6 - 0.7 which is taken advantage of when scanned or printed by being turned into a positive image of about 1.5 - 1.7. A reversal film is built to have a contrast range of 1.5 - 1.7 to start with. In the end, they have the same approximate final visual result, but the color reversal has less latitued and less color correction.

In conventional printing, color reversal suffers from having a toe and shoulder which can be compensated for electronically by scanning, but color negative has no toe or shoulder and therefore does not suffer in the transition to either analog or digital copying.

PE

 

[Photo Engineer]

Color negatives have finer grain, better color and a longer tone scale than reversal. The problem is that since there is no comparison to work from, many software developers have problems with the mask and the tone scale of negatives not to mention the fine grain which is smaller than the sensor's capability and causes a rollback or rollover that makes a negative appear to have high grain. (See an explanation and a reference by Pat Gainer in the Rodinal II thread)

Ok, we know that color negatives have a long tone scale, as much as 3 stop on the high side and 2 stops on the low side. Thes are used in over and under exposure, but basically result in a straight line for the usable portion of the scale with no toe or shoulder. The electronics and software introduce a tiny bit of toe and shoulder.

If you scan a slide, it has a toe and shoulder right in every scene! That is if you use the slide properly, otherwise it is all on the toe or shoulder. In any event, scanning this scans the toe and shoulder, but then imposes the electronic equivalent on top of this (same thing happens when you print a slide) and you compress the data.

You see, a copy of a copy is slope of curve * slope of print process = slope of final print. Since a negative slope is 0.6 throughout, then the only error comes from the slope of the print process and by minimzing this you maximize the quality of the final print, but in a slide the slope varies throughout the curve of the original thus lending distortion to the entire curve of the print. The print thus has a distorted compressed tone scale.

This is why pos-pos never gained popularity in motion picture. Repeated generations or SFX with multiple generations ended up looking "dupey", but with neg-pos, the final result was as we see today even through 15 generations of SFX creation.

PE

 

[keithwms]

DW, I agree, and my experience with drum scanned colour film is that chromes scan much better. I also agree with the notion of slides as original art.

Ron, whatever the theoreticals, [well exposed, appropriately used] slide does scan better, according to my results. I mean I can show you results taken on the same format with an Aztek drum scanner, arguably the best in the biz. I'll see if I can dig up some full MF scans if you wish. The problem with side by side comparison is that in practice, I'd never shoot the same kind of scene on slide and colour print film. I reach for one or the other depending on the scene, the range, the final intention I have for the shot. So this is a case where a 'standard' scene or test chart just doesnt give the right answer because we would be asking the wrong question.

I have been repeatedly disappointed by drum scanned negs. Now the big proviso, of course, is that the range fo the scene has to be appropriate for slide. Now, I am a natural, gentle light lover, and I almost always shoot everything in low, muted light. If a scene has 6+ stops of range in it then I am not going to shoot it to slide anyway. I am just saying that wide range contrasty scenes are not what I go for in my colour stuff, so that you can calibrate what I am saying. And what I am saying is that slide films like velvia 100, astia 100F, provia 400x, when used appropriately, give spiritually moving results!

Ajuk, velvia 100 is not 'uebersaturated and unnatural'..... first of all there are three (arguably four) velvias, all with different colour palettes. Second, some people like the oversaturated look and will deliberately underexpose to get it. Soem folks rate velvia 50 at 64 for that reason. Indeed, if you do like that look, then slide film is an easy way to get it. But that does not mean that that it is intrinsically unnatural. Pretty much all of the colour images that people tell me they really like were taken on velvia 100.... and not once has anybody said they look oversaturated or unnatural. Hmm maybe they are just being polite Also, Fuji's newest 64T gives extremely accurate colours. Frankly. People who harp on the issue of colour accuracy probably haven't shot any of the recent slides films... astia 100F, provia 400x.... these films are absolutely superb.... and to get back to the original point, they drum scan fantastically well. Another possibility is that some folks expect slide to stand up to neg film in a wide-range scene... of course it will not. And personally I could care less, if I want an HDR look then I will get it digitally

 

[Photo Engineer]

Keith;

See Patricks reference on apparent grain in the other thread.

We used a K. S. Paul scanner at EK in our early work, but IDK what was used later on. We did build some ourselves of course.

The argument here is one of tone scale Keith. Draw a negative curve and a positive curve. Then overlay a positive curve on each one and plot the resultant line as the derivative of each multiplied (slope x slope at each point). You then plot the pos-pos curve and the neg-pos curve. Then you will see what I am talking about.

Remember too, that software is trying to do all it can to fix this error introduced by multiplication. And, remember that unmasked imaging is lossy in color reproduction as well.

PE

 

[keithwms]

I already see what you are talking about, Ron. But the thing is, I use slide and print film for two totally different purposes. I am not going to use print film for a 4 stop scene, and I am not going to use slide for an 8 stop scene. And it turns out that most of the scene that I like, that appeal to me visually, are ~4 stop scenes!!! Again, if I want 20 stops of range then I know very well how to get it... but that typically doesn't appeal to me. Slide does what I want... and the scans are, well... maybe you just have to see for yourself.

Also, there is no issue of software manipulation here as far as I am concerned, I am talking about *straight* drum scans. As you are aware, a true drum scan involves no sharpening or tone manipulation. There isn't any lensing. It's as true to the actual info in the film as technically possible.

Overall, I think it's much too easy to get led astray by technical arguments about how much information we record, how high the resolution is, etc. A big part of the appeal of photography, to me as somebody who frankly doesn't have trouble grasping the technicals, is that the technical stuff is there in the background but it does not ultimately decide what makes a compelling image. Because of my own experience, I am 100.5 % convinced that my favourite images on slide would not be as effective on print film. 100.5%. That is even higher than 100%.

 

[Photo Engineer]

You are correct about the drum scan but most here use flatbed scanners and I was only thinking narrowly on that.

On the matter of the scene, it really does not matter how long the scale is, there is loss in the slide and not in the negative when all else is identical as far as scene and proper exposure goes. The point is that the loss does not appear to bother you nor others. I have done the experiments and compared the two and can see the difference.

I converted to negative, and when needed, I made slides from the negative using print film.

PE

 

[keithwms]

But hold on Ron. When we do b&w, the (almost universally accepted) best of the best output is had when one uses BTZS methods... i.e match the range of the scene to the range of the film and ultimately aim for the best print. That said, why is it that when we speak of colour imaging, suddenly that principle is thrown out and it's assumed to be better to use a colour film with very high latitude (colour print film) on a small range scene (e.g. 4-6 stops) as it would be to use slide wt a range that matches the scene perfectly??? To put it another way, I think I get my best results when the two ranges match up as much as possible and the knee and toe just enter the scene.

To put it another way: why would I want to use an 8 stop film on a scene with much less than 8 stops of range? I like my knee and I like my toe!

Or to cast it in b&w terms: can you imagine using, you know, POTA-developed neg film on a scene with 6 stops of range? I mean, sometimes people apply way more film range to a noncontrasty scene and then wonder why they can't get compelling tonality without doing darkroom backflips to essentially rebuild the knee and toe back in. I did that for quite a while, I admit it. Split grade and scan+photoshop, you name it. The source of my troubles didn't actually occur to me until I started shooting some b&w polaroids for proofs and was wondering why suddenly the tones looked so nice and snappy to me... yes there were highlights and shadows on the ragged edge or beyond the ragged edge, but the midtones, yum! And then I realized that, lo and behold, polaroid actually has very limited range- quite similar to slide, in fact. So wide range isn't always a good thing- I was previously working hard to get full range out of my development, even when the scene really didn't want that. End of digression....! So these BTZS folks may just be on to something

Back to the point. I can get good results with print film, don't get me wrong, I know that can get very satisfying results, and for many scenes it's the only colour film that can handle the range. But if I use a colour neg film with, what, twice as much linear range as the scene brightness range, then I am going to have to curve the neg like mad to get what I want, no question about it. Bottom line is that I don't like to have to scan my negs and then photoshop 'em into what I can get straight away... no manipulation... from slide.

Anyway I have two drums scans, one from velvia 100 and one from 160 film, both shot on a rb67, drummed to ~30-40 megapixel res with a top-o-the-line Aztek (regarded as one of the very best of the drum scanners), and I am trying to somehow get them on my site for you to look at. What you will see is that the neg shot, despite being about as perfectly exposed as humanly possible (spot metered TTL), is far grainier and consequently doesn't enlarge nearly as well. I have printed both at mural sizes and the slide kicks the snot out of the print film.... despite the fact that they are both (ahem, if i may say so myself) pretty good shots, equally well exposed. To get good enlargement form the neg shot, I find that I have to do a nasty little trick: I have to curve it substantially and I have to overscan and then downsize to effectively average out some of the grain. That is not an enjoyable process, frankly.

This is an interesting discussion, may it continue peacefully

 

[Photo Engineer]

Keith;

Both B&W and color films are built to about the same spec. The negative films have an approximate best slope of 0.6 and several extra stops in latitude. The reversal films have a slope of about 1.8 and not much latitude. In fact, to be realistic, there is none, as the average scene is encompassed by the entire curve.

If you draw a straight line negative curve, with a slope of 0.6 and mark of what, to you, is the proper latitude, then superimpose a positive curve over that with a dmax of 3.0 and have it lie within the limits of your imposed latitude. Now, in addition, on the positive curve impose a toe and shoulder that introduce curvature over about 1/4 (each) of the positive curve. (this varies from 1/8th to 1/4th depending on the shape of the toe and shoulder)

You will see that in an average scene the positive curve distorts from about 1/4th to 1/2 of the entire scene due to curvature.

Now, you didn't say how you printed the scene you describe, but a digital print from neg and slide scans may contain the pseudo grain artifacts in the negative that I mentioned before and described by both myself and Pat. Gainer. An optical print likely would not.

PE

 

[dynachrome]

Earlier in the thread it was mentioned that b&w film has random grain. In most cases this is true but the Imagelink films have a monodisperse emulsion. The film is already so slow and fine grained that a random grain pattern is not needed. Whether you can get perfect registration in grain between two or more frames is another matter. Many people feel they get more punchy prints out of slide film in the right circumstances. I think that has more to do with the fact that it is more contrasty to begin with. This is necessary if it is to look good when it is projected. It doesn't mean that equally good prints can't be made from print film if it is exposed, developed and printed (optically or digitally) properly.