Medium Format Magic

[Bill Anderson]

Robert, there is no magic in large format per se- but
there is more information, and that equates to, well,
lots of decent things: the illusion of "depth" not being the least.
Smoother tonalities and greater fidelity go a long way
in promoting convincing interpretations of reality.
If you've ever seen fine art contact prints made from
8X10 negs then you'll know what I'm talking about.
Even early examples of large format photography,
with their "inferior" optics, can show remarkable
qualities when compared to many contemporary images.
Of course, paper and developer combos (not to mention the artists)
had a great deal to do with this. I've been a fine art photographer
for about 30 yrs and have used many formats- 35mm to 8x10
and occasionally larger- and many printing methods,
from gum bichromate, cyanotype, platinum, and so on; now
I use digital all the way from acquisition to printing- but instead
of using large format I tile my images (taken on a very modest Nikon d200)
on location and blend them back together.
This is superior to using a single medium format exposure.
I could do the same (tiling) thing with medium format and that
would improve things again, but as far as your query goes:
a number of 35mm combined can produce a superior image
to medium and large format single exposures.
Which tells us a lot about the nature of various formats.
As an aside, with multiple frames I can "deconstruct" a scene
by photographing each important element separately, this also
allows me to reestablish spacial, tonal, and color relationships
(impossible with a single exposure no matter how large the format)
to give, among other things, a greater sense of place.
Now we can circumvent many of the mechanical limitations
of our tools, and we are no longer bound to one fstop, one exposure,
one focal length... and no one element gets to dictate at the
expense of other equally important ones. If you want greater depth
of field in a single exposure, shoot 4x5 or 8x10, they have independently
movable front and back elements that allow you to "swing" and/or "tilt"
toward the subject plane, etc.. and you can shoot at f 90 if you desire!

I was going to post a couple of images but it seems we're only
allowed to post 19k file sizes. My original files are 4 gigs, and
there isn't much left at 19k:

http://www.williamanderson.ca/works/show/id/27

http://www.williamanderson.ca/works/show/id/26

All color images (24 x 50inches) are from a 35mm DSLR. I'm looking forward to the next generation of RED sensors for stills and motion work.

 

[Bill Anderson]

And don't get me wrong, larger sensors/formats are
hugely desirable, and the tiling approach is only possible
with fairly static subject matter, and obviously not cinema.
But it is a decent example of how a small format can be
made to exhibit all of the attributes (and some) of a large format.

By the way, these works will be exhibited at Winsor Gallery
in Vancouver (opening reception Thurs OCT 8th; 5:30 to 8:30)
and The Whyte Museum, Banff, Alberta, FEB 6th, 2010, reception 7-9 pm)

if any of you RED cats are in the neighborhood, please consider this an invitation.

 

[Radim Schreiber]

Interesting topic, although I do not see any conclusions here. I have heard a lot about this microcontrast. What is that? Does it relate to chip interpolation or resolving power of lenses? Do you know where I could see side by side comparisons? How can you see microcontrast on a 640x480 photograph on a web??? Scaled down from 4000 pixel original after unsharp mask?

 

[Radim Schreiber]

Now that I have access to P65+ and PhaseOne body I should make some comparisons with my 5D Mark II.

 

[Daniel Browning]

I have heard a lot about this microcontrast. What is that?

It's contrast at fine scales. Getting high contrast at large scales is easier than getting high contrast on small details.

The following example demonstrations how diffraction causes a loss of microcontrast (mouse over the image to see the comparison):

Diffraction causes loss of microcontrast

Does it relate to chip interpolation or resolving power of lenses?

Yes, both of those affect contrast, along with many other factors. Contrast is measured by the MTF.

How can you see microcontrast on a 640x480 photograph on a web???

You can't. 640x480 is a wallet-sized print. The level of detail (spatial frequency) is very low. At these frequencies, most systems will have a nice, high contrast, and there will be no visible difference in contrast.

Larger print sizes can be simulated by resampling and cropping. Then you will see contrast at other spatial frequencies. Contrast also tends to drop as you get to the edge of the frame (image height) for many lenses and f-numbers.

Now that I have access to P65+ and PhaseOne body I should make some comparisons with my 5D Mark II.

That would be great.

 

[Roberto Lequeux]

Bill, if you would like I can hose the picture/s you wanted to post on my server. I am very interested in seeing any of the snow covered fields in full glory.

About the technical analysis, I guess I understand how if the image is squeezed into one point within the lens then the resulting projection is dictated and not a factor of the sensor size but the other way around. I guess I didn't think of how you can grab the same light path and project it onto a S35 sensor or a wall.

I guess it is all about bigger is better in terms of resolution, lens quality included. The more information you have the more accurate a photo will be, and the more accurate they get the more realistic they will be when we compare them to our stored images from real life in our brains. We see a lot of detail when we focus on a point, so we much want an exorbitant amount of detail in our photos so that we can do the same and get a decent amount of detail when we focus on a point in a picture.

 

[Radim Schreiber]

Daniel Browning, thanks that makes it clearer. Micro contrast is than basicaly contrast generated by the lens. Some lenses are better than others. I take it that MF and LF has better microcontrast in their lenses.

Funnily enough, I can take your samples and create same contrast (not sharpness and detail) by adding USM in Photoshop with these parameters:

Amount 50%
Raidus 100 pixels
Treshold 0 pixels

If I scale it down to web size nobody will be able to tell the difference, because sharpness and detail loss due to diffraction will not be visible at that scale.

So what is than the magic of MF and LF?

 

[Daniel Browning]

Daniel Browning, thanks that makes it clearer. Micro contrast is than basicaly contrast generated by the lens.

There are a lot of things that contribute to contrast of a given spatial frequency, but the lens is usually the biggest one.

Some lenses are better than others.

Yes.

I take it that MF and LF has better microcontrast in their lenses.

Kind of. There tend to be at least 3 factors that give their photos more microcontrast:

* Reproduction magnification
* More resolution
* OLPF

The first one is the biggest one. What it means is that for any given print size, say 30x20, APS-C images have to enlarged by a huge factor, because they are only 24x16mm (or thereabouts). Whereas MF is enlarged by a smaller factor.

So for the two formats to have the same contrast in a 30x20 print, APS-C needs to have high contrast (80% MTF) at 80 lp/mm (very high spatial frequency), but the Medium Format only needs high contrast at 32 lp/mm (medium spatial frequency). In other words: the crop factor has to apply to the MTF (e.g. MTF-50) in order for the resulting image to be the same. That can be difficult for lens designers.

The second factor is resolution. Even if the MTF-50 of the lens scales with crop factor, above, larger formats tend to have more megapixels or film area, allowing them to have more total resolution. That increases contrast due to higher MTF from the sensor or film.

Third is the Optical Low Pass Filter (OLPF), which reduces aliasing artifacts. MFDB tend to not use them, which increases contrast at the cost of artifacts.

Funnily enough, I can take your samples and create same contrast (not sharpness and detail) by adding USM in Photoshop with these parameters:

Amount 50%
Raidus 100 pixels
Treshold 0 pixels

Yes. You will get even better results if you used an algorithm built specifically for deconvolving diffraction (e.g. Richardson–Lucy).

So what is than the magic of MF and LF?

Diffraction at low ISO with equal resolution is the very best case for sharpening overcoming a difference in microcontrast. There are many other cases (most, I'd say), where sharpening can't do that. If actual resolution is different, for example. Noisy images, for another example. Most commonly, because the point spread function (PSF) of lenses cannot be deconvolved perfectly. A junky lens with sharpening is not enough to replace a quality lens. In the same way, small formats with sharpening are not enough to replace larger formats.

 

[Roberto Lequeux]

I don't suppose you can re-post that last one using only 3rd grade level words?

 

[Pietro Impagliazzo]

Daniel you said perspective can be matched in any sensor size.

But let's compare 1/4" chips to FF35 (maybe i'm being a little extreme in the example).

For smaller formats isn't there the problem of distortion in the wide angles?

Enlight me please.

 

[Daniel Browning]

Daniel you said perspective can be matched in any sensor size.

But let's compare 1/4" chips to FF35 (maybe i'm being a little extreme in the example).

For smaller formats isn't there the problem of distortion in the wide angles?

No. Perspective distortion (e.g. anamorphosis with rectilinear lenses) is exactly the same.

[You might be talking about barrel, pincusion, or mustache distortion. That is not a problem either. Of course, smaller formats *tend* to have cheap lenses with big zoom ranges. Some of those have barrel distortion as a natural result. But that has nothing to do with the sensor size: you can find many 1/4" lenses that do not have barrel distortion and you can find many 35mm lenses that *do* have barrel distortion.]

The following two examples have exactly the same perspective.

1/4" chip (3.6x2.4mm)
5mm lens
horizontal angle of view: 40 degrees
===============
FF35 (36x24mm)
50mm lens
horizontal angle of view: 40 degrees

If you have cameras of multiple sizes, I suggest you test it yourself to prove what I'm saying is correct.

 

[Daniel Browning]

I don't suppose you can re-post that last one using only 3rd grade level words?

Modulation Transfer Function and Spatial Frequency in Four Letters or Less, by Daniel Browning:

There are a lot of things that contribute to contrast of a given spatial frequency, but the lens is usually the biggest one.

Take a shot and show it on the wall. Some of the shot has good snap, or bite, like an edge that cuts. But the rest is more dull, like a bald head. You only get out what you put in. Move too much when you push "go" and it will have blur in the shot. Put in a dull chip and it will come out soft. Most of all, the lens is what you use to not get a dull shot.

contrast of a given spatial frequency,

Each shot has more than one part. A big part and a tiny part. The big part is what you see at once: the face, the car, the sky, the sand, etc. The tiny part is what you see when you look very hard: the one off hair, the dust, the tiny line, the one inch spot, etc. You even see the boom (ha ha).

Each part has a look to it, from the one very big part all the way down to each tiny part. The look can be good or bad. It can be an edge that cuts, or it can be dull and soft. Most of the time, the big part has a good look, but the tiny part has a dull look. That is just the way it is most of the time. It is hard to make it all so that the tiny part has a good look too.

Kind of. There tend to be at least 3 factors that give their photos more microcontrast:

Bert is a box. A big box. It can take the big shot. It has to have big film. The lens it uses can cost a lot. You can call Bert "MF". Tim is a tiny box. Tim can try to take the big shot too, but Tim only uses tiny film. Tim is "S35". Bob is not too big and not too tiny: "FF35".

Some say that Bert is good. Even more good than Tim and Bob. Most of the time it's true. I'll tell you why.

* Reproduction magnification
The first one is the biggest one. What it means is that for any given print size, say 30x20, APS-C images have to enlarged by a huge factor, because they are only 24x16mm (or thereabouts). Whereas MF is enlarged by a smaller factor.

So for the two formats to have the same contrast in a 30x20 print, APS-C needs to have high contrast (80% MTF) at 80 lp/mm (very high spatial frequency), but the Medium Format only needs high contrast at 32 lp/mm (medium spatial frequency). In other words: the crop factor has to apply to the MTF (e.g. MTF-50) in order for the resulting image to be the same. That can be difficult for lens designers.

When you blow up a shot to put on the wall, it can look very nice. But Tim is so tiny, that you have to blow up the shot very big to have it on the wall just so. Bob is not so tiny, so when you blow up the shot from Bob, you don't have to blow it up as much to have it just so. (Let's say it's a four by two foot spot on the wall.) When you blow up the shot that came from Bert, you don't have to blow it up so much, it can fill the spot on the wall with more ease.

The less you have to blow any shot up, the more it can look like a very nice shot. This is a very big help to Bert, and that is why Bert can look very good most of the time, when you put it next to Tim and Bob.

* More resolution
The second factor is resolution. Even if the MTF-50 of the lens scales with crop factor, above, larger formats tend to have more megapixels or film area, allowing them to have more total resolution. That increases contrast due to higher MTF from the sensor or film.

Bert also gets help from one more area. That is, most of the time, Bert has many more dots. Dots can help to see each tiny part of the shot. Tim and Bob tend to have less dots, so they can not see each tiny part as well.

* OLPF
Third is the Optical Low Pass Filter (OLPF), which reduces aliasing artifacts. MFDB tend to not use them, which increases contrast at the cost of artifacts.

Most of the time, a shot has more than one tiny part. More is good. But if the part is too tiny, then it is not good. If the part is too tiny, then it can not be in the shot. But what if it gets in the shot? That can ruin the shot. One of the bad ways it can ruin the shot is with what may look like a grid. When you get a grid in the shot that was not in real life, it is not good. So we try to keep the grid out of the shot.

One way to keep the grid out is with blur. Yes, I said it: blur. We put a tiny blur on top of the chip, just to keep out any part that is too tiny for the shot. Then we don't have to have fear of the grid.

But Bert is not the same. The guy who made Bert did not care to keep out each tiny part. He let each tiny part in the shot, even if the part was far too tiny. What does that mean? Bert has the grid. I do not like that part of Bert, but some love it. Oh well.

Yes. You will get even better results if you use an algorithm built specifically for deconvolving diffraction (e.g. Richardson–Lucy).

When a wave goes into a hole, it can make the shot too soft. The more and more tiny the hole, the more and more soft the shot. One way to fix this is with code. An ipod has lots of code. So does your car. Code can do a lot of good. One way to make a soft shot look good is to use code. One guy, Bill Rick, had a good idea to use code to make the soft shot look good. Leon Lucy also had a good idea. They did work as a team to make code. When a wave goes into a hole, it can make the shot too soft, but we can use the code from Rick and Lucy to make it good again.

Diffraction at low ISO with equal resolution is the very best case for sharpening overcoming a difference in microcontrast. There are many other cases (most, I'd say), where sharpening can't do that. If actual resolution is different, for example. Most commonly, because the point spread function (PSF) of lenses cannot be deconvolved perfectly.

The code from Rick and Lucy can not fix each soft shot. Only the ones that are soft when they came from a wave in a tiny hole. If the shot was made with a lens that is far too soft, no code can fix that.

Noisy images, for another example.

The code they made has some ways that it can not work, too, like when the tiny dots jump all over too much.

A junky lens with sharpening is not enough to replace a quality lens. In the same way, small formats with sharpening are not enough to replace larger formats.

So even if code is very good to make a soft shot into a good shot, it can not beat a good shot all the time. And a big box like Bert can make a very good shot.

 

[Bill Anderson]

Daniel; perspective isn't a function of sensor size. There is no magic in large format's view of the world when it comes to giving you anything outside the information that it stores.
Large format doesn't have an exalted relationship with physics, or the muse.

Quality?: yes, the final magnification factor is important; but this is all moot: who'd want a small sensor when a larger one is available (all things being equal)? Unless of course the larger sensor costs 40K and the smaller one 2K- and you're filming for web. Then you might be forgiven for squeezing as much out of the small sensor as is possible- not to compete with the large, but to get closer to your pre-visualized image. There is no competing with the large: whatever tools or methods are used to enhance the small sensor, they are available to do so on the large one. Checkmate.

 

[Daniel Browning]

Daniel; perspective isn't a function of sensor size.

Did you address me by mistake? There are a few people in this thread who think perspective is a function of sensor size, but I'm not one of them. I'm the one who is telling folks that all format sizes are capable of the exact same perspective.

There is no competing with the large...

Generally, I agree, although I did mention several exceptions above, including low light and dynamic range for FF35 compared to MFDB. (Those are two areas where the economies of scale give an R&D advantage big enough to overcome the size difference.)

 

[Bill Anderson]

Robert, thanks for the hosting offer, but I can't post my final files
because they'd very likely end up as place-mats in a coffee shop
in Outer Dzhibrovia. Besides, they wouldn't tell you that much
more than the jpegs, since my reason for showing them was
to do more with angle of view and depth of field relative to a
single large format image.

Something that you might find interesting: the "Pole and Weathered Message"
image was made using a 300mm lens. And yet it has a field of
view (not to mention depth of field and critical focus throughout)
of a much wider lens because many frames were used for the final.
A "wide zoom" so to speak- a very unusual relationship.
If I had used a single wide angle lens then the trees would have all
but disappeared and would have had little relationship with the pole
and its "weathered Message": broken passages from the bible
(quite legible in the final print).

Longer focal lengths didn't have much of a place in the fine art canon
before the advent of digital acquisition.

 

[Bill Anderson]

Did you address me by mistake?

Oops, that wasn't meant for you Daniel.

As far as sensor performance is concerned, we're talking all things being equal- for there is no point assigning imagined attributes to one (theoretical) sensor over another. We don't yet know what RED is going to do. Besides, there are MANY more practical advantages to large than there are drawbacks at this point in the game.

 

[Daniel Browning]

As far as sensor performance is concerned, we're talking all things being equal- for there is no point assigning imagined attributes to one (theoretical) sensor over another.

Agreed. I was only talking about the MFDB and FF35 cameras that exist on the market now. I did not mean to imply that the same will be true of RED's future products.

Besides, there are MANY more practical advantages to large than there are drawbacks at this point in the game.

For many types of photography, I agree with that.

 

[Roberto Lequeux]

HAHAHA :smilielol5:

I am going to tell my DP that I want to shoot on something that is between Tim and Bob, but hopefully closer to Bob. I wonder what he'll say to me. Thank you Daniel, you made me laugh really hard and I learned more than I would have if you didn't break it down. What more could I ask for? :thumbsup:

Bill, I was thinking more along the lines of a decent size,not too big, and with a massive water mark going across the whole thing. Please don't worry about it though, I've found several examples online. I was just really curious to see your photos because I was drawn to them because of the snow and composition. Good work.

Daniel, so what would you say might be the sweet spot for sensitivity and resolution? FF35? A bit smaller? Bigger?

 

[Daniel Browning]

You're welcome, Roberto.

Daniel, so what would you say might be the sweet spot for sensitivity and resolution? FF35? A bit smaller? Bigger?

Theoretically, the bigger the better: up to as much weight/bulk as you want to carry.

In practice, economies of scale will give some advantages to smaller formats. For low light stills, that advantage is currently with FF35.

Larger formats don't always have to be much that much bigger or heavier, though. Look at how tiny FF35 film rangefinders can be.

 

[Jonathan Payne]

"Larger formats don't always have to be much that much bigger or heavier, though. Look at how tiny FF35 film rangefinders can be. "

A slight deviation and a somewhat extreme case in point: a 6x9 folding camera. I recently acquired a 1955ish Zeiss Super Ikonta c 531/2 MX (I know, 24 characters is too long for a camera name) that takes 8 6x9 images on a regular 120 roll of film. This is going to be my travel cam on my next vacation. There is something very refreshing about this bizarre mixture of features in one camera. Granted this is really a stills only example at this point.

Upside: Medium format film look, large 6cm x 9cm negative, smaller, lighter and easier to travel with than MF SLR or TLR (or a view camera) (estimating from memory i'd say it's 5.5in wide x 3.5in tall x 1.25in deep when folded), sturdy all metal contsruction, relatively sharp tessar lens. Easy to walk around with all day in smallish P&S camera bag or if I were of the fanny pack persuasion, that would work perfectly.

Downside? Still weighs approx 2lbs, Fixed focal length, single lens coating, not low light capable (F3.5 min). Slow to operate.

For 250.00 in moderate to daylight lighthing, this thing trounces any digital point and shoot or small form SLR at an equivalent focal length. It packs a very large negative in a relatively small package.