How much better will digital get?

[Ben Rubinstein]

I'm afraid I can't compete in this discussion as far as the math or physics goes, however I found that shots I had taken using a Mamiya 645 with a 210mm lens set at f32 were incredibly sharp and easily beat similar sized 35mm prints shot at f8 (same technique, great glass, blahdiblah and yes I had knocked it of f22 by mistake!).

I was very suprised at this, but to me it makes sense. At 18X12" 35mm runs out of steam, therefore this could be said to be a pretty large enlargement of the frame, which will show any faults and problems.
645 however handles a 16X12" print without even getting warmed up. It is not at the enlargement stage yet where faults will show, or at least make problems.

To my mind, although there may be the same diffraction at f32 on both medium format and 35mm, the diffraction will only show up as the same if you magnify the image the same amount. If you magnify a 35mm frame by X to get a print size where diffraction is noticeable and a problem, you have to magnify the med format print by the same X before you see the same diffraction. The med format print will of course be much larger.

35mm can only be enlarged so much before it gives up for multiple reasons, sharpness being only one of them. I can therefore enlarge a med format frame more than a 35mm frame, and still not have reached the X point where diffraction is a problem.

If I can enlage med format more than 35mm, even for digital, then med format is better for prints over a certain size despite having to stop further down. I'm sure this is even more apparent with 8X10"!

Of course the mathematicians may say I'm wrong, I'm just looking at the prints.


Mamiya 645 210mm f32, every single twig is razor sharp on a 16X12" print...

[didger]

Good lenses will probably be the greatest factor in image quality once sensor technology matures

What do you meant "once sensor technology matures"?  1ds is already beyond most lenses and 1dsMKII is way beyond all but a mere handful of lenses, none of them the wide and ultrawide lenses I care about most.

I think he's got it! This almost uniform increase of various relevant length scales in my main point.

Yeah, but as Jonathan explained, you don't get relatively more diffraction problems with larger formats, since it's absolute aperture diameter and not the f stop value that determine this.  The smaller f stop value required for comparable DOF for a larger format is no problem because that larger f stop value is NOT a smaller diameter in absolute terms.  

If you additionally take into account your observation that larger lenses are relatively easier to make because absolute machining and grinding accuracy tolerances also get relaxed proportionately with absolute lens size, then it's win/win every which way for going to a larger format.  Moreover, 35mm format has been seriously compromised by consumerish gimmickry and convenience features that rule even in the rarified $8000 DSLR market.  You can't buy an ultrawide prime for EOS and ultrawide zooms just don't quite hack it, with only rare individual sample exceptions, and then never for wide open apertures.  For MF there's practically nothing made except pro quality prime lenses and they're not even expensive and there's lots of used ones around dirt cheap.  Oh, boy, I can hardly wait, though you won't be seeing my 1ds stuff on ebay either.  1ds is great for now, but only for now.

[howard smith]

didger, because yo udon't own a TV, amybe you aren't aware, but:

For the best camera/lens/tripod/printer, there's MasterCard.

"... do free schlepping for the privilege of being in the presence of photographic fame."  Priceless.

Gag me with a spoon.  Please.

[Ray]

Where is there any grounds for optimism that there will be lenses to adequately match 1dsMKIII sensors or 1.6x sensors of ever higher sensor site densities?  It's all fantasies.  The only way upward is medium format, and hopefully some day 4x5 digital.  

Well I agree generally, except I think it's a fallacy to believe that current 1Ds and Mkll sensors are so good that they've left the lenses behind. There's no reason to blame the lens for inadequacy any more than the sensor. Improve either one of them, not necessarily in pixel count for the sensor but pixel quality, and you'll get sharper, better images.

[bob mccarthy]

Peter, I like all your ideas and maybe most of them are even doable, but not for free.  However, what are "legacy lenses" that would allow you to use 25 Mpixels effectively in 35mm format?  I don't know of any lenses shorter than 35mm that can even keep up with 16 Mpixels or more than barely 11.  What lenses are you talking about?  If Zeiss distagon lenses are "legacy", I'm afraid you'll be disappointed in the ultrawide realm.  

My belief is that post processing could yield big improvements with a marginal situation. I've seen it in audio and video, why not still photography.

DxO (or something like that) is just the beginning, I would assume. A film camera is somewhat of a binary device, either a grain of silver gets enough light or it doesn't. Its just that their are multiple layers of silver (maybe better to say depth than layers) in an emulsion. With a sensor, its just one layer.

 Their can be a number of ways to gain info. Give me enough data points and I can easily guess whats between. The bayer filter(?) has to process like that. I believe there is much to be gained by "in camera" and "post processing. To me anyway there's a great deal of advances available in the area.

Processing could mitigate the whole lens/ sensor size conumdrum.

Question: How does the "well" work. Is it a counter or a reservoir?

Bob

[Bobtrips]

Howard -

You're an engineer, are you not?  If you've got the background, what's your take on the ability for modern technology to produce better, and more consistently better lenses?  

Why, for example, is not every element ground to the highest precision?  Why is not each element precisely placed in the barrel?  And adjusted as necessary?

We're moving single atoms around these days.

(Maybe the issue is cost of tooling.  Or maybe I have too much faith in clever people....)

--
And let's not forget that smaller lenses are easier to make.  Are they not?  My entire trip here is about making smaller high resolution cameras rather than pursuing the old path of 'Bigger is Better'.

[howard smith]

"Well, we know that lenses as good as a Zeiss can be made.  

So given enough market demand one of the manufactures is going to revise their manufacturing process to crank out a flow of them."

Likely true.  Reading forums like LL gives the impression the market demand is there.  But LL readers represent a very very small fraction of the camera public.  How any 1ds bodies does Canon sell per every sub-5mp camera?  I have no idea, but my guess it isn't many, and Canon doesn't need 1ds sales to stay in business.  I suspect the 1ds type cameras are a subsidy for and a platform for R&D.

"A third party battery (presumably just as good) cost $100. A bit of research on the net can locate a supplier of another third party brand costing just A$35. Hope my camera doesn't explode one day. If it does, I'll probably think, 'you get what you pay for' "

The world economy is not perfect.  In a perfect economy, I could not sell an equivalent product for a higher price.  There are people who will buy "Canon" (or whatever) at a higher price an an equivalent or maybe even better product.  Then there are who will buy a more expensive Nikon product because it is in effect cheaper (I have all these Nikon lenses, so I can't a bettter Canon 1ds and all new lenses.  I'll settle for a lesser Nikon body.).  Even Canon competes with Canon by selling different versions of the same product.  How many similar zoom lenses does Canon make?

A megamegapixel 4x5 back will likely be severly limited by today's or tomorrow's LF lenses.  The resolution of any lens/sensor combination will never be as good as the weaker link.  A perfect back would no better than a really great one, yet cost much more.

And in the final analysis, it is the print that matters.  That makes printers a limiting factor.  Then there is the human.  I don't expect any great advances in human eye sight soon.  In fact, I think it will decline, IMUO.

[howard smith]

I just had a recollection.  About 40 years ago I saw a photo of two men plying golf.  I sould easily see the golf ball on the green.  It helped to tell it was a golf ball to see the green and two golfers.  I could see the dimples or the name on the ball though.  The most amazing thing to me was the photo was taken with a MF camera on an airplane 65,000 feet above the golf course and traveling at 1200 mph.  That was 40 years ago and film.  How good can digital get?

[Ben Rubinstein]

I used to give the negs anyway, it was my selling point. What I meant was that the price included film + processing, it wasn't invoiced seperately.

I don't charge less when shooting digital, the money comes to me not the lab that's all, still, if you worked it out in $ per hr it's not a huge amount for your computer time.

[Ray]

I don't want to sound disagreeable

No problem. I always try to look at the strength of the argument. I can only speak of what makes sense to me.

If we were to devise a test chart for the R pixels alone, it might consist of dots of shades of 'white' (actually pale grey), all of which would be of greater intensity than the S pixels could handle.

If we were then to remove all the S pixels from the sensor, because they wouldn't be serving any purpose, and record a scene containing, say a hundred different shades of white, many of those shades of white would fall in the rather large spaces between the very small R detectors with their small microlenses.

The small R detector is serving some purpose. It wiil pick up a proportion of those dots (which could be described as representing the texture of sunlit concrete), but will ignore the dots that fall in between.

According to Fuji's schematic diagram of those pixels, shown in Michael's review, the distance between the R pixels is much greater than the diameter of the S pixels. I would therefore deduce that most of the excess photons constituting blown highlights will fall on stony ground, ie. serve no purpose.

[bob mccarthy]

hehe, I know your right. Most of my digital capture for work does not require any more than a lower/mid range dslr. 6 megapixels are more than enough. I just have a project later this year thats going to require more. The d2x will get me there. I hope anyway. If not, I shoot film,

I sense Fuji is on the right path

[Bobtrips]

However, what i am saying is that we have made this mistake before with computers. 10 years ago, no one would image we need a 1ghz processor, now, people are talking about dual 2.5ghz.

But there's a basic difference between cameras and computers.

My early desktops - Apple II types - that ran at a blazing speed of 6 mHz were plenty for word processing and spreadsheet work.  I'm a fast typer but I couldn't type fast enough to overload the buffer.

But new uses for computers developed.  Graphics and games required a lot more speed.  Over time these got more sophisticated and required even more speed.  Computer manufacturers have been pushing to keep up with these demands.

Cameras take pictures.  Most people have a maximum print size that they very rarely wish to exceed.  We're reaching the point at which each of us can find a camera that meets our printing needs.  

It's unlikely that there are going to be new uses for cameras that will cause them to drastically change in the way computers have.  (Sure, there will be needs for better video, better camera phones, but those are different tools.)

[Sfleming]

Jo,

Thanks.  I just  emailed Mr. Fuerbringer.  We'll see what happenes.  I found this site on Plaubels:

http://medfmt.8k.com/mf/veriwide.html

Know any other  good ones?

Yes I could see just  from my limited  study that  tha later  f/5.6 47mm SA  would be preferable.  So the camera was never originally made  with  this  lens?  As to your Horseman back?  What  is  the advantage to this?  220 over 120?

[Sfleming]

Come on Quentinn and BJL...

Are we honestly discussing present day reality .... or just  tryin to win arguements?

It's not just the press pit and both of you know  it.  It's fashion and product and fine art  and wedding and and and and and .....

The GD 1Ds II is so much better than ANYTHING outside of the 400% more expensive and clumsy by comparison DBs that as I said ....

It virtually stands alone.

This is  a problem and it is  half the curx of what I'm always whining about.  It's gonna ruin much if not most  of this game.  Can you imagine if there was only one  brand of car?  Home appliances?  Clothes?  etc.

Can you imagine how much more of a dick Bill Gates would be if there  was no Unix and Apple?

Thank goodness they are charging such rapine prices for the 1Ds and the MF DBs.  If forces thousands like me to shoot  MF and LF film.  Thus the  'end of the world as we know it' is postponed.

[Sfleming]

Some little upstart like Foveon could make a breakthrough but they  will be bought out or forced to keep the prices high.  

Foveon is manufactured under contact by National Semiconductor.  Hardly a little upstart, that National Semiconductor.

So what?  The big boys have already got the whole Pro world anxious to pony up $8k for a lousy camera body.  The 'cream of the crop' are willing to cough up $30k for a back.

They will back off these figures exactly as far as they are FORCED.

I don't see any army on the  horizon.

[didger]

If you magnify a 35mm frame by X to get a print size where diffraction is noticeable and a problem, you have to magnify the med format print by the same X before you see the same diffraction. The med format print will of course be much larger.

This is the "common sense" assumption I was making all along as well, but with no personal MF experience.  However, all this math stuff (in which I'm also not knowledgeable) apparently insists on just the opposite; namely that you'll see diffraction and DOF limit blurriness at the same print size in both systems.  You would presumably only see the superior sharpness of MF at optimal apertures (NOT f32) and then only in a very small area of perfect focus.  I'm inclined to go with your print rather than the math and I'm still not discouraged about a 22Mpixel MF DSLR in my future in the next year or so.

I like the tunnel through the trees shot; it looks like a study in organic fractal geometry.  I don't understand fractal math either, however, though I may hold the world's record for number of fractal images generated as animation frames.  Thank God that you need neither advanced math nor any rules of composition to do personally satisfying and commercially viable photography, music, and art.

[BJL]

Everything increases proportionately as image size and lens size gets bigger.

I think he's got it! This almost uniform increase of various relevant length scales is my main point.

Except that the length scales which lenses can resolve probably do not scale up quite so fast, giving a larger format at least a bit of an edge. Good lenses will probably be the greatest factor in image quality once sensor technology matures, and bigger has an edge, since manufacturing tolerances do not have to be quite so tight.

[didger]

Well, considering the incredible complexities involved in choosing and using these tools competently, I feel truly fortunate that I ended up with a kit that's working as well as it is.  I don't see anything whatsoever available in the world today at anything like a price I can afford that I'd rather have, and that even includes my Apple Cinema monitor.  I never calibrated it, taking the word of a pro photograper friend that said they're good enough straight out of the box.  Well, just a few minutes ago Jonathan's 24x36 print that he made for me arrived and it's incredibly close to what the monitor shows and also incredibly sharp.  The sharpness and absolute absence of any artifacts from pretty massive processing really amazed me.  Of course, I had to do a lot of cleanup before I sent the file to Jonathan, but the whole procedure from shooting through printing really worked.  If 22 Mpixel MF turns out to be twice as good as 1ds, maybe I'll never need to think of going beyond that in my quest to rival David Muench 4x5 prints for technical quality.

Now if I could just train a swarm of bumblebees to carry some of my stuff for me; or get famous so that college students will do free schlepping for the privilege of being in the presence of photographic fame.

[Bobtrips]

However, because photonic noise is equal to the square root of the signal, we can arrange for it to be less significant by making bigger pixels that receive larger signals, which is the general idea of large format.

Consider a really big pixel on a large format sensor that can absorb, say 1 million photons. Photonic noise will be 1,000 photons or 0.1% of the total signal.

Consider a really small pixel that can absorb only 1000 photons to maximum well capacity. Square root of a thousand is 32. Photonic noise is 3.2% of the signal. A big difference.

Are you saying that large photosites can generate a stronger signal because their greater surface area will be struck by more photons than a small photo site?

Or are you saying that large photosites simply have more 'capacity'?  

(Or some combination of the above?)

[Bobtrips]

I haven't followed the DxO stuff to any extent.  But isn't it based on the concept that the data can be tweaked to make adjustments for lens problems?

OK, put each lens on a good test bench, fully automated.  Analyze the lens.  Store that correction data in a chip embedded in the lens.

Teach the camera to either use that data when building the file or to attach the data to each file for in-computer use.