Tuesday 17 February 2009

fixing a floppy Bessa folder

readers of this blog will know that I quite like my Bessa I and Bessa RF folders. They are great cameras but can be less than ideal producing soft images at times. I'm an analytical kind of guy so I quickly applied a little investigation into this camera to sort out why its not so perfect.

It turns out that a little wear had occurred and some play had been introduced in the rigidity of the front element. Carefully folding and unfolding it I noticed that this part of the mechanism which completes the locking down of the 'front standard' needed assistance.

lens support

If you grab your camera (assuming your a Bessa RF owner) you can see here the place I mean. The suraces which the red arrows point to meet when unfolding, and the
arm with the white masking tape on it merges in behind that "can opener" shaped one to lock the front standard (when thinking of the bessa as a non adjustable view camera). This had developed some wear in the mechanism and was allowing the front element to swing a little. This would be desireable in a view camera but its not what you want in a rangefinder ;-)


How I've solved it is to place essentially a small shim over the small arm there. In the photo above I've used masking tape to test the idea, but I've now settled on another simple method that is non-destructive and reversable. I have bent a small section from a aluminum beer can over the arm and secured it with a thin bead of blu-tac. I put the bead of blu-tac it on the inside of the aluminum along where it will sit on the arm. I used the edge of a metal ruler to form the section of aluminum shim so that it folds in a neat "3 sides of a box" and sits neatly over the arm.

This rig now alters the Bessa to have a nice solid front and has returned a snappy feeling to the unfolding (rather than the wobbly feeling before). I've checked the alignment of the lens carefully with both film and ground glass and its been working fine ever since.

Here is a sample of the full width of the horizontal middle of a 6x9 frame at f8



I think that you can easily test your front element by
  1. unfolding the camera as if taking a photograph
  2. opening the film door of the camera exposing two light sealing plates at the back (these will allow the camera to sit on the film plane with the lens facing up
  3. place a flat and light spirit level across the front of the lens

using this you can get an idea how far tilted the lens is from the focal plane (measure in X and Y axis to make sure).

My Bessa is now pretty spot on (as can be seen above)

Monday 9 February 2009

lens tests: looking for better pictures, what's missing?

Amateur Photographers (perhaps especially landscape photographers) are often almost obsessed with how to get the most of scenery detail into their pictures. I guess that its because most photographers seem quite attached to their particular equipment (in some cases almost religiously) that the first thing examined is getting a better lens. So sites with lens tests are consulted, to see if this or that lens has more or less resolution than another seeking a better lens than they have.

This seems like the right path but it misses out on understanding a couple of critical points.

  1. Lens test charts are not scenery.
  2. There are other limiting factors in the system.

Perhaps you've read this before, in my reading in the past its never been shown clearly just what this means, so in this post I'd like to shed a little light on thes important points.

It may seem (later in this comparison) that I'm attempting some sort of film vs digital discussion, and I'm not. I'm trying to discuss something bigger ... in fact its bigger and denser capture. I just happen to have only MF and LF in bigger capture.


Part 1: Lens Testing charts

To the left you can see the result of a lens test using Norman Koren's test system. The lens being tested is my "nifty fifty" the EF 50 f1.8

Looking at this it seems that the lens runs out of grunt shortly after 50 line pairs / mm (line pairs per millimeter), and you could argue the toss that it goes to 60.

This is not a bad result with some saying that getting better than 50 lp/mm in the real world (with all the factors like camera shake, focus issues .. against you) is as much as you can get. Being a digital system what I've got here is a final usable result which will directly translate onto a print.

(BTW If you're not familiar with terms such as line pair / mm, I suggest reading this.)

So I thought that I'd test also my EF 24mm f2.8 lens (which in 35mm full frame is my favourite landscape lens). At left is a 100% crop of an image taken (in RAW) using flash as the primary illumination (for increased contrast and to eliminate any issues of shake even on a tripod).

Again not a bad showing for what is not thought of as the hottest lens in the canon stable. It seems to run out at about 60 lp/mm.

This all looks good, but what I didn't mention yet is that the method for the test is to photograph the target at a distance of 51 x "focal length".

Since this is a 50mm lens, this means that the target is 2550mm (or 2.5 meters) from the lens or 1.2 meters with the 24mm lens.


Now, just think about this for a moment:

How often do you stand that close to the subject when you take a picture with a 50mm lens or 1.2 meters with the 24? Especially with an APS sensor DSLR (not a full frame one) it is just what you'd do for a portrait and not what you'd have in most scenery.

What is more common is using a 24mm a few more meters away in a scene like this:


So, I put the lens test chart over there on the wall ... to see what I'd get ... well here's the 100% view if it..

Wow, nowhere near as good is it.

You can see that its not giving the wonderful results of the test chart above. Well, that's because its now further away in the scene, and the subject detail is now smaller (being further away). Inspection shows that we can barely read the details just past 10 lp/mm.


So from this perhaps you can see what I mean when I say: lens test charts aren't like the real world).

So knowing how the lens is able to perform is one thing, but what you can get with it in the real world is another thing. It all comes down to what is the "feature size" in your picture and how does this relate to the parameters in your testing.




Part 2: " other limiting factors in the system"

Ok, so now that I've shown the issues involved with "Lens test charts are not scenery" I'd like to move on to "other limiting factors in the system"

Consider your camera, a camera like the 10D (or the 5D for that matter) will have a limit of something like 50 lp/mm because of the array of the sensors (see Bob Atkin's article on that here). Look again at the details of the 24 and 50mm lens above, notice how they coincidentally run out of detail at almost exactly the same point?

Sure cameras like the 20D and 50D have higher detail sensors than the 10D, but this essentially means no matter how much better lens you get, you'll see no difference without spending more money on a higher resolution body.

Welcome to the one of the reasons why camera companies love digital cameras.


Conclusion: Alternatives to the lens arms race.

So, if you're going to spend money on sharper lenses, you'll need something better to capture that. You could use a Canon 5D II or something better spec (Nikon and Canon both make things like that. There is even the alternative of the MF digital backs.

But its all starting to sound expensive for the landscape photographer isn't it, here's where I'd like to show another test with a different camera system.

I'll start with my old Bessa RF camera, this is a 1940's folding 120 roll film camera which uses a 6x9 format *(6x9 if you're not familiar with film and its sizes the please take a quick look at this article which sheds some light on that topic).

The Bessa is by no means any kind of bench mark, in fact its about the cheapest 120 roll film on the market. My Bessa RF can be had for less than a new lens for a DSLR (I see them on eBay regularly for between $100 and $300).

The Bessa RF uses a 105mm lens and so using the same test target and system I need to position the camera 5.36 meters from the target. This is a scan of the entire 6x9 cm image on the film (pardon the gudge at the top, that's from my tape where I taped a bit of film into the camera because I'm too cheap to sacrifice a complete roll for a test image).

Now, another important point here is that these film scans were made with an old Epson 3200 flatbed scanner, which cost a couple of hundred bucks. There is no doubt at all that with a more professional scan (send it out if you have a beauty which begs to be printed big) you will be getting even better results than shown here.

Now, this time I'll start with the overview image of the test setup.

Did you notice that this is exactly the same angle of view in the scene that I used in my "real world" use of the 24mm lens from above.

Why? well because of the bigger format film the 6x9 to get a wider angle it needs a longer focal length lens. In this case its 105mm.

So in some ways this old camera is designed to take best advantage of real world photographic and optical issues to make the most from what it has.

Using the same method above I tested the lens on this camera and found that it had a much lower result than the EF 24mm in the lens test.

Looking at the result (below) perhaps you can pick its somewhere just after 20 lp/mm that the lines merge into mush.

So in test setup the Bessa scored 20 while the 10D scored over 50. This makes the Bessa look bad and the 24mm on the 10D look good ... but only if you forget to look at the scene and focus on the test result numbers.


When comparing scene for scene the older camera actually makes a sharper image for the same scene than the camera with the lens which tests better.

In a usable scene the Bessa ended up with 20 lp/mm available in the scene VS only 10 lp/mm for the 10D.

So what small formats gain in terms of specifications, they can also loose in terms of what you can get in your image.

Thus you'll end up with features that are much finer in this image than in the 10D's image. How interesting, a 60 year old camera with an ordinary lens is doing better in some ways than a modern DSLR.

Why? Because what the Bessa lacks in sophistication and accuracy it makes up with in generous lashings of image area.

Hopefully I've shown that by using the lens test charts for reference that taking images of real things shows up the issues of buying gear based on lens test charts and there is more to it than just simple numbers. I have an image which has the same "angle of view" using a lens which 'tests to much lower resolution' but in the real world makes an picture which is much more detailed than another camera system. Here's the over view of each again to show you the relative sizes.



So if you're thinking that you'll spend a thousand bucks on a 'better' lens for your DSLR to make sharper landscape images, perhaps you might now consider adding new tools to your methods of making images not just spend more money on nifty lenses for your DSLR.

This represents only the beginning of the scale in MF gear, and with great cameras and systems out there for literally give away prices its really hard to go wrong.

But if you go for a higher quality MF camera (to make higher quality images) you'll need a higer quality scanner in your work flow. I don't personally feel like making that sort of investment, so I went to another alternative again. Large format.

Size matters when it comes to images as I've just shown with the comparison between a 6x9cm (in reality 56mm x 87mm) image negative and a smaller 22 x 15mm. I prefer larger formats again for my landscape photographs. Using the smallest of Large format 4x5 (which has an image area of 92 x 118 mm) I can get even larger capture, beter control over my image focus and still make use of cheap scanners to get great images.

Here is a sample of the same scene with my 4x5 camera mounted on the tripod.


Because its using a 90mm lens (quite close to the 105mm in the Bessa) its about the same distance from the wall. The angle of view though is significantly broader, showing in landscape the same height that the others showed in portrait orientation. No looking at the detail, obtained from the lens its nearly the same (although being a 90mm it was a little closer to the target than the 105mm lens, creating a little bigger image of the chart).


this is almost the same result in sharpness but with a system which captures more than double the area of view than the digital. For example this picture

?

has detail like this:

?
click for full detail


and as its 14 000 pixels wide (while a DSLR will give you about 5000) will make nice detailed prints well over a meter wide if you want it.


I hope from this long (and perhaps difficult) analysis of the subject you can see that the bottom line is that if you want to make bigger more detailed prints of your favourite scenery, then rather than going to all expense of newer and 'better' lenses you will end up making much sharper prints if you use the right tools for the job.

the bail out

According to the BBC

The US government has unveiled a $250bn (£143bn) scheme to purchase stakes in leading American banks.

Announced by President George W Bush and US Treasury Secretary Henry Paulson, the move is the first stage of the government's wider $700bn bail-out plan aimed at rescuing the US financial system.

I was just reading that Barry Ritholtz, author of Bailout Nation, had put together a historical comparison of the cost of this to other substantial outlays. This will apparently be “the largest outlay in American history” costing more, in inflation adjusted terms, than all of the following historical expenditures:

  1. Marshall Plan: Cost: $US12.7 billion, inflation adjusted cost: $US115.3 billion
  2. Louisiana Purchase: Cost: $US15 million, inflation adjusted cost: $US217 billion
  3. Race to the Moon: Cost: $US36.4 billion, iflation adjusted cost: $US237 billion
  4. Savings & Loans Crisis: Cost: $US153 billion, inflation adjusted cost: $US256 billion
  5. Korean War: Cost: $US54 billion, inflation adjusted cost: $US454 billion
  6. The New Deal: Cost: $US32 billion (est), inflation adjusted cost: $US500 billion (est)
  7. Invasion of Iraq: Cost: $US551billion, inflation adjusted cost: $US597 billion
  8. Vietnam War: Cost: $US111 billion, inflation adjusted cost: $US698 billion
  9. NASA: Cost: $US416.7 billion, inflation ajusted cost: $US851.2 billion

Total cost: $US3.92 trillion

I'm no economist, but gosh, wouldn't it be cheaper just to give them all a decent social security pension and then let the market get back to being a free market?

What happened to Mr Smiths invisible hand?

Monday 2 February 2009

film VS digital capture: revising old work

Setting the Scene

An important thing for learning and discovery is that as you learn more its handy to go back and revisit old works that you've done. Soon after I got a 20D digital SLR I was interested to know how well it compared with my (used for a long time, comfortable with the results) 35mm film. It didn't take long to sus out that the smaller sensor on the 20D (compared to 35mm film).

So, soon after getting my new toy I wanted to test which was better (as clearly I understood which was cheaper to take photographs with ;-)

Initially (we're talking 2005 here) I used only ACR to convert the RAW files and had to process then in Photoshop to make my images.

Since I like to take an image with my camera that looks like what I want (not simply what the lens forces me to) it was not easy to compare the two systems I took an image with each camera using two lenses from the same spot for evaluation.
  1. Canon EOS 630 EF 50 f1.8 (@ f8) I used 200ISO negative film.
  2. Canon EOS 20D EF 18-55 @ 35 (@f8)

My first results were as to the left with quite different colours on both systems. Initially I felt that the image from the 20D was more neutral and was impressed with how close they were. The shadows rendered on the digital from the 20D were cleaner than the film and there was over all.

So while I liked the digital image for its clean appearance I thought that the film image had better colours and tones. Note: while it was much clearer the 20D image was much smaller than the film scan.

I've downsized the Epson segment to match the native size of the segment from the 20D and pasted them side by side here to see.

The image from the 20D seems sharper but you can't get it to show as nicely on a print because 1) you can't see 100% view on the paper without a magnifying glass, and 2) if you print it so magnified then the pixels look funny.

So essentially while it has fewer pixels they're 'nicer looking', so you can upsize it about 70% to effectively the same size (as the 20D could tolerate uprezzing to the same print size as I could get from the film). I placed that discussion here on my personal pages which contains all the details.

I didn't have access to a good scanner at the time I did the above mentioned www page, I only had an Epson 4870 scanner and so when I eventually sent the negative to a friend with a Nikon LS-8000 I was surprised how well it did. Below is a sample from each scanner (I've resized both to 4000 dpi) they are closer than I expected.


The Nikon brings the balance back more towards film being a better image than my 20D in both tone and clarity at a large printable size. So now with the background set, lets look at the tone mapping image.





New Tools applied to old RAW files

Recently I've started using Photomatix for making better "goes" of rendering RAW images. I've been pleased with the results (as in that blog entry), so it recently occurred to me to revisit the above tested RAW image and compare it with the film scan.


My first thought when I got it how I wanted it to look in Photomatix was that looks really good. then I thought: I wonder what it would look like compared to the Negative.

So here it is ...

Surpirsed? I was!

The tone mapped RAW image brought out the sorts of contrast enhancement which the natural 'organic' feeling compression (heel and shoulder) that negative film does naturally.

But the now the digital (which does a better job of recording low contrast details than film does) brought out the clouds and textures of the clouds better than my negative and certainly better than my ACR conversion had done.

This scene is quite a low contrast light scene (taken in Finland in late spring) which neatly fits within the range of a single digital capture. Negative on the other hand would be able to cope with a much brighter or contrasty scene because it would be able to still expose for the shadows and the highlights would not clip as the digital would.



For those with an eye (or interest in) the details I've put a 200% magnification of the tone mapped 20D RAW file below to compare with the film scans above.



I think that it shows that tools like tone mapping goes close to bettering what the results from scans of negative film can achieve with far less expense and technical (photoshop and or scanning) skills needed.

So, 7 years into the age of the digital SLR we now have the tools to get the image from a high quality APS sensor digital SLR into the same ball park that supermaket grade film and a good scanner can do.

I've phrased it like this to draw emphasis to what I believe is an important perspective:
  • that the tools are both very similar in what they can do
  • as the digital tool sets continue to develop (with tools like HDRI and Tone mapping Photomatix) we can make better images from our RAW images (another reason to use RAW)
  • that as I come to learn more about digital I find how really good the results were from my film images
I used to argue that the main drawback against digital was the upfront cost of investment, back when I got my 20D (2006) it was AU$2000 for a body only and that would perhaps also require you to buy different lenses (because a 24mm is no longer wide and a 50mm standard becomes a telephoto) if you bought one. At the same time I could buy a Nikon LS-V scanner for AU$700 and get scans as good or better than the one above from the LS-8000. Back then I argued that one was better off to get a film scanner and keep using you 35mm SLR.

But since 2006 that field has changed, with used 20D's and equivalent cameras to be had for less than $500. Its nearly a toss the coin decision.

If you already have a compact point and shoot digital camera I think it comes down to how much you expect to use your camera for serious images and perhaps personal choice.

But still an old 35mm SLR like my 630 has a much bigger and brighter view finder (helpful for critical focus) and with just a 24mm and a 50mm lens makes a much lighter travel companion that my DSLR with the lenses it needs.