Depth of Field - Apps with Depth of Function

June 07, 2013  •  2 Comments

I have tried almost as many DOF apps as I have camera bags. Really. It's that bad/sad. And as with camera bags, DOF apps never really seem to get it quite right. Either they are not sophisticated enough, or are insufficiently flexible, or they work from assumptions that went out with the Ark. Sometimes, their developers just give up the ghost, failing to update their apps in the face of an onslaught of new cameras, lenses, sensor sizes and resolutions.

So what was needed, from my personal POV, was an app that had the following two chief characteristics:

  • Parametrically independent of specific equipment (no need for camera specific app updates).
  • DOF definitions that can be tailored to the user rather remaining mired in the past.

Browsing the App Store recently I found a new candidate, TrueDOF-Pro and, even more useful, its sibling OptimumCS-Pro. I have been trialing them extensively and, though they are not yet perfect (compared to me 'dream team' fantasy) they are already very useful to me on a daily basis. Even better, their Author, one George Douvos, is astonishingly friendly, knowledgeable, responsive and keen to engage with his customers. So things can, as they say, only get better.

Before looking in a bit more depth at each app specifically, I want to make a few points which are general to both and which indicate the originality of thinking involved in their conception.

Traditionally, DOF has been defined in terms of Circles of Confusion. I won't go into the theory, you probably know it already and it is too long a subject to trot out again here but in short, old-style DOF calculations looked at what precision of level of detail the human eye could not discern beyond in a print, given certain assumptions about what distances the print would be viewed from, usually in terms of a multiple of the diagonal dimension of the print. Hence the laughable but widely held belief that people in galleries automatically step back to a viewing distance in proportion to the size of the image. They might well do for a moment. But then they will get up to as close as they can comfortably focus, probably a foot or eighteen inches, and stare intently before stepping back again. This behaviour is particularly evident in experienced, big-budget purchasers of high end fine art photography.

Now this is absolutely NOT to say that they won't buy an image that is less than technically perfect at nose-distance: they are a sophisticated audience, quite capable of discerning whether the creative intent of the image does or does not require enormous resolution. But there are photographers out there (Burtynsky, Crewdson, Kander, Gursky and many, many others) who make and sell very large prints at very high resolution.

These large print sizes (say, over a meter on the long side and sometimes much larger) do not just test the peak resolution of the imaging chain: 'print big, look close' also places enormous strain on the DOF in the image. DOF needs to be treated by both makers and consumers of photography as part of the creative intent of the image, and as such it needs to be understood, controlled and accurately deployed. For some images it is in effect a production tolerance.

Which is where Mr. Douvos comes in: his apps let the photographer set their own parameters for what is 'acceptable' DOF rather than relying on ancient tables or meaningless scales engraved around the edges of lenses. Better still, they allow for those parameters to be conceived of and entered in terms either of traditional COC or, IMHO more usefully, print size and resolution. And all this in a system that, uniquely, factors the effects of diffraction into the equation - so finally COC is truly useful again.

This diffraction accounting is very important: with high resolution digital systems the photographer can no longer count on stopping down to F22 (let alone F32 or F64) to get the required DOF. F16 is a desperate measure for most of us and many won't even stray beyond F8 unless we really have to. The Medium Format Rodenstock 40mm HR lens that I will use in some examples later in this piece comes with the manufacturer's recommendation to shoot only at F5.6 and F8 for best results. But traditional DOF scales just keep on in a linear fashion to F64 with no concern for diffraction. 

Not any more: both these apps will take diffraction into account and I do believe that is 'a first'.

In detail:

TrueDOF-Pro

Screenshots tell most of what you need to know:

 

photo copy 8 The main working screen allows you to enter your lens focal length, aperture and the shooting distance that you have set. These, calculated according to the parameters you have chosen in the settings screen, determine the DOF readout. In the example above, my 40mm lens and aperture of F8 combine with my the shooting distance I have set of approximately 2.5 metres to suggest that everything between two and three meters will be in 'acceptable focus'.

Notes: 

  • By pressing the 'S2' or 'S3' buttons, you can choose to show the distance scale with more of an emphasis on mid or far distance work.
  • Pressing the 'HF' button sets the focus distance to its optimal hyperfocal distance: in other words the 'far' focus range (top red line) is set to infinity and the 'near' (bottom red line) is set to wherever the app calculates it - assuming you set your camera's focus to the distance indicated by the blue line, which is the indicated hyperfocal distance. The numerical readout windows are self-explanatory.

Simple, eh? Let's take a look at the setting screen:

Using my preferred parameters, it looks like this:

photo copy 7 This requires more detailed explanation.

  • The key choice is to select between 'Blur Spot Diameter' and 'Print Quality'. The above screenshot shows 'Print Quality' mode selected. I will come back to this.
  • Next, using the top left buttons, set the physical width of your sensor.
  • Then mid left, your intended physical print width.
  • Finally, bottom left, your desired print resolution.

This is all simple with the exception of that final point: print resolution. It is expressed in terms of Line Pairs Per Millimetre and this is not a measure that will be familiar to all (or even most) users. It took me quite some time to calculate how to enter this value, taking as a starting point the measure of resolution with which I am most familiar. Let me explain.

When I make an image with a view to making and selling a fine art print, I generally shoot with a final print size in mind. Of course sometimes I can't know this in advance, and sometimes an image made on a 35mm sensor might be good enough as a piece of work for me to want to print it larger. But in general I shoot with the following in mind:

  • 35mm sensor 18 to 25mp  - 22 inches (equal to between about 250 and 272dpi).
  • 35mm sensor 36mp - 34 inches (equal about 216 dpi).
  • Medium format 80mp - 100cm (equal to about 258 dpi) but with the possibility of up to 125cm (equal to about 206 dpi).

The above list is absolutely full of human inconsistencies and quirks: the smallest print size is for historic reasons that I have almost forgotten but relates to a show I had where most of the images were shot on a 5DII but some were on smaller format cameras and I needed a compromise print size that worked across the board. It then became my standard size for editions of prints from frames shot on 35mm. Then along came the Nikon D800 and, if I was showing only work from that camera, I would go as far as 34" on the long side because that gives an image about as large as I can print with margins on my 24" roll printer.

The medium format print size is determined by the fact that from experience, when I send these frames out for commercial printing on larger format printers, 125cm is about the largest I can go and get a print to the quality that satisfies me when viewed up close,  at a distance of around 12-18 inches. On top of all that, I flip flop quite happily between inches and centimetres, having been educated in both systems. This is useful because common photographic calculations also flip flop between imperial and metric.

In other words, my criteria are a mess. And I don't think I'm alone in that. 

However, there is one clear message that comes through: for a good camera with no AA filter and assuming a good capture, I know, reliably, that I will be happy to print to around 210 dpi (actually if pushed, I will go as low as 180 dpi as long as the capture is really good and I am printing on textured fine art paper, which is more forgiving than gloss). And it is that base level of  resolution that I need to use in my DOF calculations. In other words, if an object in an image appears acceptably sharp when printed at around 210 DPI* from an un-cropped frame, then I consider it to be 'in focus' - which means 'within my DOF'. So a DOF calculator which lets me specify print size and resolution is absolutely perfect for me.

Trouble is, my head for resolution, like most photographers', works in DPI (accurately, PPI but for this purpose they are the same thing) but these apps work in line pairs per millimetre.

Here is an example of how to convert:

I know that I can get a good print at an output resolution of 210 DPI

1 inch = 25.4mm

210 PPI /25.4 = 8.26 pixels per mm

8.26 pixels per mm = 4.13 line pairs per mm (two adjacent pixel columns, one black and one white, equal one 'line pair').

In other words, to get DPI translated into Line Pairs Per Millimetre, I merely take my preferred DPI for printing and divide it by 50.8  - and then enter that value into the app.

Eagle eyed readers will spot that though my 210DPI equates to 4.13 LPPMM, I have entered a value of 5 into the screenshot above. This is because the maximum print width that can be specified is 1000mm but I want to print to 1250mm: so to get the 'right' value I have to adjust the 4lppmm in the ratio of 1:1.25 and that gives a value of 5 (trust me, I have checked all this with the app's author and he agrees my arithmetic).

This is a very long-winded way to describe what value you should choose to enter into one parametric window - but it is necessary. Once you understand it, it won't need to be changed often at all.

BTW there is also the option to work with Blur Spot Diameter rather than Print Quality. This will please those who are happy with traditional COC values as a way of thinking of resolution. I am not going to cover this because it isn't how I prefer to work but I will say that the difference between Blur Spot Diameter and COC is irrelevant when it comes to knowing which value to enter: just dial in the COC that you are happy with as if they were the same thing. The true difference is that BSD will take diffraction into account without your needing to think about it any further. Again, I checked this with the app's author.

For those wanting to establish the BSD value to enter into the app, simply calculate a normal COC by taking your sensor's pixel pitch (width of sensor in mm divided by number of pixels on that side), doubling it and multiply by 1000. For example, a D800E's COC calculated this way is 10 microns, though a D800 might resolve slightly less well due to the AA filter. You get the idea.

Having set those parameters according to the above explanations, you can just go back to the main screen, set your shooting distance and fire away in the knowledge that whatever objects fall within the distance parameters of the two red lines will appear in acceptable focus at your chosen print size and resolution or COC settings.

Or can you? Does it work?

Mostly. There are some problems though.

  1. Firstly, very few camera systems let you set shooting distance with the requisite degree of accuracy. I used a High Precision Focussing Ring on my Rodenstock and even that isn't quite precise enough.
  2. Secondly, even if your lens were to allow you to set focus distances to the exact millimetre, the means of entering and reading data in the app are not refined enough for truly accurate use: the scales are simply too course for you to read a very accurate value from.
  3. The inaccuracies in 1) and 2) above might cancel each other out - or they might stack up. To get an accurate system, every component must be as accurate as possible individually.
  4. There are some aspects of optical systems that no calculator (except possibly made by the lens manufacturers using data they don't share) can help you with. Focus shift (residual spherical aberration) can blow all this out of the water, for example: the app has to assume no focus shift because it is a parameter that varies from lens to lens and behaves differently depending on how that lens is deployed. And the individual shape of a particular lens's field of focus, which will vary with aperture and distance, means that the placement and depth of the field of focus varies across the frame. Finally, individual lens designs (asymmetrical for example) can change the math a little, especially at macro distances. Again, nothing is going to help you with any of that other than 10,000 hours of practice and a sharp eye.

Nonetheless I have already found the app very useful in the field - and far more useful than anything else I have tried. I will give some examples below. I want to add this though: the thing that sets experienced photographers apart from the 'my cousin's got an SLR and can do it for half your price' crowd is the 10,000 hours of practice we have. That 10,000 hours means that we have a sixth sense for DOF and can often judge it very finely. So I think this app is of most use to the less experienced user, or to the expert who is using some new equipment and wants a short cut, or to those who want a sanity check or require greater precision than usual and need the best starting point for some bracketing. All that said, I have been using it blithely and getting really nice results.

All of this is a very useful update to a fundamentally traditional way of using DOF. For something more revolutionary, there is another app: OptimumCS-Pro has a very neat special trick: you tell it the distance to the nearest and furthest objects in your field of view that you want in acceptable focus and it will tell you if it can be done and, if it can, exactly where to focus and what aperture to set. Neat, eh?

It basically works in a way very similar to the above description but there are some wrinkles so let's take a more detailed look.

Here's the main screen:

photo copy 11 You will recognise the left hand input dial for lens focal length. But in the central panel, instead of you entering your shooting distance you enter your near and far object distances. The app then calculates the optimal focus distance, indicated by the red triangle marker. It also tells you, in the left-hand panels, which aperture to set and, with the blue triangle marker, what resolution you will get expressed in either line pairs per millimetre (shown) or optionally in Blur Spot Diameter (if that's how you roll). The keen eyed reader will see that though this suggested combination gives you a suggested F10 and indicates that within the zone indicated by the distance markers, a resolution of around 22 line pairs per millimetre will be achieved in a 250mm wide print. The very observant reader will also notice three little dots below that, and they refer to the resolutions that would be achieved were you to opt for apertures of one, two and three stops lower than suggested.

Now all of this is excellent in theory but there are some problems.

  • For those who think in DPI rather than LPPMM, some math is required to understand what the resolution numbers mean. Effectively, it is the reverse of the process I described for the first app, above, but with the additional need to convert for the print size, which in this app cannot be set parametrically and is fixed at 250mm. In the arithmetic above I showed that 210DPI is, for a 1250mm print, 4.13 lppmm. So for a print one fifth that width we translate that into a requirement of around 21 lppmm. But this is an irritating further step when working quickly and the inconsistency between the two apps is notable.
  • The slider blobs for entering distance are hard to set accurately (they move as you lift your finger) and they are so big that you sometimes can't get them close enough together to set the distances you want to set. Also, the scales are too impressionistic to either set or read a distance to the degree of accuracy required.

Before we look at how well this works let's take a look a the setting screen, where some other neat possibilities are revealed.

photo copy 4 Firstly, you will see that this is the place to set your sensor width. Next, aside from the buttons for Info (help) and useful external links, you will note two further options: one is to work with Blur Spot Diameter (I described this above in reference to the first app and the implementation and implications are similar here). But more excitingly for users of technical cameras, there is the option to work in Image Space rather than Object Space. If you don't know what that means, it won't matter - but if you do, then it might prove extremely useful to you.

Does it work?

I have played, used and tested - three different activities. In play mode, it's a hoot, I really like it despite the niggles. In real world use, it provides at worst a very useful close starting point to getting things exact and it will now be a part of my normal workflow. In fact on a challenging shoot today it just worked, every time. You can't say better than that. Here's an example, and clicking it will load a 50% resolution image in a new window. The aim was to get the front of the desk and everything on the desktop in focus and everything else out of focus. Not easy at close distance and with 80mp... but it worked perfectly: in fact, if you look very closely you will see that the right hand table leg (square carving at top) is sharper than the left - and that is because the one on the left was very slightly closer because I was very slightly off perfectly square to the subject. Now that really is precise!

CF002559

But in testing, the app doesn't quite prove itself as well,  performing more as a great guide than an indisputable oracle. I have my theories as to why this might be but first, let's look at two quick tests.

For these, I used an IQ180 on an Alpa STC technical camera with a Rodenstock 40mm HR lens fitted with a High Precision Focussing ('HPF') ring. In order to measure the exact distances involved I used a Leica Disto D510, an extremely accurate laser distance measure. I also used as much care as I could muster. Theoretically this combination should give about the tightest results currently possible. 

The first test involved two barcodes placed at 2 and 3 metres distance. For reference I put two further barcodes in frame at 50cm before and behind my main subjects. I then used the Disto to focus on the 2 metre subject and shot it accordingly at F10, then repeated the exercise for the 3 metre subject. These provide good benchmarks for 'the best the system can do'. Next, I used the app to determine the optimal focus distance and aperture (f10) then I shot at that distance as marked on the lens's HPF ring. Then I looked at the results at 50% zoom on a 100dpi monitor, which emulates my required DOF quite well, especially given that this combination of shooting parameters should have given at least 22 lppmm resolution for both near and far targets compared to my requirement of 21 lppmm. Then I also shot a frame at F8 which should have given me resolution just below my requirement. The results are below in the order stated and they clearly show that for the F10 shot, whilst the 3 metre target is close to acceptable when compared to its benchmark, the 2 metre target isn't. 

(All frames reproduced at exactly half of shot pixel dimensions. If you are unsure that your browser is rendering them inline at the right resolution, clicking any of them will load the image into a new window from which you can drag it into Lightroom or similar for reliable viewing)

CF002515

CF002514

CF002516

CF002517 I put these results in 'close but only a cigarillo' territory:  not quite accurate enough for critical use. But this is quite a close range test, and at close ranges all inaccuracies are magnified by the fact that DOF is much narrower at close distances. And clearly, I was trying to shoot with as little headroom as possible to see how the system performed at the margins.

So I ran a 'close to real world' test, with two Reikan targets at more or less random distances but significantly further away. The nearest target is at 4.73 metres and the furthest (hard right) is at 6.473 metres. Behind that, for blur reference, is a Rodenstock box at 7.1 metres.

Entering 4.73 and 6.473 gave a suggested focus distance of around 5.4 (it is impossible to read this distance exactly from the app's scale) and an aperture of F6.3 so I set the camera accordingly at that aperture, at which I should have achieved a massive overshoot of over 30 lppmm compared to my required 21 lppmm. So I also shot a frame a stop wider, which should have given me 26 lppmm - still a significant overshoot.

The results are as follows and in the following order:

  1. Disto focus near target at F6.3
  2. Disto focus far target at F6.3
  3. App suggested focus distance and aperture (f6.3)
  4. App suggested focus distance but shot at F4.5

CF002520

CF002521

CF002522

CF002523 It is clear to me that the results are acceptable but not perfect, despite the indicated resolution overkill. In fact the near target appears slightly closer to its benchmark frame in the F4.5 shot than it does in the F6.3 shot, which is initially counterintuitive. In any event, the results are better than in the first, closer range test, as we might expect.

So what's going on? Here are my theories (and remember, the app is performing as part of a complex system with many other parts which all contribute to the result):

  • I screwed up ( I don't think so but hey, it happens).
  • The Disto measurement should be taken from the diaphragm rather than the sensor plane, as I did (try finding the answer to that one in a hurry but I believe I did this correctly).
  • The math or the theory in the app is wrong (I seriously doubt this).
  • The input and readout scales on the app are not usable with sufficient accuracy (I am quite certain that is true and I suspect this to be the main, but not the only, cause of end-result inaccuracies).
  • The High Precision Focus Ring on the Rodenstock is either not perfectly placed (I installed it myself, standard practice, but I was very careful) or not perfectly accurate or is insufficiently finely calibrated (quite possible: the near and far benchmark frames are pretty good but they might well not be absolutely optimal).
  • The lens has some focus shift when stopped down (I tested it, it does, a little - and this is probably why the near target is a touch better in the F4.5 frame).

For whatever reason or combination of reasons, the imaging chain of which the app was a part in these tests wasn't quite able to 'nail it' in either of these tests. I think there are too many steps in the process which are vulnerable to minor imprecision for the end result to be perfect. But this might well not be the fault of the app at all and at the very worst, its suggestions work as a very close starting point. At greater distances and in my real world use it is proving extremely useful.

I have been in contact with George about these issues and he is actively considering making updates that address some of them. In particular, he is interested in making the transition to using DPI as the resolution measure, rather than LPPMM. And he is keen to make the input and readout accuracy more refined. He was also very responsive to some suggestions I had regarding the sorts of FAQs that might be addressed in the already excellent Info/Help section.

Conclusion

Neither of these apps is a quite a miracle - but they are both a true blessing. I haven't found anything nearly as useful elsewhere, and they are hardly expensive. Any serious photographer would find them useful but, for users of lenses with impressionistic distance scales, the results might be less predictable. For my purposes, both apps taken together constitute about the best value 'bang for your buck' of any photographic purchase I've made. Everyone should have them.

* this 210 DPI is a very convenient number: it makes good prints and it is also very close to both a traditional 100dpi monitor displaying an image at 50% zoom or a Retina screen showing an image at 100% zoom - in other words, it lets you preview your resolution and your DOF quite accurately before you print, with no effort)

 

 

 

This site is not for profit but I do support the charity Photovoice.  I wrote about it in depth a while back and that article is here. If you have found this article useful and are feeling generous, I would hugely appreciate a donation to the charity, even just a pound or a dollar: every little helps. You can donate here and the Virgin Giving site is secure and takes cards and PayPal. The Gods of Great Photography will smile on you if you donate. I promise.
 

 


Comments

Tim Ashley Photography
Hi Scott,
Disto is the way to go: accurate to 1mm and reliably so, by far the most accurate part of this imaging chain. The objects I was measuring for this test were all in the very central part of the frame so that was fine but actually the Rodenstock is a comparatively very flat filed lens in that using an accurate central focus distance will give you sharp edges on planar subjects - way better than most other lenses in that focal length range other than the Zeiss on the Sony RX-1...
scott kirkpatrick (HUJI)(non-registered)
Why not use a tape measure instead of the laser rangefinder for your distances. How accurate does the Leica discrete RF promist to be, anyway? Also, when you measure from the camera to an object in the corner of the field of view, don't you want to use the distance from the object plane to the extension of the image plane instead, assuming (hah!) a flat field of focus?
No comments posted.
Loading...
Subscribe
RSS
Archive
January February March April May June July (24) August (11) September (8) October (4) November (5) December (4)
January February March (2) April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December
January February March April May June July August September October November December