Hyperfocal distance and DOF experiment

I tried an experiment using the HFD. Please give me some additional advice, because I'm not sure I'm doing this correctly.


The camera and lens: 5D Mk 2 and 24-105 L. IS was off. Camera was tripod mounted.


I disengaged the shutter button, using #3 in CFn IV. I used a remote cord to release the shutter.


Lens was at f/4, 24 mm, on aperture priority.


With the Mk 2, the Circle of confusion is 0.030 mm, so for f/4 at 24 mm, it's supposed to be 16 ft for the hyperfocal distance.


With a tape measure, I set the camera back 16 feet from a flat surface and that it was approx. 16 feet to the plane of the sensor. I then used the AF-ON button to focus.


I swung the camera around, then fired the shot with the remote switch. There was no wind (sorry about the exposure)


I don't see the distance in sharp focus at all (at least, not on my monitor).


It's approximately 200 feet to that chain link fence, behind that second tree.


The second picture is 100%, cropped.


I must be doing something wrong, since "infinity" is not sharp.


Your advice is appreciated. Thanks.


[img]/cfs-file.ashx/__key/CommunityServer.Components.UserFiles/00.00.00.25.34/_5F00_MG_5F00_8869res.jpg[/img]


[img]/cfs-file.ashx/__key/CommunityServer.Components.UserFiles/00.00.00.25.34/_5F00_MG_5F00_8869crop.jpg[/img]

did you try moving the camera forward 1 foot, then try 2 feet. There are alot of variables at play. Could be your lens is back focusing.

Are you referring to when I initially focus?


If it's after I swing the camera around, why would a back focus affect the "infinity" end to this degree?


I'll try another lens or two, to see what happens.

Assuming your measurement of distances and shooting methodology is correct, one must conclude that either (1) the aperture is not actually f/4, (2) the focal length is not truly 24mm, or (3) the circle of confusion is not 0.03mm.


The precise hyperfocal distance with those three parameters is 4.824m or about 15 feet 10 inches.


Here's what I think is probably happening. The 21MP sensor of the 5DmkII has a resolving ability that exceeds the traditional 35mm film and therefore when viewed at 100% crop you are no longer dealing with a CoC = 0.03mm. In fact, I believe that if you are going to use the hyperfocal technique to achieve sharp focus at infinity at 100% crop, you will need to use a much smaller CoC, resulting in a longer hyperfocal distance.


I would explain this in more detail but I think the Wikipedia article is a better reference: http://en.wikipedia.org/wiki/Circle_of_confusion

wickerprints, it is my understanding that "precise" distance measurement from the focus point is not critical, since some claim that you can use the focal plane of the sensor, or the end of the lens. In this case, it falls within the 15 feet 10 inches, since the lens length is beyond the 2 inch difference.


I don't understand the CoC not being 0.03, when the 5D Mk 2 is listed as such: http://www.dofmaster.com/digital_coc.html


How then can I ever use a graph or equation, if I can't presume the CoC is as listed? I'd never be able to find the hyperfocal distance if I can't start with what is known.


And, as for the f/stop and focal length, if I can't rely on those (especially with a lens that shows that I did, indeed, have the correct f/stop and focal length in the EXIF data), then combined with the "unknown" CoC, I might as well throw the graphs and technique in the trash bin.


There must be more to it than this.

First, I only included the precise measurement for the sake of illustration.


Second, you have to understand that a CoC is not a fixed property of the imaging system. It is a number that is chosen based on a semi-subjective analysis of the desired viewing sharpness. If some site just gives you a value, you can't just arbitrarily take it for granted because it does not explain the context in which the CoC was calculated. What is the desired viewing resolution? The image enlargement from the format size? The viewing distance? All of this is explained in the Wikipedia article.


The CoC is not like f-number or focal length or shutter speed or focusing distance. You don't look it up in a table somewhere because it's not fixed. You need to estimate it empirically based on viewing conditions. What you are doing--looking at 100% crops on a screen--is not the same as making an 8x10" print and looking at it from 25cm away, which is what the 0.030mm figure is roughly based upon.


Third, I did not say that the f-number or focal length were reported incorrect by the camera or the lens. I merely stated that if your methodology was correct, the only way to explain the discrepancy was if one of those three parameters were not as claimed. From this I deduced the likely parameter that was not correct is the CoC--implying that it is highly unlikely that the other two are incorrect to the magnitude that would affect the calculation of the hyperfocal distance.


Finally, let me give you an example of how I would go about empirically determining a CoC. First, I do a calculation based on my viewing conditions to give me a ballpark estimate:


If I view the image at 100% crop, then what is the enlargement factor? A 5DmkII has 5616 pixels per 36mm sensor width. This is equivalent to (5616 px)/(36 mm) * (25.4 mm)/(1 inch) = 3962.4 ppi. On my computer screen, I have 1440 pixels in 13 inches of horizontal screen width, for a resolution of 110.8 ppi. Therefore, my enlargement factor is 3962.4 / 110.8 = 35.8 if I view the image at 100% crop, as you have done.


Now assume I view my screen from a distance of 50cm.


Next, what is the maximum resolution of my screen in line pairs per mm? 110.8 ppi = 4.36 p/mm, which is a maximum of about 2.2 lp/mm. I can't really distinguish more than that, even if I got super close to my screen. I would say my eyes could resolve the full 2.2 lp/mm at a viewing distance of 35cm.


Okay now that we've got all these numbers, we calculate the CoC = 50 / 35 / 2.2 / 35.8 = 0.018mm. This is significantly smaller than the 0.03mm figure you used, and would affect the hyperfocal calculation accordingly:


Hyperfocal distance @ f/4 @ 24mm @ 0.018mm CoC = 7.963m = 26.125 feet.


Is that going to work? The problem is, I don't know. You need to verify it empirically by then going out on a clear day and taking photographs and determining at what focusing distance you are getting acceptably sharp images at infinity. Once you do this, you will be able to establish a CoC figure that should then be independent of focal length or aperture, but is suited for your viewing conditions. You will need to do other calculations for CoC if you decide to print those images. Do you see how this works now?

wickerprints, yes, I do. Thanks.

As others mentioned, 0.03 is not small enough for a 100% crop of the 5D2 (which is like looking at a 6 foot wide print up close).


But more importantly, why are you bothering with a tape measure and charts? At their very best they will still be less precise than 10X liveview, and are much slower besides.

Daniel, I'll switch to that for the Mk 2, but I also still have my original 5D and the 40D, so I'd still need a tape measure for those if I'm out in the field.


The experiment was to see how well the HFD thing worked.

Quote:

---

Originally Posted by Alan

Daniel, I'll switch to that for the Mk 2, but I also still have my original 5D and the 40D, so I'd still need a tape measure for those if I'm out in the field.


The experiment was to see how well the HFD thing worked.
<div style="clear: both;"]</div>

---

Of course; that makes perfect sense.

Try manual focusing on the hyperfocal distance using your distance scale. AF doesn't always hit the right spot.

when you are hyperfocal distance focusing you are placing the focal point on a subject matter. What this means is that obviously there is DOF in front and behind your focal plane. However, the calculation alone will not gurantee perfections in the lens opticals and what you are doing is placing too much demand on the lens at that aperture.


I would have used live view if i were you, and then focused on the tree, this would help with focus to begin with.


I woul also stop down the lens to F/8 or F/11.





What you also need to realize is that, at the far ends of the DOF buffer, that resolution is lost you aren't resolving as much. However, when viewing the print "normally" the background trees appear as sharp and within the DOF.





I believe if you stopped down to F/8 or F/11 you would see a noticable increase in the resolved trees.





FOcus on car @ F/2.8


http://farm4.static.flickr.com/3594/...b391d847_b.jpg





aperture down to 4.0


http://farm4.static.flickr.com/3357/...f1f0e397_b.jpg





Aperture to 5.6


http://farm4.static.flickr.com/3258/...c4bb402d_b.jpg





Aperture to 8


http://farm4.static.flickr.com/3624/...72987c63_b.jpg





While my focus did not change at all, I left all of that stationary, I did stop down my apertures and you will see how the items in the rear are better resolved. However, you just can't expect that great of a performance at F/4.

Also the following scenes are with the 17-40 L on an XSi


THe slices are full size and taken from images that look like this:





http://farm4.static.flickr.com/3217/...2bebc03062.jpg





F/8http://farm4.static.flickr.com/3101/...bd7cfa19_o.jpg





F/11


http://farm4.static.flickr.com/3118/...b4b575e6_o.jpg


F/16


http://farm4.static.flickr.com/3130/...b688ae63_o.jpg





I believe this shows the other side of the Hyperfocal problem, where you resolve better at F/8 and the narrower you go, the more your image suffers from diffraction at 100% views.

The impression I got from Alan's post was not how to maximize DOF through hyperfocal technique, but rather, how to determine the hyperfocal distance for a given f-number and focal length. Therefore, the calculation demonstrates three principles: first, that the determination of the hyperfocal distance is sensitive to the CoC which in turn is dependent on the viewing conditions of the image; second, that the commonly stated CoC of 0.03mm for 35mm format is not suitable for 100% crop on a 5DmkII, as this corresponds to a particularly high enlargement ratio due to high pixel density; and third, that it is ideal to not have to determine these things empirically at the time of capture, and that simply stopping down is not the relevant approach for understanding this subject.

Thats fine.


If he used a relevant COC such as .019 he would have found:


<span style="word-spacing: 0px; font: 16px 'times new roman'; text-transform: none; color: #000000; text-indent: 0px; white-space: normal; letter-spacing: normal; border-collapse: separate; orphans: 2; widows: 2; -webkit-border-horizontal-spacing: 0px; -webkit-border-vertical-spacing: 0px; -webkit-text-decorations-in-effect: none; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class="Apple-style-span"]<span style="font-size: 11px; font-family: Arial;" class="Apple-style-span"]



<table summary="results of depth of field calculation" width="100%" style="padding: 4px;" class="resultsTable"]
<tbody>
<tr>
<td width="55%" align="left" valign="top" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Subjectdistance</td>
<td name="resultsDistance" align="left" valign="top" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="resultsDistance" class="textClassResults"]16ft</td>
</tr>
<tr>
<td colspan="2" align="top" style="padding: 1px;"]
</td>
</tr>
<tr>
<td colspan="2" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Depthoffield</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; text-indent: 20px; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Nearlimit</td>
<td name="dofNear" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="dofNear" class="textClassResults"]9.75ft</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; text-indent: 20px; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Farlimit</td>
<td name="dofFar" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="dofFar" class="textClassResults"]44.5ft</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; text-indent: 20px; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Total</td>
<td name="dofTotal" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="dofTotal" class="textClassResults"]34.7ft</td>
</tr>
<tr>
<td colspan="2" style="padding: 1px;"]</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; text-indent: 20px; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Infrontofsubject</td>
<td name="dofFront" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="dofFront" class="textClassResults"]6.2ft (18%)</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; text-indent: 20px; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Behindsubject</td>
<td name="dofRear" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="dofRear" class="textClassResults"]28.5ft (82%)</td>
</tr>
<tr>
<td colspan="2" style="padding: 1px;"]</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Hyperfocaldistance</td>
<td name="hyperFocal" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="hyperFocal" class="textClassResults"]24.9ft</td>
</tr>
<tr>
<td style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" class="textClassResults"]Circleofconfusion</td>
<td name="cocused" align="left" style="font-weight: bold; font-size: 12px; margin: 0px 5px 5px; color: black; font-family: Verdana, Arial, sans-serif; padding: 1px;" id="cocused" class="textClassResults"]0.019 mm</td>
</tr>
</tbody>
</table>






Therefore, I advise consider stopping down to be with in the performance specifications he desired for his IQ.

I don't necessarily want to resurrect this discussion, but I found this article informative.


http://www.trenholm.org/hmmerk/DOFR.html


He says that if you focus at the hyperfocal distance, the background will be disappointingly blurry.


With my experiment, this is what I found, as well.


He doesn't believe in the "old story" of focusing at the hyperfocal distance.


I think that I might do better by using Photoshop, and focus stacking two images, than taking chances with hyperfocal distance experiments.

Quote:

---

Originally Posted by Alan

He says that if you focus at the hyperfocal distance, the background will be disappointingly blurry.

---

He's doing it wrong. :) The "old standards" for hyperfocal distance (like the ones printed on lens barrels) are based on small print sizes and low resolutions. These days we view images at 100% on the monitor up close, which simulates a print size of 6 feet wide. The hyperfocal distance varies by the acceptably sharp CoC, which is smaller for 100% crop viewing than it is for a 8x10 print, for example.

Well, if that's the case, I'll try it again.


Maybe it's worth focusing (no pun intended) on f/8 (which is mostly where I'd use the camera for landscapes, anyway), then experiment around the theoretical hyperfocal distance, then just print some images to see how it goes.