Accurate slicing

Shooting in macro mode, techniques, tips & tricks
Post Reply
Borberg
Posts: 4
Joined: 19.02.2005 19:09

Accurate slicing

Post by Borberg »

Ever wonder how many slices you have to shoot to get 1 mm of object depth at a given magnification?
I use a Manfrotto macro rail which advances 1.25 mm per 360° turn of the crank.
This translates into the following number of slices and turns for 5 different magnifiction factors (using the Canon mpe-65) at f8:
M1: 1x.0.75, M2: 3x0.25, M3: 5x1/6 (2 hour intervals on a watch dial), M4: 7x1/8, M5: 9x 1/10.
In other words you need from 1 to 9 slices per mm ob object depth, depending on the magnification.
Don't forget: any speck of dust in your optical path will also be multiplied by the number of slices, leaving a trail like comet dust. Fortunately the dust map in Helicon Focus will "subtract" the dust.
Dan Kozub
Posts: 355
Joined: 24.03.2004 18:14

Post by Dan Kozub »

Peter,

Do you have kind of formula to calculate DOF? Or you get it by experementing? I do not believe in formula much. It seems to me that formula cannot take into account all the properties of the real optical system.

Have you tried to measure DOF? For instance to shot a textures surface with 45 degree slope? The width of the focused area in this case will be equal to the DOF. I also wonder if DOF in the center of image and near the border is the same...

Danylo
Borberg
Posts: 4
Joined: 19.02.2005 19:09

Post by Borberg »

For most of the macro lenses the manufacturer gives the DOF at certain magnification factors. I always shoot at f8 because quality deteriorates and exposure times become long (at 5x up to 15 seconds).

The trick is to calcualte how many slices you need to achieve the desired DOF and by how much the camera position needs to be adjusted between each slice.

The world is full of inaccuracies. The auto focusing mechanisms are not accurate enough (for the Canon MPE-65 there is none). It is difficult to accurately measure the distance from the focal plane (chip or film) to the object. And even precision macro rails cannot be adjusted within better than 0.05 mm. And finally, it is virtually impossible to accurately measure the object depth for the total DOF needed; this figure is usually a rough estimate.

The following DOF figures for the Canon MPE-65 given in the Canon manual are, therefore, nothing more than a starting point for the calculation.

1:1 = 1.120 mm, 2:1 = 0.420 mm, 3:1 = 0.249 mm, 4:1 = 0.175 mm, 5:1 = 0.134 mm

Helicon Focus prefers a certain overlap and I try to keep it at around 25%. More does not hurt, but less might.

The less the camera position is adjusted by one full turn of the control knob on the macro rail, the easier it is to position accurately. On the Manfrotto 3419 which I use one 360° turn corresponds to 1.25 mm.

This results in the following basic formula:
Y = 360 * 0.75 / 1.25 = 216
where 0.75 allows for 25% overlap, and 1.25mm is the advancement per 360° turn of the knob on the macrorail.

Calculate the number of degrees by which the screw needs to be turned:
Y * theorectic DOF at the given camera setting. This figure can be obtained either in the lens manual or obtained on the WEB from DOF calculators.

Using the above optical DOF specified by Canon the, the turns are 242°, 92°, 54°, 38° and 29° for the magnification factors 1 to 5.

A knob without a large dial scale cannot easily be turned with such accuracy. This is how I do it.
I printed character No. 146 (clock dial) from Windings2 at font size 550 in Word, cut it out and pasted it on an old CD. I fixed this CD to the control knob of the macro rail. In the case of the Manfrotto the center hole fits just over the knurled knob.

The increments that can be visually achhieved with this clock dial are: 242° = 8 hours, 92° = 3 hours or 1/4 turn, 54° = 2 hours, 38° = ?? and 29° = 1 hour. The only problem is with 4x magnification requiring 38°. You can shoot either at 45° intervals (8th) or 12th interval, depending on whether you wish more or less overlap. Alternatively you can create a second scale with 1/10th intervals and glue it on the flip side of the CD.

See picture of Manfrotto setup.
Attachments
Manfrotto railS_small.jpg
Manfrotto railS_small.jpg (67.35 KiB) Viewed 38785 times
Canon Macro setup3 P1010385_small.jpg
Canon Macro setup3 P1010385_small.jpg (115.88 KiB) Viewed 38785 times
Dan Kozub
Posts: 355
Joined: 24.03.2004 18:14

Post by Dan Kozub »

Peter,

I still do not believe in the estimations from the canon manual.
Have you tried to _measure_ the DOF?
Again, I am not sure DOF in the center and near the edges is the same.
Or these figures are just right?

Danylo
Museumtech
Posts: 3
Joined: 27.09.2006 06:59
Location: Australia
Contact:

Depth of Field

Post by Museumtech »

Peter,

Thanks. I was just reading the thread and you've solve a problem for me. Re:Helicon likes a 25% overlap. I downloaded Helcon a couple of hours ago and stiched some images together but there were blured sections that I couldn't avoid. I was using f 2.8. After downloading a copy of a depth of field table for my lens I was able to calculate that I needed a DoF of 8.0. The resultant image is brilliant.

Peter
TomMeeks
Posts: 45
Joined: 05.09.2006 00:10
Contact:

This has been useful

Post by TomMeeks »

Dan,

IS there a 'sweet spot' when it comes to a percentage of overlap of DOF that HeliconFocus likes? The 25% seems reasonable; but, do the developers have a particular preference.

This is where have a camera having the ability to do a live view into a larger screen monitor can be a great help. If we know the overlap that HeliconFocus prefers we could actually physically mark the monitor (with 'post-its' or a non-permanent marker or crayon) at the edges of the focal areas and ensure that the next image overlaps that area by a given percentage.

Since most DLRs for not have live preview, I'm hoping that my purchase of the Zigview S2 will give me that capability. We'll find out.

In any case, this thread has been useful. It's also been EXTREMELY useful to see the setups that people are using. I've been constructing a PVC shooting table with integrated lighting specifically for macro and ObjectVR work. If it works as I hope, I'll post some images of it.
Dan Kozub
Posts: 355
Joined: 24.03.2004 18:14

Post by Dan Kozub »

Tom,

There is no particular preferences in overlapping degree. Important is that all pixels are well focused at least on one image.

Generally there is no problem if the pixels is focused on two or more images. But in practice this can make the resulting image worse. Why? Well, the same focused detail can move a bit (1-2 pixels) while you are changing distance to object or focusing distance.

Another reason to avoid excessive overlapping: The program tries to align images. But nothing is perfect and the program may make a error for one pixel or so. Again, two focused areas but not perfectly align may result in blurry image.

That were mostly theoretical speculations. I often noticed that my assumptions do not work with real images. So my suggestion is to use the lowest overlapping degree as a starting point. The lowest means that you should not notice blurred zones on the resulting image.
TomMeeks
Posts: 45
Joined: 05.09.2006 00:10
Contact:

Almost ready for serious testing

Post by TomMeeks »

Thanks for the additional information. That's adds a lot for when I'm fully set up to begin developing a workflow. I've built a custom shooting table; but, I still need to take delivery on the Zigview and still need to pick up a Manfrotto Junior geared head.

I'd planned to buy the geared head for some time; but, now it seems imperative. Precision is the name of this game. :)

It's the nature of what I'd like to do that using a "by-the-numbers" approach will result in quicker learning and more consistent outcomes... at least in the serious testing phase.

Here is a shot of the shooting table I've built for this project. There are more images up there also. The table is built for stability and compactness. I shoot in a 12x12 artists studio with a kiln and other art tools that leave very little room for shooting.

Image

There are other, more detailed, images at:

http://s101.photobucket.com/albums/m47/ ... R%20Table/
Dan Kozub
Posts: 355
Joined: 24.03.2004 18:14

Post by Dan Kozub »

Let's wait and see the first results : )
Post Reply