Green stream.
Having a DSLR is fine, you have fast and accurate autofocus, clean high ISO and long exposure images in a reasonably priced system. But what if you really wanted a view camera? And what is it a view camera can do that you miss by only having a DSLR?
Lets have a look at what sets a view camera with a digital back apart from a common DSLR:
Will it be possible to get some or most of these traits in a view camera by "rebuilding" a Canon EOS 5D? Lets try!
This is a list of what is needed to convert your Canon DSLR into a view camera the McGyver way.
After buying the equipment above you will realize that you have spent quite a bit of money, so you are now close to satisfying the high price requirement of a view camera, even if we are far from the price of a real view camera with a digital back. And since the lenses needed only have manual focus , that requirement is also met. That was easy, but what about the rest of requirements?
This is built into the TS-E lenses, so this is easy to do. But it's a bit tricky to get the plane of focus right by only looking through the viewfinder of a DSLR. The lenses have the possibility of up to 8° of tilt or swing, but not both at the same time.
Focusing Rail & Slider from RRS.
Front rise, fall and shift is also built into the TS-E lenses with +/- 11 mm of movement.
To get the back rise, fall and shift movements we use the focusing rail & slider B150B + LMT + B2-FAB from RRS (see picture). If you want a left shift of the back, you shift the lens to the right, and use the focusing rail to shift the whole system the same amount to the left. That way the lens will stay in the same position and you have effectively done a left shift of the back.
There is only one way to get this from a DSLR: stitching.
By using the capabilities of back shift, rise or fall, you can take two or three separate exposures of the same image though the same lens. Because the TS-E lenses cover a larger image circle than needed (58.6 mm), you are in fact just using more of the available circle than one frame can capture by itself. This is why there is no distortion or registration problems when doing the stitching, and that makes it possible to do flat stitching without the need to do any cylindrical projection.
By doing this it is possible to get up to a 16.3 Mpix 3:1 ratio picture, or a 20.3 Mpix 1.6:1 ratio picture with a 20D/30D camera. This gives you the equivalent to having two sensors in the sizes 15mm x 44.5mm and 22.5mm x 37mm. Wit a FF 5D camera you will get a 20.5 Mpix 1:2.4 ratio picture, or a 24.4 Mpix 1.3:1 ratio picture.
Process the two or three raw files using the same white balance and setting in your raw converter, and then process them into PhotoShop. Create a new file with a large enough canvas, and then copy the files onto the new one. Each will be on its own Layer. Line them up to roughly overlap, and then enlarge the join area to about 300%. Reduce the transparency of one of the Layers to about 50%-75% and then move them so that they are in registration. Or use software like Realviz Stitcher that can automate this process.
Update: In PS CS3 the capability to stitch together images is much improved with the enhanced photomerge plugin. So today most stitching can be done perfectly withe photomerge. No more manual work.
The following table summarizes what is theoretically possible with the different Canon DSLR sensors available today when doing flat stitching with the TS-E lenses.
Sensor | Size (mm) | Width (px) | Height (px) | Mpix | Format |
---|---|---|---|---|---|
20D/30D | 15.0 x 22.5 | 6954 | 2344 | 16.3 | 1:3 |
5782 | 3520 | 20.3 | 1:1.6 | ||
1D MkII | 19.1 x 28.7 | 6201 | 2336 | 14.5 | 1:2.6 |
5027 | 3504 | 17.6 | 1:1.4 | ||
5D | 23.9 x 35.8 | 7042 | 2912 | 20.5 | 1:2.4 |
5593 | 4368 | 24.4 | 1:1.3 | ||
1Ds MkIII | 24.0 x 36.0 | 9063 | 3750 | 34.0 | 1:2.4 |
7188 | 4625 | 40.4 | 1:1.3 |
So we see that we have been able to increase the effective size of the sensor and increase the resolution. As a bonus we also have gotten a great panoramic format 1:2.4 with the 5D. And to do all this with stitching and back shifting we must use a tripod, so that requirement of a view camera is also satisfied.
When shooting this way you have to use the camera in manual mode. All exposures must be equal to make stitching possible (or at least easy), and when you are using shifts the cameras internal light metering will not see the whole picture to evaluate the "best" exposure. The solution is to use an external light meter like the Sekonic L-558 to make a good measurement.
With all the things mentioned above going on you are definitely slowing down, you are now at view camera speed.
In short: yes! The images below is the first two images I have shot this way, and the result is two perfect 20+ megapixel images.
Log in the river.
Small river.
It's possible to get much of the capabilities of a view camera this way and at much lower price, but as can be seen there is gap up to the resolutions of the current best digital backs like Phase One P45. To get that kind of resolution you have to do spherical stitching and software like PTgui or PhotoShop CS3 that now have become very good at this. If you have a Canon 1Ds MkIII you will also get very high resolution with just til/shift lenses as described in this essay (see table above).
Ronny A. Nilsen
Norway, June 2006
Here is some links to other articles that describes much of the same as I have done here:
Yucca
Factory Butte, first sun rays