Graflex 4×5 Digital Back DIY

A Graflex Super Graphic 4×5 camera with lenses popped up on Craigslist for $125.  The seller, a photographer Timothy Wilson, explained that he no longer had time for the camera.  He knew the price was a bargain (Thank you, Tim!).  After buying it, I began to research and discovered that other tinkerers are breathing new life into the 4×5 format.  New55Film is manufacturing instant film, which can be used like the now-defunct Polaroid 55. They also sell film and single-bath (monobath) development kits.  Timothy Gilbert is developing a painless 4×5 development tank through Kickstarter, the SP-445.  I’m happy to learn that if my experiments fail, I have options in film.

My goal with the Super Graphic was/is building a low-cost DIY digital back–a high speed one preferably, for portraits!  If a 4×5 digital sensor exists, it must cost a fortune.  The least expensive medium format camera (2 1/4 inches) runs around $7,000 without lens, the Pentax 645Z.  Professional medium format photographers typically use $50,000+ worth of gear.

My goal was some sort of raster-scan digital back using a mirrorless camera, like the Sony A7.  Fotodiox offers a few adapters that allow one to use a Nikon, Canon, Pentax or Sony A-mount on their 4×5 camera, using one of these adapters.  Here’s a more sophisticated Vizelex Rhinocam from Fotodiox for medium format.  I have many Fotodiox Nikon/Canon/M42-mount to e-mount adapters for my Sony cameras and they are great value for the money.  However, there are some limitations to their Rhinocam-type solutions which I’ve always had an itch to solve!

For starters, I had to determine how difficult it was to get anything workable.  This was my first model

Camera mount adapter has been removed
Camera mount adapter has been removed

From my first attempts I immediately noticed the benefits of higher resolution, less chromatic distortions from bayer sensor images.

Cut from 300MB 4x5 stitched image
Cut from 300MB 4×5 stitched image
Cut from 24MB A6000 Image
Cut from 24MB A6000 Image

The general idea behind the approach is to create a virtual 4×5 sensor (102mm by 127mm) by tiling a full-frame digital sensor (36mm by 24mm).  In a perfect word, we could place our digital camera sensor 16 times to capture almost a full 4×5 image:


Unfortunately, we can’t be certain that we can align each exposure perfectly, where there isn’t missing space between tiles.  To be confident that we’ll get the whole 4×5 image, we’ll need software to figure out where our images overlap so it can “stitch” them together.  The more overlap we have, the easier it is for the software to create the composite image.  Generally 50% overlap is safest.  Therefore, a real world capture pattern is this:


The corner and center tiles are transparent, so you can get a better idea of what’s going on.  Essentially, there is some overlap between all images.  We need 30 images, double our theoretical 15 full-frame shots (102 x 127 = 12,954mm, 36 x 24 = 864mm; 12,954/864 = 15).

Anyway, I attached the A7 to the black board, held it against the white board that was attached to the Graflex (w/Nikkor W 180 f/5.6), fitted with a 4×5 picture frame on the other end.  The grid helped me position the camera in a way to expose a grid of 5×6 overlapped images.

Original 20,413 x 15,086 (307MB) JPEG
Original 20,413 x 15,086 (307MB) JPEG

You can download the large JPEG here  You can also download the original 30 images.

I was very excited!   The free Microsoft Image Composition Editor (ICE) stitched the images together well.

Then the honeymoon ended.  The biggest obstacle to stitching-based large-format photography is subject and/or camera movement.  Here’s a shot where a combination of the camera moving (disconnected girders in background) and my (sliced arm), result in a “problem” photograph.


(Nontheless, this was the first attempt my friend Mark Wylie made using my crude device on a gravel surface!)

I have yet to figure out the best approach with stitching software; each has its benefits and drawbacks.  For now, I’ll just list the options, with what I know so far.  In freeware there is Microsoft ICE (link above).  ICE’s “structured panorama” visual tool is the best approach for this type of stitching in that you can see what it’s doing.


It’s Windows-only

For a cross-platform solutions there is Hugin.   Other commercial solutions are PTGui, Easypano.  Adobe Photoshop will also photo “merge” images.   The problem with the non Microsoft ICE solutions is they don’t have a strong “structured panorama” approach to image stitching; indeed, ICE is almost perfect in that regard.  Most stitching software assumes you’re a panorama head and uses the focal length of the lens to pre-calculate stitching parameters.  I’m hoping that I can use a tool like ImageJ to create scripts where I can say, ‘these are the exact placement of these images so please stitch from there’.  It’s a lot to do.  Or, I may figure out a fool-proof procedure in ICE.

In the meantime, I’m working in improving the digital back so that I can take 30 shots as quickly as possible.  The faster that can be done, the more one could use this method for portraits.

Here is a link to a Flickr album I have set up for this project.  Thanks to Mark Wylie for shooting the video above!

I’m working on a 3rd version of digital back.  Part II of this coming soon.  I hope!



WIFEL for Panoramas

Here’s the acronym for my pre-panorama checklist: WIFEL.

W: White balance, set MANUAL, like Daylight if you’re outdoors
I: ISO, set MANUAL, to the lowest you can use, like 100 outdoors.
F: Focus, set MANUAL and focus for the important depth
E: Exposure, set MANUAL (scan the scene and set to important stuff)
L: Level, is the camera/tripod level?

In other words, take the camera OFF AUTO!

If you have a better checklist, or can improve this, please let me know.

Here’s my latest panorama head

I just shot this panorama using it and a Sigma DP2s.  Naturally, I didn’t follow my own WIFEL and left the camera on ISO 800 from the day before.

Click on my name, on the pano, to view others I have done recently.

I also set the camera to a 2-second shutter timer. That way, when the shutter fires the camera is still. This is also important because it gives the camera time to buffer/save the image.

I’m hoping to do some high quality multi-image stiches using the pano head. Here’s a quick test I did recently, freehand. You need to see it large to appreciate the DOF effects.

Sigma SD14 as Panorama Camera

I built a quick manual panorama maker for the Sigma out of particle board and lazy susans.  Here’s the setup:

I used the 50-200mm lens, and set it to 50mm.  Although the camera takes nice pictures, as expected, the buffer is too slow.  After taking about 7 images I’d have to wait around half a minute while the buffer emptied, and then take 7 more.  I took 3-row panos, 36 columns, or 108 images.

Microsoft ICE was updated recently.  Unfortunately, it wasn’t a major upgrade.  The 64-bit version is fast, but the structured pano come in at lower resolutions than the panoramas created automatically.  I’m sure there’s a good reason for this, but I don’t have time to read the newsgroups.

If you’ve never used ICE before don’t get me wrong.  The software is a lot of fun and HIGHLY recommended.  It’s free.

Here’s a panorama of the Dartmouth Alumni Gym (there are a couple of dead spots because I got distracted talking to a friend).

And here is one of Berrill Farms, which is a quarter-mile up from my house. The clouds were moving, so it created light problems.

Just in case you think I’m knocking the Sigma camera, look at this shot. When it comes to color the Foveon sensors really do shine.

That’s it from here.