DIY digital medium format photography?

[rjpierrard]

<p>Hi,<br /><br /><br />I've been thinking about making a DIY digital medium format camera myself to cut the price tag slightly.<br /><br /><br />There's a group of engineers at the university I attend who contract outside projects, and I'll be speaking to them about the technicalities, but I just want to get an initial impression from the photography community first.<br /><br /><br />Here's what I have laid out so far:<br /><br /><br />If I use two full-frame sensors (eg. from a Canon 5D, the cheapest available) side by side, then the total sensor area would be 36X48mm. Given that there are two sensors, I'm thinking parallel processing would be the best route: save the image from each sensor to a different file, and assemble them in post-production (to save the camera's battery and processor). The obvious potential flaw is that the sensors would have to be exactly aligned next to each other, and completely flat to the image plane.</p>

<p>Question 1: Is the entire sensor area exposed, or are there peripherals that are not exposed (eg. maybe 23x35mm exposure area)?<br /><br /><br />Obviously full-frame lenses would not cover a 36x48mm sensor, so I'd be using some 645 MF lenses, most likely the Mamiya 35/3.5N and 80/2.8N. Crop factor for this size sensor is 1.16X, so the 35/3.5 would look like a 25mm on FX; the 80/2.8 would look like a 58mm on FX.<br /><br /><br />The next issue would be that the 5D sensor resolves at around 3000dpi to give 12MP on FX. At this resolving power, two sensors would simply be 24MP.<br>

I had a look at the DigMF sensors out there, and they resolve anywhere between 2800 and 4200dpi (with the outside being the 4890dpi of the IQ180).<br>

Full-frame sensors resolve between 3000-5200dpi (5200: Nikon D800).<br>

Crop-frame sensors resolve between 4200-5600dpi (5600: Canon 7D).<br>

From these numbers, I think I can safely assume a resolving power of between 4000-4800 is not unreasonable, and at this size, would render 45-62MP images, which could be printed 300dpi at 19x25 to 22x30.<br>

The difficulty there would lie in reprogramming the algorithm to something more akin to a crop-sensor's algorithm.</p>

<p>Question 2: is there any sort of interface to be able to play around with these type of details? I know of the firmware hacks that allow the camera to accept otherwise impossible settings, but I don't know if it can be extended to my project.</p>

<p><br />I've tried looking for retail image sensors, without real luck; however I think that simply buying the DSLR cameras would be a better option, as they already have the processors, body, etc.<br>

Question 3: is there enough room in any FX body for a 36x48mm sensor with dual processors? I haven't been able to find detailed diagrams, just the see-through images of camera bodies from dpreview.com.</p>

<p>I think I can assume that, with the lens adaptor, the image circle would diverge at approximately the flange/mount. Given that the flange-sensor distance is 46.5mm, the angle of divergence would have to be 58.1*, assuming the full 41.5*56mm = 69.7mm image circle. 645 medium format lenses have an angle of divergence of 72.3*, so having a minimum required angle of 58.1* for full coverage, that looks like it would work.<br>

For comparison, a normal FX lens covers 24x36mm with an end of lens-sensor distance of around 36mm (approximating using my Nikon 50/1.8D). The angle of divergence here is 42*.</p>

<p>Question 4: (out of curiosity) do different format lenses have different angles of divergence? I know that view cameras require the lens to be quite far away (from the large film sensor), and mirrorless 4/3s cameras have the lens very close to the (small) sensor.</p>

<p>Just to show the price range I'm looking at with this setup, two 5D bodies would come to around $2000, used (looking through the local classifieds); the two lenses would come to around $700, and the adaptor $80.<br>

<br />Materials cost is then $2800 or so.<br />Labour cost is as yet unknown, but I've heard the group at my university are quite reasonable, if you give them an interesting challenge.<br /><br /><br />Off-the shelf DigMF sensors alone are between $8000-$18000 for the dimensions I'm looking at, and offer between 22-50MP. So - significant savings, so I think I can deal with manual focusing.</p>

<p>If I wanted to extend this into the extreme, a 2X2 FX sensor system would give 90-120MP at 4000-4800dpi (or 47MP at 3000dpi), 1.08X crop on 6x7 lenses (if using 48x70mm rather than the full 48x72mm), and have a materials cost around $4600, but likely much more difficult in design and implementation.</p>

<p>Question 5: what sensor would you recommend for this venture? Using Canon's 5D II would give an easy 42MP without having to tinker with the algorithm, but at over double the price, plus tax. If buying the individual sensor and processor is an option, maybe I could simply insert that into a cheap body like the Nikon D90 - $500 used.<br>

<br /><br />What are your thoughts?</p>

[bob_sunley]

<p>The first problem is getting the sensors close enough together so you aren't missing a big strip out of the middle of your combined image. Just buy a Pentax 645D. :)</p>

 

[rjpierrard]

<p>Agreed, that is an issue, one I asked about and will look into. A Pentax 645D - though it looks nice - is out of my price range, along with all other digital MF cameras. That's why I'm considering DIY.</p>

[richardsperry]

A cheap LF camera, $200

 

One of those sliding back/lens mounts, $200

 

A nice used lens on it, $600

 

Stitching the frames together in PhotoShop, Free

 

Cannibalizing two 5Ds on something that most likely won't work, Priceless.

[rjpierrard]

<p>Continuing costs of 4X5 format photography: $4 per shot, having to be sent off-island, taking several weeks for processing. Not something that I want to be doing.</p>

<p>I have nothing against large format photography, it's just not optimal in my location.</p>

<p>Besides, I'm interested to see if this is even possible.</p>

[jslabovitz]

Your project sounds like a lot of fun, but I'm sorry to say that it's probably impractical.

 

Resolution and lenses are the least of your problems. Even the particulars of mounting a pair of

sensors so they are directly side-by-side may be a problem you can solve.

 

The bigger issue is going to be figuring out how to get two independent camera systems to work as

one unit. If you think about the guts of any DSLR, it's not just the sensor & the processor: the system

also includes all the controls (switches, knobs, etc.), and other smaller signal sensors (IR for focus,

lens coupling, etc.). This is designed as an "embedded system"; it's not like a PC where you can just

mix & match components, or remove parts and have it still work. I suppose you could leave the two

cameras as separate systems, and simply trigger both of them remotely, but I think you'll find it

tedious to keep the settings of two cameras in perfect sync.

 

I've been looking at Hasselblad-compatible digital backs on eBay. Old 16MP tethered backs (e.g., no

display) like the PhaseOne P20 sell for less than $3000. Those backs are well-known to offer very

good image quality, their pixels being generally larger (and hence can capture more light) than

smaller, high-density sensors. Throw in an old MacBook (with Firewire) and a 500cm or the like, and

you've got yourself a very decent system. It may not be super-portable, but neither would the solution

you're proposing.

 

--John

[bob_sunley]

<p>Robert,</p>

<p>Richard was talking about a sliding back that mounts a DSLR body. Not using 4x5 film.</p>

[John Crowe]

<p>In your budget you could get a used 5DII and one of the TS-E lenses used. I combine shifted images with a 5D II and 17 TS-E providing files as large as 42 MP with an equivalent view of a 12mm lens on 35mm format. There are also 24mm, 45mm and 90mm TS-E lenses and you can also use teleconverters on them with good results.</p>

<p>A Nikon D800 and PC-E lens will be more expensive but the combo would get you to 72 MP. Canon has more TS-E lenses so the 5D III which should look much like the D800 is another possibility.</p>

[rjpierrard]

<p>@Richard: sorry, I misread your post. Both you and John are talking about multi-image stitching, which I already do (except without tilt/shift).<br>

I researched the options for multi-image stitching, including standard panoramics, tilt/shift lens, shifting the sensor, and a panoramic head. For multi-image stitching, I prefer the panoramic head option, as there is no limit to the FOV one can do, nor lenses you can use.</p>

<p>What I'm interested in with this potential project is a high-resolution single-exposure. Obviously 4X5" photography is a very sound option, but impractical without a nearby developing lab.</p>

<p>My personal preference for film is 6X7cm due to its relative portability compared to 4X5"; and relatively large film size compared to 35mm/645.</p>

<p>With digital, however, there is the obvious issue with cost. Multi-image stitching is obviously easiest with digital, but the requirement for this project is using a single-exposure.</p>

<p>Does anyone have any suggestions or replies relating to my posted questions?</p>

[peterbcarter]

<p>Something you should look at - <a href="http://www.betterlight.com/">http://www.betterlight.com/</a></p>

 

[rjpierrard]

<p>I've had a look at scanning-back systems as well, thanks - the long exposure duration doesn't appeal to me: I'd prefer to have a large range of exposure times to be able to use, not a long minimum time. Also they carry a similar price range to digital medium format systems, so I would prefer those over a scanning-back system.</p>

<p>Basically what I'm looking to do is build a two-sensor array, the smallest possible version of the multi-sensor cameras in space telescopes like the LSST, and the Earth-based OmegaCAM.</p>

[ondebanks]

<blockquote>

<p>Basically what I'm looking to do is build a two-sensor array, the smallest possible version of the multi-sensor cameras in space telescopes like the LSST, and the Earth-based OmegaCAM.</p>

</blockquote>

<p>The LSST is ground-based. The Hubble WFPC2 and ACS cameras are in space.</p>

<p>Even with their "big-science" budgets, they have large gaps between the sensors. You just won't be able to avoid this with off-the-shelf components like existing DSLR sensors. The astronomy mosaic cameras use exposures at multiple telescope pointings to fill in the gaps, so the final image looks complete. The LSST camera specs for each of the 190 CCDs is "4-side buttable package, sub-mm gaps". A "sub-mm" gap is still a gap of up to 100 pixels wide! (and they're big 10-micron pixels too).</p>

<p>What you've come up with is a cool idea, but that doesn't necessarily mean it's a workable idea.</p>

[jan_h._voigt]

<p>If you don't want a big gap between the two half pictures, you could try a beam splitter which would cost you 1 stop but at least you might be able to get it work. If you use lenses with enough flange focal distance, it might even be possible to do it without disassembling the two cameras. Just use an IR shutter release to get them at least close to being synchronised and that's it.<br>

Have a nice weekend!</p>

[rjpierrard]

<p>Thanks Jan, I just heard about that today - had a talk with one of the profs in mechanical engineering.</p>

<p>It looks like the build will take a lot more than I thought with regards to the components and electrical, but I've been shown a few areas to look into to make this actually feasible!</p>

[nathan_bonsal]

<p>I hope you have a professional engineer who owes you 4 months of his time at 40 hours per week. At a minimum. Seriously, this is a fool's errand.<br>

Just buy a Pentax. Seriously. It is ALWAYS more expensive and difficult to reinvent a thing in order to build one than it is to initially invent it with the intention of selling tens of thousands.<br>

Unless you have the practical skills to actually DIY (you don't). your results will always be kludgy, will lack essential features, and be too fragile for practical use. Just aligning the two sensors such that they are in the same plane, while having them both soldered to circuits without ruining them is going to be a total pain.<br>

The physical work, while finicky, is nothing compared to the circuit layout and assembly. You will NEED a professional circuit board, the ability to work with BGAs, and some technique to not melt the AA filter and IR cut filters that were glued onto your sensors after they were installed. It's going to be messy, and you'll be risking the entire project and all funds involved at every step of the operation. You'll have no spares. If anything goes even slightly wrong, you will have nothing to show for your money or have an aborted or hopelessly inferior output for similar money to the professional rig.</p>

<p> </p>

[jan_h._voigt]

<p>That's why I suggested a beamsplitter. The only big problem will be aligning the two cameras in a perfect 90° angle and with the same distance to the beam splitter. No soldering, just mirroring one of the images in post processing and stitching them together. Of course, you will have to use manual mode in order to get a homogenious exposure and using a wireless shutter release will not synchronise the two cameras perfectly, but the cost will be less than 100 $ and there is no harm done to the cameras.</p>