A full-frame digital M?
The Leica M8 has a 1.3x cropping factor, for the Epson R-D1 it is 1.53x.
What everyone really wants is a full-frame digital rangefinder body which doesn't have any cropping. Ideally like the Leica DMR, it would also be available as a clip-on digital back. After all if Canon can do an affordable full-frame DSLR in the EOS-5D, then why not the same kind of thing for a rangefinder camera? I mean, how hard can it be? What's holding this thing back?…
In February 2008 Steven Lee, (soon-to-be-ex) chief executive officer of Leica Camera AG, dropped hints that Leica were considering a full-frame option to be implemented at some unspecified future date. Which is nice, if a bit vague. Unfortunately at the end of the month Leica issued a clarification of Mr Lee's remarks.
Therefore — don't hold your breath. A full-frame 36 x 24mm digital rangefinder body may be years away, mainly because non-retrofocus wide-angle rangefinder lenses work poorly with the current generation of full-frame chips. Leica may be currently investigation a solution, but it is (at the time of writing) not yet a viable product.
Full-frame a long way off
The problem is that microlenses on current sensors only work properly when illuminated by nearly perpendicular light. Strike them from an acute angle (as you would with deeply recessed wide-angles), and you start to get drop-outs and erratic behaviour.
In May 2002 Sal DiMarco Jnr posted the following to confirm this:
Below, is the OFFICIAL Leica AG explanation as to why a digital M camera is not possible at this time. They are looking into the problem and hopefully will eventually find a solution.
Thank you for your e-mail regarding digital LEICA M or R solutions.
First of all, we agree with you that a digital M or R camera would make sense, because you (the customer) could use your current lenses for the new system. On the other hand we (the manufacturer) could sell the new product into an existing consumer base. Because of these reasons, this idea was subject to intensive research at Leica. As a result we must say, that a digital M camera is not possible with the current CCD or CMOS technology. Let me try to explain:
One great advantage of the M camera system is that there is no mirror between film/shutter and the rear lens element. This makes the lenses very compact and at the same time very fast. Due to that fact the Light rays hit the film in a flat angle. On a SLR lens, there is more space between rear lens element and the film/shutter to accommodate the mirror, the light rays hit the film in a rectangular or almost rectangular (telecentric) way.
If you use film, there is no difference in performance. If a digital sensor is employed instead of film, there is a difference:
1. To protect the sensor, a protection glass and filters (anti aliasing, IR) in front of the Sensor will make an additional and unwanted refraction, if the light rays will fall on the sensor in a non-rectangular way. This destroys the optical performance of the lens. (Still good performance in the center of the image, bad at the edges).
2. To increase the sensibility (speed) of the sensor, the pixels have to have microlenses to collect more light. Again, if the light rays will fall on the film in a non-rectangular way, less or no light will fall on the pixels at the corners of the image. This will cause a vignetting (dark corners).
These are major effects which will give the whole system a poor performance. For these and other reasons, we had to give up the idea at the actual time. As soon as new technology is available on the sensor side, Leica will continue the efforts for a digital M-Camera.
The above mentioned problems are less difficult with SLR lenses. The distance between film/shutter and rear lens element has to be larger because of the mirror, the light rays are almost telecentric and can therefore be used for digital. Possibilities for a digital R-Camera are under study at the moment.
(Indeed they were, and it bore fruit as the DMR 10 MPixel digital back in June 2003.)
Full-frame corner fringing
In Sept 2003 architectural photographer Henning Wulff spelled it out for the slow ones — until this problem is solved, there won't be a full-frame Leica M Digital anytime soon:
[…] DSLR sensors use microlenses over each photosite to compensate for the fact that all image producing rays do not come in parallel to each other, but there are severe limits on this form of compensation; just look at the colour fringing problems Canon 1Ds users encounter with various wide-angle lenses. Those lenses are strongly retro-focus too, with exit pupils much further from the film plane than even Leica's 50's […]
(You can read Mr Wulff's remarks in full at <LUG - v25/msg09646.html>.)
To support Mr Wulff's argument, see digital-guru Fred Miranda's EOS-1Ds review. See also Rob Galbraith's similar comments in Oct 2002, along with the Nikkor wide angle lens reviews by Bjørn Rørslett. Notice how all three remark that SLR wide-angle lens corner performance is considerably diminished when mounted onto digital cameras. Imagine how much worse it is for closer-to-the-sensor M lenses.
Furthermore, see the DxO Optics Pro photoshop plugins, Silkypix Developer Studio and Bibble Pro RAW developers — all of which are specifically designed to software correct (ahem) lens corner aberrations. Now why do you think they did that?…
Finally, see the December 2004 Luminous Landscape review of the Epson RD-1 by Sean Reid. Notice his conclusion that many W/A rangefinder lenses readily vignette with the RD-1, even though it isn't remotely full frame (1.53x cropping factor).
So this problem is real. It isn't a Leica or "Film Lubber" conspiracy. Until someone finds a way to handle acute-angle light from deeply recessed non-retrofocus lenses, there won't be a full-frame digital camera by anybody which can utilise the entire range of rangefinder optics :?(
But Leica solved the vignetting problem with the M8!
Not for full-frame they didn't. They used a 10.3 MPixel Kodak KAF-10500 sensor with a 1.33x cropping factor (18mm x 27mm) and especially thin (0.5mm) sensor IR filter. Even Leica agree that this is only a stop-gap solution.
A special feature of the 10500 chip is its "optimized microlens configuration", which helps deal with increasing amounts of CA and Vignetting as you approach the sensor corners. The technique appears to work well, but clearly it wasn't available as a full-frame solution when the Leica M8 shipped in late 2006.
Backyard Bozo full-frame digital M
While waiting patiently for a full-frame Digital M, check out Huw Finney's "M2aD" 2006 project, utilising a full-frame 14MP monochrome chip:
Corner fringing be damned! Maybe this is just the thing to put a firecracker down Leica's pants?… Now if only Mr Finney would finish the camera and get it to work…
Clip-on M digital back?
Some critics maintain that, corner issues aside, it would be impossible to mount a digital device onto the rear of a mechanical M body anyway. After all, how would the two ever communicate?
A similar problem was faced by Ixpress for their Hasselblad "V Series" digital backs. Their solution, the $US 9800 Ixpress V96C, (with its 36.9mm x 36.9mm capture area), works perfectly well with fully mechanical 500c/cm or 501cm bodies! How did they do it? Simple — they ran a flash sync cable from the body PC terminal to the digital back to let it know when the camera fired. Clunky, but it works!
( The newer Hasselblad CFV back doesn't even require a sync-chord. Unfortunately it utilises an internal Hasselblad prong which communicates with the film-back whenever the shutter is pressed. Of course there is no such thing in a Leica M. )
In theory there is nothing to stop Leica (or anyone else) from using the flash-sync trick for a Leica M. Of course the M7 or M6TTL won't fire a flash at speeds faster than X, but this could easily be fixed by a camera electronics upgrade.
Hopes were dashed in May 2006 however, when Christian Erhardt from Leica's USA office announced there were no plans for a clip-on digital back. If you want a digital M then you have to buy a dedicated digital M body — see the discussion at <Photo.net: #00GegZ>.
Marty Forscher to the rescue?
To work around the corner-fringing problem, why don't camera designers use a "Fused Fiber Optic array" to transfer the "non-telecentric" image light from the M focal-plane to the CCD surface, similar to the technique used by the NPC "Proback II" polaroid back (the current version of the original early 1980's Marty Forscher polaroid back)?
Quoting from the NPC www site:
[We] use a free-floating fiber optic lens to transfer the image from the camera film plane to the Polaroid film plane without any degradation of the image. Because the fiber optic bundle is in complete contact with the Polaroid film, the image is transferred with absolute sharpness and fidelity. There is no light loss and no color shift.
Why not do this for a digital M? Light from the lens would strike the front of the optical array at any angle, then travel down the "light-pipes" to emerge at the other end, perpendicular to the CCD surface. The "telecentric rays" problem would be solved in a stroke!
In Feb 2004 I ran this idea past an optics lab computer engineer. Unfortunately it turns out it wouldn't work because (a) you have to illuminate fused-optic arrays head-on to transmit light efficiently (the "telecentric" problem all over again) and (b) light travelling down an array suffers from leakage and cross-talk, resulting in a lot of image degradation by the time it comes out the other end. Although the resulting image may be good enough for quickie 36mm x 24mm previews on Polaroid film, it wouldn't cut it for hi-rez 10MP (and beyond) masters.
After reading the above remarks in March 2004, Jim Stolper of the NPC Photo Division sent me the following email explaining that NPC's fused-optic arrays don't actually suffer these problems:
[… Image degradation] would, indeed, happen except for the manufacturing process NPC uses in its fiber optic bundles.
Each fiber starts out as a glass rod with a specific refractive index. [… Glass rods are grouped together, pulled, broken off, grouped together and pulled again. …] The process is continued until the drawn fibers in groups of six attain a diameter of 3 microns or smaller if desired.
At this point the haxagonal combination of six, 3 micron rod-tube combinations is assembled mechanically with many other such combinations, placed in a rectangular compressible mold. (Some of our assemblies contain 150 million fibers). This assembled mold is put into an oven / press. This results in a fused block of many rod-tube combinations, all of which attain a multi-hexagonal, fused rectangular shape.
In essence, this resulting fused assembly has no cross talk from any one fiber to another since any light striking one surface of the assembly at any angle, up to and including an angle of 180 degrees, enters an individual rod-tube combination and cannot exit radially due to the different refractive indices of the rod and the tube. In fact, each ray of light that enters is successively reflected at an ever increasing angle and when it emerges from the other end of the bundle all rays are for all practical purposes, parallel.
Should you use Polaroid Positive / Negative film which has a resolution capability of 180 line pairs per millimeter, you would be able to enlarge the image by a factor of 20 before you would see the hexagonal pattern of the fibers. Essentially, what I am saying is that there is no cross talk and no "loss" of light and using the right kind of glass there is no color shift.
So who knows, maybe NPC's fused optical arrays may prove useful in avoiding digital corner artefacts and fringing!
What about "Silicon Film"?
The "Electronic Film System" (EFS-1) was developed in early 2001 by Silicon Film Technologies. It was supposedly a digital film insert which loaded into film cameras without modification. You dropped the "digital film" into your camera and shot as normal, with 24 images being stored in the self-contained electronics of the EFS-1 canister. When done, you removed the insert, plugged it into a dock and transfered the images to your computer.
Superficially this sounds great. It would also provide an instant solution to the digital M "problem", just drop in a digital insert and go. Neat! So what went wrong?… Unfortunately "Silicon Film" was never much more than a cool idea. They never moved beyond a 1.3 MPixel prototype and it is unlikely they ever will.
In May 2008 I received a long and detailed email from Jon Stern, a senior engineer at the former "Silicon Film Technologies" corporation ("SFT"). Edited extracts of his letter are presented below (with Mr Stern's permission). In it he quotes from my previous version of this page, and then carefully addresses each issue:
Silicon Film - an inside story by Jon Stern
I was a senior engineer at the company, the third employee and I even came up with the name for the company. Since it has been over 6 years since SFT collapsed I think that I can reveal some of the true story without violating any NDAs that I was previously bound by.
[AZN:] Silicon Film was never more than a cool idea. It was vapour-ware, merely a piece of hype from a tech-bubble start-up who wanted to attract venture capital.
Prior to the dot-com crash, we could have raised more VC money than we actually did, but instead we raised what we thought we needed and tried to avoid diluting everyone's stock. Raising VC money for it's own sake doesn't seem to be of much value to me. Sure you can pay a group of people's wages for a while but that's all you'll ever get out of it. At Silicon Film we were trying to build a real company with a real product. Each of the employees was working hard, both because we believed in what we were doing and because we all were working towards building the value of the company so that one day we could all make a bit of money off our stock options. Of course in the days before the market crash this was a common model.
When Silicon Film Technologies filed for Chapter 7 bankruptcy in Sept of 2001 EFS-1 was probably 3 months (my estimate, our management was saying 2 months) from reaching the market place. One of the engineers even managed to keep one of the pre-production models and continued to use it for several years until he purchased a Nikon DSLR.
The company was bought out of bankruptcy by Quest Manufacturing who was a vendor of shaped metal parts for the SFT. While the head of Quest was passionate about the product, it probably wasn't the best fit and he struggled to raise funding for the company. This was post-911 and VC money was very hard to come by. I was not part of this attempt at a relaunch, although I did not a little consulting for them at one point for which I was never paid.
As for resolution, the original product used a relatively small 1.3Mpixel sensor, as this was the only real off-the-shelf sensor available at the time. However, we had a custom 4Mpixel approx. 30mm x 20mm sensor manufactured that was successfully tested before the company closed. This was intended for the second generation of the product.
[AZN:] One of the insurmountable problems they faced (aside from image capture and storage), was that just about every camera has a different internal geometry for film positioning and transport.
Think about it. The way film is held in a Nikon F2 is different to the way it's done in an F5 or F100. Then there's the Canon FD and EOS variations. Then - of course - there are Leica M's, which have a radically different way of loading and positioning film. The permutations are endless.
This was a smaller obstacle than you might think. Silicon Film took accurate measurements from a wide range of camera bodies and found that just 6 versions would cover most of the popular high-end camera bodies. The adjustment of the sensor position was carried out as one of the final steps in the manufacturing process.
[AZN:] So even if it were ever made, it is unlikely it would work in Leica Ms without extensive modification. Then you would still have to address corner artefacts from our old buddy ? "non-telecentric rays"?
Leica M-series cameras were never a target camera for the original Silicon Film product. The EFS-1 cartridge powered up when it detected the sound of an SLR's mirror flipping. Seeing as a range-finder camera does not have a mirror, there was clearly no way of powering up the unit sufficiently early to allow the sensor to be integrating light by the time the shutter opened.
Silicon Film was, however, responsible for proving to Leica that a digital back was feasible (we had to carry out many tests and provide a lot of evidence to them) and specification discussions for a unit for the R8 and R9 (the R9 was in development at that time) were well advanced when Silicon Film folded (I personally made several trips to Solms). The plan was to follow up that product with a digital back for M-series cameras. I not sure I can recall how we were going to wake the unit up in advance of the shutter opening in the M-cameras. I think there are connections that either inside the back or on the base of the M7 body that are shorted when the shutter is pressed that we had planned to use (after 6 years I can't remember all the details, I'm afraid).
Of course in the absence of SFT, Leica had someone else develop the Leica Digital Modul R and ended up going for an integrated digital M-series camera instead of a back (the M-series certainly did present some "interesting challenges" for a back).
[Joseph Wisniewski:] You cannot make a full frame drop in insert, the sensor and frame have to protrude through the film gate. The Leica Modul-R (at 1.37x crop factor) is the largest sensor that will actually successfully go through a film gate. Since you can't have full frame, you'll have metering sensors outside the image sensor area, and for some cameras with a lot of AF sensors (EOS 1V and EOS 3 come immediately to mind) you'll even have AF sensors outside the image area. Even if the AF sensors are in the image area, they will be awkwardly placed for the 1.37x (minimum) crop sensor.
AF sensors outside the sensor area could be a problem although we found that under most shooting conditions this didn't seem to make a difference. Of course center-weighted and spot metering worked without any issue.
I largely agree with the statement that full frame is a problem, although I did file a patent for a novel kind of sensor package that would have been able to overcome this limitation. I'm not sure if would have been cost effective, but the concept made sense.
As you mention though, the high chief ray angles in wide, fast M-series lenses presents a problem for a full-frame sensor. It is possible to customize the sensor (microlens shift, etc.) to over-come this, but then you de-optimize (sorry, I know that's not a real word) the operation for more telecentric lenses.
[Joseph Wisniewski:] Sensors don't reflect like film, so you don't get TTL flash. Fuji and Pentax tried, and found you cannot do it without modification to the camera, increasing the sensitivity of the TTL sensors in the camera body.
This is a difficult issue, although we had patented the idea of coating the sensor glass to make the total reflectance similar to film. This would have come at the cost of a little bit of sensitivity (not a lot).
For the 1.3Mpixel sensor we had a black mask outside the sensor area, our idea was to change the reflectance of this slightly to get the average reflectance of the whole sensor package to match that of film. By chance we found that our default black mask worked perfectly for this and we were getting correctly exposed TTL flash shots (this was just about the only piece of good luck we had during the entire development).
[Joseph Wisniewski:] There's no communication between the digital and film parts of the camera. Want to change ISO? Set it on the camera's ISO control, then set it again on the "control module" part of the silicon film system...
Sure this was a limitation of the EFS, although this was no more of a limitation than that of film. We did however have an idea for an infra-red remote that would shine though the film speed window on the back and allow the user to make speed changes without removing the cartridge from the camera.
[Joseph Wisniewski:] A lot of cameras (especially newer Canons and Nikons) won't work unless the film sprockets are turned by a piece of film moving through the camera in response to the winding motors. They detect this condition as a film jam.
We heard this a lot, but it's based on an incomplete understanding. These cameras have a mechanical sensor that detects if a film can is present or not. If this sensor is pressed then the cameras will not work unless the sprockets are turned. But if the sensor is not pressed then the cameras will work as normal (this is necessary to allow in-store demos and for users to "dry shoot" without film). Our solution was to mold a depression in to the "film can" section of the cartridge so that the sensor would not be pressed. A single, extended depression was all that was required to get the unit to work with all the cameras we examined. Of course this depression slightly reduced the volume of the can and make fitting the electronics and batteries in to it even harder!
[Joseph Wisniewski:] So, Silicon Film allows you to take an expensive film SLR like a Nikon F5 or Canon 1V and turn it into a DLSR with badly placed AF points, problematic metering, no TTL flash, and a really awkward user interface.
EFS was only ever going to be a short term product. We knew that DSLRs would fall in price and eventually make the system obsolete. This meant that time to market was critical. Unfortunately it was the board of directors' unrealistically short development deadline that doomed the company. Short cuts were made (despite the protests of some of us engineers) that proved costly. As I've often said, if you try to develop a product in half the time, it'll end up taking you twice as long. First you'll try it the quick way. Then you'll spend just as long trying to fix it. Then you'll have to start again and do it properly!
EFS was a difficult product to develop, but we had overcome all the significant challenges by the time we closed our doors. Unfortunately, the management mistake mentioned above delayed the product and then after the dot-com bubble burst it was next to impossible to raise VC money. The majority share holder kept the company funded for around a year longer, but seeing as they had a few subsidiaries in similar positions and their own market was far from healthy at that time, this situation could not continue indefinitely.
What happened to SFT after the bankruptcy is another story and one that I played little part in, so I am not able to tell that part of the tale.
So forget about it. Silicon Film is DOA. Time to move on :?)
A note about possible broken links
This FAQ has over 900 external links. Over time it is inevitable some of them will break. If you are bothered by this, see this detailed topic elsewhere in the FAQ.