An R10 sensor prediction

[adan]

Advertisement (gone after registration)

This just occured to me, and I will stand by it until facts prove me wrong.

 

I've concluded the digital Leica R10 will use a "double-M8" sensor. 27mm x 36mm (give or take .3mm), standard Kodak 6.8-micron pixel-pitch, 22.4 total megapixels, 21 effective megapixels, 3:4 "ideal-format" picture shape.

 

[edit: the diagonal of a 27 x 36 sensor is only 0.84mm larger than 24 x 36, (44.1mm vs. 43.26mm) so it should be safely within the image circle of existing R lenses]

 

(Why 21 effective Mpixels from doubling the M8's 10.3, instead of 20.6? Because sensors have a border of pixels that don't contribute to the final image directly, but by putting two sensors together (or, more accurately, never cutting them apart in the first place) the "borders" where the two sensors meet become effective pixels and part of the image.)

 

27 x 36 counts as the rumored "oversized" sensor - even though it is only larger than 24 x 36 in one dimension. It does make shutter design and construction a bit easier than increasing the size in both dimensions, though.

 

For folks who like the "ideal format", it offers an even bigger advantage over Canon/Nikon 24 x 36 sensors in that they must be cropped to reach that format - to about 18-19 Mpixels. So the R10 will have a very viable market position in between those cameras and the 36 x 48mm MF digital backs (especially since Leica will continue the "no-AA-filter" policy).

 

I also predict the camera will have a cropped option to 24 x 36 for those of us who like the "Barnack format" - with viewfinder masking to match, as in the Nikon D3 or some similar technology (LCD darkening plate or crop lines overlaid with the groundglass). About an 18Mpixel image.

 

I could really go out on a limb and hypothesize a "square-format" in-camera cropping option as well - an effective 27 x 27mm sensor - which comes out at 16 Mpixels, a direct match to Hasselblad's CFV back for the 500/501/503C/CM cameras.

 

(With "edge-recovery" software one could get the full 27 x 36 image out of the RAW data, regardless of cropping applied in-camera, of course).

 

I suspect it will have a standard Bayer RGB color pattern to begin with, but that Kodak will use several R10 bodies as testbeds for their newly described high-ISO colr-matrix patterns that mix unfiltered "white" pixels in with the filtered ones to catch more light, and that this pattern will eventually show up in subsequent Leica cameras if it works out - and may even be retrofittable, since the pixel count and data volume will still be the same - it will just need new firmware to interpret the different color pattern. This could offer up to 3x the ISO for a given amount of noise, depending on which pattern is finally chosen and how it is implemented.

 

As an add-on - I suspect that the M9, if Leica solves the continuing angle-of-incidence problem for a sensor bigger than the M8's (which I suspect they will eventually) will use tis same sensor, but permanently cropped to 18 Mpixels because it is much trickier to deal with a format change with RF framelines than it is with a groundglass.

 

(OTOH Leica is tricky, and could surprise me with the LCD frames other have dreamed about - which would make changing the format shape as well as the field of view a snap).

 

The M9 will be the prime candidate for the new Kodak high-ISO pattern, which sacrifices a bit of color precision for better low-light performance, given the M's reputation and effectiveness as a low-light and monochrome camera.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Angora]

Hey, that makes sense.

Re-using a mastered sensor (maybe the M8 one will be enhanced a bit, in order to solve the IR thingy) permits to spare a considerable amount of time.

[Bob Ross]

Hi Andy,

I offered the same possiblity to the ......"FF?...why not larger" hint in a thread a few months ago. I based my idea on Kodak's propensity to make larger sensors in multiples of smaller ones. In this case, the base unit is the 4/3rds sensor, the M8 has a 2X and the "R?" might have a 4X. It is an integrated tile approach and may have to do with the stepping machines used. Your ideas fill out the concept nicely.

Bob

[ho_co]

Brilliant deductive reasoning, guys! Makes perfect sense.

 

As a matter of fact, the only thing I see wrong with the prediction is just that it does make perfect sense.

 

As Andy said, Leica is tricky, so the fact that this is a predicted possibility may reduce its probability. (Correct me on that, Bob...)

 

Andy, you speak of a couple ways of masking the frame when the camera is adjusted to Barnack Format (heck, if Nikon calls it FX, why shouldn't Leica call it BF?--good call, Andy!).

 

One way that you didn't mention would be a mechanical mask that would require removal of the focusing screen to install. Wouldn't that be more like the traditional, mechanically-based Leica of the past? The screen mask would be, say, a $300 accessory and would come with a second one that could be installed if wanted on the LCD. The mask would of course set the camera to BF and adjust the metering pattern automatically. Like the first Leica interchangeable screens (R4), it would take about three minutes to make the change until one got the hang of it, when the time would drop to about 20 sec.

 

Oh, and then there's the second finder insert for the 16:9 format. Just listen to that mirror flop at 30 fps for HDTV recording!

 

Seriously, as I said above, your prediction is brilliant!

 

--HC

[ptomsu]

While this seems all very reasonable and I myself am using the 4/3 format in the FT system with the E-3, I still cannot see what advantage this 4/3 format makes today.

 

Most screens on Laptops and Computers have a 3:2 - which makes already small grey or black stripes to the left and right if you diplay 4/3 photos on those - but if you go to Plasma or LCD Video panels to show your photos (which will be the dominant display system of the next decade) you even have 16:9 - and you will then have even larger stripes left and right.

 

So why step back to 4:3 ????

 

For me it rather would make sense to go to a 16:9 format, which of course would mean a very different shutter, metering and lens design.

[luigi bertolotti]

Agree that the hypotesis is brilliant and has a tech sense (apart the little error about the 27x36 diagonal... ); me too wrote time ago that announcing a standard FF DSLR would put Leica into a difficult position... too direct a comparision with Ni/Ca (while M8 stands by itself, a unique product), and so very difficult to play the price game. I had envisioned a SQUARE sensor in the 31-32 mm size, but also the 3x4 option can be viable; I continue to prefer my idea... to explain, in marketing terms, WHY a 3x4 format has "something more" than the standard 2x3, I think is difficult... on the contrary, the square neg has been for decades an appreciated option by pros and amateurs.

[adan]

Advertisement (gone after registration)

Luigi: Yah - should have been 45mm diagonal - I calculated for 24 x 37 by mistake the first time (Doh!). Still likely safe from vignetting.

 

I "like" 1:1 and 16:9 formats, too. But obviously the marketplace is not crying out for these formats as prime sensor shapes, seeing as there is one camera or back using each (D-Lux for 16:9 and Hassy CFV for 1:1) compared to countless 3:4 or 2:3 cameras. In five years or so when HDTV becomes far more universal we may see more still cameras moving to that format.

 

I'd love to see a 1:1 digital "Rollei" or "Mamiya 6" or whatever in the sub-$5000 range with at least a "50mm wideangle equivalent" lens - It doesn't even have to be "medium-format" in size - just a squared-up 25mm x 25mm APS sensor would do me. But I suspect I will have to live with cropping in the RAW converter for quite a while - unless the R10 comes through.

 

In making a prediction I am saying what I think Leica WILL do, not what I WANT them to do - a distinction.

 

Bob: yes, I knew I had seen a mention of a quad-4/3rds format among the various R10 discussions.

[Iron Flatline]

Very cool. I have no idea if you're right or wrong, but it sounds great and I'd vote for it!

[marknorton]

That's a neat idea, we'll be able to tell R10 pictures by the bar down the middle where there's a gap betweeen the the two sensors...

 

What about the sensor readout? If the sensors were turned through 90 degrees, there'd be a readout top and bottom, left and right, which raises the spectre (specter?) or a vertical AND horizontal centre-fold line when things go wrong.

[andybarton]

The gap in the middle is what's been taxing me too...

[earleygallery]

(Why 21 effective Mpixels from doubling the M8's 10.3, instead of 20.6? Because sensors have a border of pixels that don't contribute to the final image directly, but by putting two sensors together (or, more accurately, never cutting them apart in the first place) the "borders" where the two sensors meet become effective pixels and part of the image.)

 

If I read this correctly, there wouldn't be a gap, it would just be a sensor area the size of 2x M8 sensors? (not two separate sensors stuck together).

[Riley]

i dont think its as simple as all that. Where would the wiring go for the join side of the sensor? Surely it has to be a single full sensor fab

[ptomsu]

If I read this correctly, there wouldn't be a gap, it would just be a sensor area the size of 2x M8 sensors? (not two separate sensors stuck together).

 

I was always assuming that there would be one "single" sensor!

[adan]

It should be remembered that sensors are not produced each on it's own little piece of silicon.

 

They are produced ganged together on a big disk of silicon (usually 8" in diameter these days) Sort of like those "picture packages" where one gets a 5x7, a 3-1/2 x 5 and two 2-1/2x3-1/2 prints on one piece of 8x10 photo paper. In both cases the silicon or paper is then cut apart to separate the units.

 

I'm not envisioning two literal M8 sensors glued together - I'm envisioning the basic pixel architecture of two M8 sensors combined side by side into one integrated circuit. Modifying the direction and number of readouts are trivialities.

[Bob Ross]

i dont think its as simple as all that. Where would the wiring go for the join side of the sensor? Surely it has to be a single full sensor fab

I think we have to step back into the dark room to figure this out:eek: Sensor cuircuts are made like we used to do enlargements...or in this case reductions...by projecting the circuit paterns on a sensitized layer on the silicon, the etched. What this would be like is making a 2 X 2 panorama, probably making four exposure passes. By redesigning the joining edges, the tiles would be integrated and no seam would be evident. Readout circuits may be per tile. The sensor diagram that is projected is designed on a computer screen and can be altered to fit, before the projection "negative" is created.

Stepping engines are a fancy enlargers. If you had a medium format enlarger, and a 4" X 5" negative, you could cut the negative up into four pieces (2" X 2.5") and project each part onto the appropriate quadrant of the paper. If you are good at it, there wouldn't be any seams showing and you'd have your big print.

The degree of precission involved for sub-micron accuaracy is mind boggling and X4 for our prediction. This isn't a new idea, because Kodak's medium format and the M8 sensors are done this way.

To add a wild side to Andy's prediction, imagine the 10MP 4/3rds unit X 4, for a 40MP sensor limited to low ISOs....the Kodachrome of the future? The max lp/mm for that would be 106, so Leica's prime lenses should be able to cope, though our computers would probably choke...

Bob

[ho_co]

Personal comment:

Four Thirds is a specific format system, not an aspect ratio. A lot of people have worked up their own abbreviations for the specification.

 

To avoid confusing matters, wouldn't it be better to refer to this possible new sensor as 4:3 (or 3:4) or 1.3:1 as is usually done in discussing aspect ratios?

 

Simply a preference from my side, merely a commend, no complaint intended.

 

 

Andy--

There are definitely people who like the idea of 4:3, though I'm personally far more comfortable with 3:2. But the dimensions you suggest would also allow a 5:4 ratio, something that a lot of studios could make use of very efficiently since it enlarges directly to 8x10, and something that Nikon offers on the D3.

 

The current range of R lenses wouldn't likely cover the full 4:3 format corners, but could automatically set the BF 3:2 ratio. New lenses introduced with the camera would incorporate the larger image circle. Existing lenses might cover the 5:4 ratio because that could be designed as a reduction from the 4:3 sensor coverage.

 

Marketing-wise, that means everyone comes out ahead: Legacy lens owners get to keep the 3:2 ratio they're used to, and get the option of 5:4 if they wish. Purchasers of the new lenses get those options plus the 4:3 possibility. Or have we now moved too far from Leica's "Keep it simple" philosophy?

 

--HC

[lct]

Brilliant Andy

[elansprint72]

"Nurse, .........the screens....... quickly"

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[ptomsu]

Personal comment:

Four Thirds is a specific format system, not an aspect ratio. A lot of people have worked up their own abbreviations for the specification.

 

To avoid confusing matters, wouldn't it be better to refer to this possible new sensor as 4:3 (or 3:4) or 1.3:1 as is usually done in discussing aspect ratios?

 

Simply a preference from my side, merely a commend, no complaint intended.

 

 

Andy--

There are definitely people who like the idea of 4:3, though I'm personally far more comfortable with 3:2. But the dimensions you suggest would also allow a 5:4 ratio, something that a lot of studios could make use of very efficiently since it enlarges directly to 8x10, and something that Nikon offers on the D3.

 

The current range of R lenses wouldn't likely cover the full 4:3 format corners, but could automatically set the BF 3:2 ratio. New lenses introduced with the camera would incorporate the larger image circle. Existing lenses might cover the 5:4 ratio because that could be designed as a reduction from the 4:3 sensor coverage.

 

Marketing-wise, that means everyone comes out ahead: Legacy lens owners get to keep the 3:2 ratio they're used to, and get the option of 5:4 if they wish. Purchasers of the new lenses get those options plus the 4:3 possibility. Or have we now moved too far from Leica's "Keep it simple" philosophy?

 

--HC

 

Four Thirds = FT = 4/3 .... is usually used as abbreviation for the Four Thirds System and yes it describes this format.

 

There may be other FT Formats as well in future, like maybe the new format of the R10. If that happens it might be appropriate to call this R10FT or something else. As long as we do only speculate about a new R10 format, we do not need to specify any further IMHO

[ho_co]

... it might be appropriate to call this R10FT or something else....

Take a cue from Nikon and use:

BF = Barnack Format = 3:2

4F = 4:3

5F = 5:4

or better, keep things more traditional by replacing 4F and 5F with their Roman equivalents:

IVF = 4:3

VF = 5:4

 

Or take a cue from Canon and use:

BF Mk I FF = 3:2

BF Mk II 4 = 4:3

BF Mk III 5 = 5:4

 

Or envisage it as a new, non-derivative sensor and use:

RFF = R Full Frame = 4:3

RRF = R Reduced Frame = 5:4

RBF = R Barnack Frame = 3:2

or perhaps instead of RRF, use:

RTF = R Traditional Frame = 5:4 (since 5:4 is the oldest of these aspect ratios)

 

 

Andy--see what you started!

 

--HC

 

PS--

If they include the square crop:

Nikon Style -- 1F or IF = 1:1

Canon Style -- BFs Mk I Sq = Barnack Frame square = 1:1

Non-derivative Style -- RSF = R Square Frame = 1:1