[barondla]

Just wondering how they pulled it off.

Thanks,
barondla

[biz-engineer]

They didn't.

[Cerebum]

Well, that was brief 😂😂😂

[Rondec]

I thought stitching sensors together was common in a variety of sensor sizes to prevent waste from flaws in the wafer.

(I am not knowledgeable on the subject, just read occasional articles on things).

[biz-engineer]

I thought stitching sensors together was common in a variety of sensor sizes to prevent waste from flaws in the wafer.

There is the possibility for connect multiple microchips into the same package , e.g multi-chip modules (MCM and variants of it, https://en.wikipedia.org/wiki/Multi-chip_module) but such assembly technology isn't suitable for image sensors (except 3D stacking).


For image sensors, it's not practically possible to stitch two sensors edge to edge without gap void of pixels. Silicon chips are singulated by a diamond saw, and due to mechanical tolerance and mechanical stress from the sawing process, there is no active circuity near the edge of silicon chips, leaving an empty space wider than several pixels (~20 micrometers without active, + ~50 micrometers for the diamond saw). Stitching multiple chips to make up one camera sensor would leave empty spaces in the image frame, and if not, the pixels at the edges would be damaged during the sawing operation. There is no currently known ways to avoid increasing cost of larger image sensors. So, one good working unit of 4"x5" CMOS sensor would cost $25 000 (without R&D and NRE costs), the market for a camera using such digital sensor would be very tiny, or you would spend the price of a house to have your own 4"x5" digital. I'd say, I live in a shelter, but I am very proud owner of a 4"x5" digital camera .

[w2ck]

Hi,

What they did isn't clear to me. My first thought is the usual multiple lithograpahy steps on one large wafer. That's what is common with sensors. Two sets, three sets even four sets as Sony does on the 100 MP sensor.

There are times when one can see these in a photo. And, I have seen the line in 645D shots and it is more defined than when I see such from other sensors.

When I look down the bore of my 645D with my microscope with polarized light, I see a defined line down the center. Looks like two chips butted together as the line is more defined than I would expect from lithography. However, I can't get a really good look at it as the distance down the bore is so great I have to use low magnification to focus the scope. What I need is a sensor out of the camera where I can get close to it and so increase the mag.

Kodak isn't saying just what they did so the only way to know is to figure it out after the fact. Anyone have a dead 645D sensor they can send me?

Stan

[kaseki]

My understanding has been that the size of the image that can be projected onto a wafer has grown over the years, and at the time of the 'D' two photo-lithographic images would have had to been projected onto the wafer for each sensor. The accuracy of the image stitching, or 'butting together' needs more tolerance than that of the details within any particular image, such as the pixel layout. Note that each half has to be read out separately; interconnections at the pixel row level would have not have been successfully aligned.

[biz-engineer]

Two sets, three sets even four sets as Sony does on the 100 MP sensor.

One guy (hobby photographer) made a very high ISO dark frame from a Fuji GFX100 and he saw the hint of two lines splitting the image in four parts. He concluded that the Sony sensor was made of four sensors stitched together. Rather , what's most likely the cause of seeing line marks from the Sony 100Mp sensor , is that the designers of the 100Mp sensor partitioned the sensor in 4 areas in order to reduce the average length of power and clock trees, and also use 4 ADC running at the same time for image readouts. The 102Mp chip may works like 4 x times 25Mpixels sensor running parallel on the same monolithic chip, manufactured like a single chip.

[mikeSF]

following with interest...

When I owned the 645D, if I compressed the dynamic range excessively in post, I could also see that seam down the center of the image, and at times there were slight variations in color response between the two sides. Curious what was happening there.

[Andrea K]

with a bit of glue?

[pentaxus]

It would have to be some super kind of glue. Kodak's eastman 910 glue worked best at molecular thinness. But maybe not here.

[Ed Hurst]

following with interest...

When I owned the 645D, if I compressed the dynamic range excessively in post, I could also see that seam down the center of the image, and at times there were slight variations in color response between the two sides. Curious what was happening there.

Me too, Mike. It wasn't obvious in 'normal' use, but could show up with extreme processing, especially at high ISOs (not that the camera was really very good at such settings anyway). I even could see a yellow line along parts of the join under some circumstances...

[itshimitis]

following with interest...

When I owned the 645D, if I compressed the dynamic range excessively in post, I could also see that seam down the center of the image, and at times there were slight variations in color response between the two sides. Curious what was happening there.

the same is the case with the GFX100s apparently.

[kaseki]

In case I was unclear, the stitching is photolithographic. No glue required. Each pair of photodetector arrays is then sawn from the wafer as a complete chip.

[cdd29]

Is this the case with the 645z?