[falconeye]

You'd need some nifty optics to split the image into 4 quadrants to account of the gap between sensors.

Not necessarily.
Even a simple prism which splits the image into two images with 50% luminosity would create two virtual copies of the focal plane. You place one sensor in each one and end up with two sensors and a seamless fit (calibration required).

[Cambo]

You'd need some nifty optics to split the image into 4 quadrants to account of the gap between sensors.

Even if it a single pixel wide, there would be a row of missing pixel. And I am sure is multiple pixel wide.

you simply overlap the images in software...easily done these days. Have the right fourth row of sensor one start at the left fourth row (for example) or sensor two.

I don't know why you'd think there would have to be a gap, or why it would have to be done optically, or anything else. I've been lining up digital images since the 90's...I know modern chips could stitch a couple of sensors together easily, without any 'content aware' anything...



Cheers,
Cameron

[melander]

Make it 12 mp so that nikon users will buy it...

[eurostar]

you simply overlap the images in software...easily done these days. Have the right fourth row of sensor one start at the left fourth row (for example) or sensor two.

I don't know why you'd think there would have to be a gap, or why it would have to be done optically, or anything else. I've been lining up digital images since the 90's...I know modern chips could stitch a couple of sensors together easily, without any 'content aware' anything...

Well, if you have two sensors, each one with a border without light-gathering receptors, and just one lens, part of the image projected from the lens falls over the non-imaging borders. So, there's no way the two half images can be used to stitch a whole image together... there will be always a many-pixels-wide column of missing pixels in the middle.

[Cambo]

Well, if you have two sensors, each one with a border without light-gathering receptors, and just one lens, part of the image projected from the lens falls over the non-imaging borders. So, there's no way the two half images can be used to stitch a whole image together... there will be always a many-pixels-wide column of missing pixels in the middle.

to get rid of that part of the image...stitch together the rest.



Cameron

[Smeggypants]

to get rid of that part of the image...stitch together the rest.



Cameron

I don't think many peopel would be very happy with an image taht's missing part of the scene. What if you were taking a portrait and the person's nose happened to fall in the gap between the sensors.

Your solution would end up with an image of someone with no nose and their eyes close together.

Falcs' solution above, whoever, woulodactually work. Split the image into multiple copies and place a sensor to gather light from each qudrant or half of the respective copies. That way you wouldn't get any gaps. Indeed there could be slight overlaps to avoid having to calibrate the physical positions of the sensors s oaccurately and let software do the stitching

but why not just make a bigger sensor?

[drougge]

but why not just make a bigger sensor?

Because it's too expensive. For a given area of silicon, there is some likelyhood of unacceptable defects. If you have on average one such defect on the area of your desired sensor size, you get almost no sensors. But if you make four sensors on the same area, you get three sensors. (I'm sure there are other problems with big sensors, but even just this problem would be enough, given the actual likelyhood of defects.)

[Cambo]

I don't think many peopel would be very happy with an image taht's missing part of the scene. What if you were taking a portrait and the person's nose happened to fall in the gap between the sensors.

Your solution would end up with an image of someone with no nose and their eyes close together.

Falcs' solution above, whoever, woulodactually work. Split the image into multiple copies and place a sensor to gather light from each qudrant or half of the respective copies. That way you wouldn't get any gaps. Indeed there could be slight overlaps to avoid having to calibrate the physical positions of the sensors s oaccurately and let software do the stitching

but why not just make a bigger sensor?

I don't get the confusion...


Cameron

[eddie1960]

Because it's too expensive. For a given area of silicon, there is some likelyhood of unacceptable defects. If you have on average one such defect on the area of your desired sensor size, you get almost no sensors. But if you make four sensors on the same area, you get three sensors. (I'm sure there are other problems with big sensors, but even just this problem would be enough, given the actual likelyhood of defects.)

This was the argument against Full frame originally and medium format then medium format (645) full frame. all have come to pass. manufacturing technology improves, costs go down and eventually it comes to pass. whether they can get the flaw rate down enough to make a 67 sensor viable price wise may take some time, there is almost no chance it won't happen though. as the tech improves and the retail on the other sensors drop they are forced to drive new sales through improved technology. Cameras for a long time went through very minimal incremental changes as film was a major factor in image quality and the cameras had reached in a lot of ways the limit of what was needed. the last decade camera companies have had enormous growth driven by digital and the upgrade merry-go-round they have us on. Most certainly 6x6 and 6x7 sensors will come down the path at some point. i would not be surprised to see 4x5 sensors at some point (though these may well be technologically stitched monsters). If they stop driving the tech people will slow down their upgrade lust and they will suffer negative sales growth. I can't see anyone letting this occur willingly at the corporate level

[thibs]

But you need market for this to happen. And people want smaler cams

[noblepa]

I don't get the confusion...


Cameron

How could the sensors overlap? Look at a picture of a "naked" sensor. There are tiny wires extending out of all four sides of the sensor. You need to make room for those. Even if you somehow had the wires come out of the back of the sensor, they would have to align absolutely perfectly (and I mean absolutely, probably 1 in 10,000 accuracy) or the image would LOOK like four poorly stitched images.

In addition, the four sensors would need to be perfectly aligned in the image plane. Any variation would cause a difference in focus from one sensor to the next.

These things MIGHT be possible (personally, I doubt it), but I think it would be more expensive than a single large sensor, which would be VERY expensive.

[Arbalist]

This is what you call big - at the moment ...

Seitz Phototechnik AG - Seitz 6x17

Vic.

[eddie1960]

This is what you call big - at the moment ...

Seitz Phototechnik AG - Seitz 6x17

Vic.

now that is a lustworthy piece of gear

[eddie1960]

How could the sensors overlap? Look at a picture of a "naked" sensor. There are tiny wires extending out of all four sides of the sensor. You need to make room for those. Even if you somehow had the wires come out of the back of the sensor, they would have to align absolutely perfectly (and I mean absolutely, probably 1 in 10,000 accuracy) or the image would LOOK like four poorly stitched images.

In addition, the four sensors would need to be perfectly aligned in the image plane. Any variation would cause a difference in focus from one sensor to the next.

These things MIGHT be possible (personally, I doubt it), but I think it would be more expensive than a single large sensor, which would be VERY expensive.

see the post below yours, i didn't realize it existed very cool piece of kit

[Parliament]

see the post below yours, i didn't realize it existed very cool piece of kit

so you want the multiple image sensors to move around inside the camera to get rid of the gaps in between them?
Because that's exactly what this panorama camera does, scan the entire 6x17 image plane with an (almost) one-dimensional pixel-array (hence the low prices starting from 21.5 k€)

The fastest exposure speed is 1/2'000 per pixel - or one second for the entire 6x17 scan at full resolution!

and minimal exposure time of one second isn't really suitable for certain kinds of photography (but no problem for landscapes i guess )

Digital scan back - Wikipedia, the free encyclopedia