Who wouldl like a camera with a SQUARE sensor?

[Ben_Egbert]

99% of all my shots will only be seen on a PC. I never shoot portrait mode for that reason. The aspect ratio should match the monitor. Ideally, the 24-inch monitor I use at work is better than the 21-inch I use at home, because it is closer to the 35MM format.

Ben

The medium format world has done well with square image capture:

1) You never have to tilt the camera sideways, just crop as needed.

In which case you have to crop every single picture. horrible
workflow...

2) A rectangle wastes more useable lens image projection than a square

Unless you're printing square pictures, a square format wastes a
lot more...

Try this. Take a lens with a 100mm image circle. It fits:

Square, 70.7mm x 70.7mm. Total area 5000mm2.

Rectangle (4:3), 80mm x 60mm. Total 4800mm2, an incredible 4% less
than the square.

Rectangle (3:2), 83.2mm x 55.5mm. Total area 4615mm2, a whopping
7.7% less than the square.

Now, lets try to do something really radical, and make a print. How
bout an 11x14? We crop like this...

Square, 70.7mm x 55.6mm. Total area 3927mm2.

Rectangle (4:3), 76.4mm x 60mm. Total 4582mm2, 16.7% more than the
square.

Rectangle (3:2), 83.2 x 55.5mm. Total area 4615mm2. That's the same
as the square.

But this won't work for consumer digicams, because you can't just
plug it in and print it without sitting down and thinking about how
you want to crop the picture.

No. You have to do something really radical, like visualizing and
planning the picture, instead. Like the master photographers, who
almost always used the rectangular view camera.

That leaves DSLRs, most of which already suffer from sensors that
don't utilize enough of the image projected by the existing lens.

Well, you've just seen that the rectangular pictures utilize at
least as much (and usually more) of the image circle of the lens.

So if fullf rame DSLRs are to become the digital version of
mainstream medium format, there would be a place for square
sensors. Who would want one?

Very few people. Look at film cameras, the square MF makers
(Hasselblad and Rollei) both went belly up. The rectangular 645,
6x7 and 6x9 makers survived.

The snapshot "shoot it square, crop later" philosophy works when
you're trying to maximize the productivity of a high paid
photographer, by throwing low paid "croppers" at the final product.
Its great for the wedding and portrait mills. Not so good when
there isn't a low paid "sweat shop" cleaning up after you.

But there's a reason every art or craft emphasizes the quality of
the input of a process. Why we teach visualization as the first
step in photography. Why we have sayings like:

Measure twice, cut once.

Garbage in, garbage out

You can't inspect quality into a product.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[BJL]

Try this. Take a lens with a 100mm image circle. It fits:
...
Now, lets try to do something really radical, and make a print. How
bout an 11x14? We crop like this...

Square, 70.7mm x 55.6mm. Total area 3927mm2.

Rectangle (4:3), 76.4mm x 60mm. Total 4582mm2, 16.7% more than the
square.

Rectangle (3:2), 83.2 x 55.5mm. Total area 4615mm2. That's the same
as the square.

Nice example Joe, and perhaps an indication of why MF film, and now MF and LF digital backs, seems to be going mostly in the direction of 4:3, not square (or 3:2).

Except for today's offical announcement of the Fuji digital back with its funky "A paper" shape of sqrt(2) =1.41...

 

[Mathias Tornqvist]

2) A rectangle wastes more useable lens image projection than a square

Unless you're printing square pictures, a square format wastes a
lot more...

Many of us prefer square images. We think they're purdy.

Try this. Take a lens with a 100mm image circle. It fits:

Square, 70.7mm x 70.7mm. Total area 5000mm2.

Rectangle (4:3), 80mm x 60mm. Total 4800mm2, an incredible 4% less
than the square.

Rectangle (3:2), 83.2mm x 55.5mm. Total area 4615mm2, a whopping
7.7% less than the square.

Now, lets try to do something really radical, and make a print. How
bout an 11x14? We crop like this...

Square, 70.7mm x 55.6mm. Total area 3927mm2.

Rectangle (4:3), 76.4mm x 60mm. Total 4582mm2, 16.7% more than the
square.

Rectangle (3:2), 83.2 x 55.5mm. Total area 4615mm2. That's the same
as the square.

Here's an easier math - let's say you want an 8x10 print. This is 80 sq/inches. On a 35mm negative (or a typical DSLR) you'd take an 8x12 image (96 sq/inches) and take a little the top or bottom. 16 sq/inches wasted. On square, you'll take a 10x10 image (100 sq/inches), and get a choice on vertical or horizontal. 20 sq/inches lost. This is not radically more than 16 sq/inches, but a lot more flexibility is gained.

No. You have to do something really radical, like visualizing and
planning the picture, instead. Like the master photographers, who
almost always used the rectangular view camera.

If you've ever worked with a magazine editor or graphic designer, you'll have found that your visualizing and planning doesn't mean much when he needs a vertical, and all you have is a horizontal.

Well, you've just seen that the rectangular pictures utilize at
least as much (and usually more) of the image circle of the lens.

I don't think we have a shortage of lens circle coverage here. What we're short on is pixel resolution.

Very few people. Look at film cameras, the square MF makers
(Hasselblad and Rollei) both went belly up. The rectangular 645,
6x7 and 6x9 makers survived.

Hasselblad and Rollei went belly up? Hmmm... Last time I checked they were still around. And don't forget Mamiya, although they (like Hasselblad) also make 645 equipment.

The snapshot "shoot it square, crop later" philosophy works when
you're trying to maximize the productivity of a high paid
photographer, by throwing low paid "croppers" at the final product.
Its great for the wedding and portrait mills. Not so good when
there isn't a low paid "sweat shop" cleaning up after you.

I think you'll find that the portrait mills and wedding people use 645 for film economy.

But there's a reason every art or craft emphasizes the quality of
the input of a process. Why we teach visualization as the first
step in photography. Why we have sayings like:

Measure twice, cut once.

Garbage in, garbage out

You can't inspect quality into a product.

I don't think these sayings originated discussing preferences in film formats...

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[BJL]

Many of us prefer square images. We think they're purdy.

A small fraction of photographers though, so do not get your hopes up about the digital camera market accomodating this niche very well

Here's an easier math - let's say you want an 8x10 print. This is
80 sq/inches. On a 35mm negative (or a typical DSLR) you'd take an
8x12 image (96 sq/inches) and take a little the top or bottom. 16
sq/inches wasted. On square, you'll take a 10x10 image (100
sq/inches), and get a choice on vertical or horizontal. 20
sq/inches lost. This is not radically more than 16 sq/inches, but a
lot more flexibility is gained.

Why have you skipped Joe's third option of 4:3, despite it being the dominant one in digital photography, both in overall volume and in the high end of digital backs? To get that 8x10" print from a 4:3 shape, print 8x10 2/3" and crop away about 5 square inches, far better than with either of your "extreme" shape options.

... And don't forget Mamiya, although they (like Hasselblad)
also make 645 equipment.

More to the point, they do not make square at all; only 5:4 (67 format) and 4:3 (645 format).

P. S. It is really true that serious photographers are incapable of either (a) judging composition even at the roughest level of choosing between vertical and horizontal correctly most of the time, or (b) taking two photos, one vertical, one horizontal, when they are unsure?

One thing that is far cheaper than either sensor photosites or image circle area is digital storage for multiple takes on the same subject.

 

[BobTrips]

99% of all my shots will only be seen on a PC. I never shoot
portrait mode for that reason. The aspect ratio should match the
monitor. Ideally, the 24-inch monitor I use at work is better than
the 21-inch I use at home, because it is closer to the 35MM format.

Same here (in terms of viewing, but I do shoot verticals at times). I've never been a 'printer'. Before digital I shot slides and looked at them on a screen. Once I could see my pictures on a monitor I got rid of my projector and screen.

My guess is that more and more people are going to do the majority of their viewing on monitor/TV, especially HDTV as it becomes more common.

And as I walk past the electronics sections of stores I'm noticing that TVs are getting very wide. If that's the type of viewing device that becomes common in homes I suspect that we'll see a demand for images that fill the screen from corner to corner. We may find ourselves shooting in moderate panorama format in a few years.

I suppose one could always display three square images side-by-side....;o)

--
bob
Latest offering - 'Two Hours in Delhi'
http://www.pbase.com/bobtrips
Shots from a bunch of places (esp. SEA and Nepal).
Pictures for friends, not necessarily my best.

http://www.trekearth.com/members/BobTrips/photos/
My better 'attempts'.

 

[Victor Engel]

To get the full benefit of the lens, what you really need is a circular sensor.

 

[Ed Pavelin]

Very few people. Look at film cameras, the square MF makers
(Hasselblad and Rollei) both went belly up. The rectangular 645,
6x7 and 6x9 makers survived.

Actually, Hasselblad, Rollei and Bronica are all still making 6x6 cameras. And they are still very popular in the MF world.

Ed.

 

[Jerry Kindall]

Some sort of anamorphic adapter that spread out a 3:2 or 4:3 rectangle over a square sensor would allow more pixels to be captured without reducing the size of the pixels and increasing noise. It would also reduce the effects of the mosaic in one direction. Of course, right now, SLR sensors aren't even full-frame, so that's the way to go to get more resolution for now.

 

[Mathias Tornqvist]

Many of us prefer square images. We think they're purdy.

A small fraction of photographers though, so do not get your hopes
up about the digital camera market accomodating this niche very well

True, but square is a completely valid choice that shouldn't be ignored.

Why have you skipped Joe's third option of 4:3, despite it being
the dominant one in digital photography, both in overall volume and
in the high end of digital backs? To get that 8x10" print from a
4:3 shape, print 8x10 2/3" and crop away about 5 square inches, far
better than with either of your "extreme" shape options.

I do like 4:3 a lot, but in DSLR's (which is what most pro or semi-pro people use) there's only the Olympus E-1. All others (I think) are 3:2, which is too wide most of the time.

... And don't forget Mamiya, although they (like Hasselblad)
also make 645 equipment.

More to the point, they do not make square at all; only 5:4 (67
format) and 4:3 (645 format).

Ah, you're correct. Well, at least they used to make squares... (which I guess supports your point).

P. S. It is really true that serious photographers are incapable of
either (a) judging composition even at the roughest level of
choosing between vertical and horizontal correctly most of the
time, or (b) taking two photos, one vertical, one horizontal, when
they are unsure?

I always know right away which orientation I want, but trying to predict what a fickle graphic designer wants is a whole different game.

One thing that is far cheaper than either sensor photosites or
image circle area is digital storage for multiple takes on the same
subject.

Ah, except what's not cheap is your time!

 

[Dominic Groß]

n.t.

 

[Joseph S Wisniewski]

2) A rectangle wastes more useable lens image projection than a square

Unless you're printing square pictures, a square format wastes a
lot more...

Many of us prefer square images. We think they're purdy.

You are "many" in the sense that you all could get together and have a nice picnic. You are not "many" in the sense that anyone will ever spend any real R&D money developing a product for you again, ever.

Now, lets try to do something really radical, and make a print. How
bout an 11x14? We crop like this...

Square, 70.7mm x 55.6mm. Total area 3927mm2.

Rectangle (4:3), 76.4mm x 60mm. Total 4582mm2, 16.7% more than the
square.

Rectangle (3:2), 83.2 x 55.5mm. Total area 4615mm2. That's the same
as the square.

Here's an easier math - let's say you want an 8x10 print. This is
80 sq/inches. On a 35mm negative (or a typical DSLR) you'd take an
8x12 image (96 sq/inches) and take a little the top or bottom. 16
sq/inches wasted. On square, you'll take a 10x10 image (100
sq/inches), and get a choice on vertical or horizontal. 20
sq/inches lost. This is not radically more than 16 sq/inches, but a
lot more flexibility is gained.

It's an "easier math" because it's wrong. You neg.lected to keep image circle the same in both examples.

No. You have to do something really radical, like visualizing and
planning the picture, instead. Like the master photographers, who
almost always used the rectangular view camera.

If you've ever worked with a magazine editor or graphic designer,
you'll have found that your visualizing and planning doesn't mean
much when he needs a vertical, and all you have is a horizontal.

I disagree, but that's philosophical. We'll leave it for another day.

Well, you've just seen that the rectangular pictures utilize at
least as much (and usually more) of the image circle of the lens.

I don't think we have a shortage of lens circle coverage here. What
we're short on is pixel resolution.

And that's exactly why you should try to match the sensor orientation to the picture format. Unless you're final output is square, a square always gets cropped. 20% loss on an 8x10 is the best case. 23% loss on a full page. 35% loss on a tabloid. THe "average loss" is going to be somewhere around 28%.

A rectangular format cuts your losses dramatically. The 8x12 only loses 17% on the worst case 8x10, 3% on the tabloid. On the averge, you lose about 8% of your expensive pixels.

The 4:3 loses about 8% for either 8x10 or tabloid. And the average loss is something on the order of 2%. (OK, maybe those "ideal format" and 4:3 folks are on to something after all).

Of course, the losses are higher on the worst case of "swapping orientations" from horizongal to vertical. But when you consider the small percentage of shots on which you attempt this in the first place, and the even smaller percentage of shots where it will actually yield something useful, you're talkign about severely compromising a camera (25% wasted pixels on just about every shot, for that square shooter) to protect for a 1 in a thousand emergency.

Very few people. Look at film cameras, the square MF makers
(Hasselblad and Rollei) both went belly up. The rectangular 645,
6x7 and 6x9 makers survived.

Hasselblad and Rollei went belly up? Hmmm... Last time I checked
they were still around.

For Hasselblad, "still around" means they're owned by a Hong Kong industrial equipment distribution company. The German facility no longer has R&D people, it only manufactures the old Hasselblad designs. The new products areFuji cameras with Hasselblad lables.

For Rollei, "still around" is even more marginal...

1982: Approximatly 88% of Rollei liquidated, the remainder bought by United Scientific Holding (UK)
1987: Sold to Schneider Kreuznach (Germany)

1992: Sold to Samsung (Korea)

1998: Sold to a group of Rollei executives (Germany)

2002: Sold to an investment group (Scandanavia).

Each sale results in even more reduction in design capability...

And don't forget Mamiya, although they
(like Hasselblad) also make 645 equipment.

I remember Mamiya. They make a very nice camera rotator. Can't say I've seen one of the cameras in a while...

The snapshot "shoot it square, crop later" philosophy works when
you're trying to maximize the productivity of a high paid
photographer, by throwing low paid "croppers" at the final product.
Its great for the wedding and portrait mills. Not so good when
there isn't a low paid "sweat shop" cleaning up after you.

I think you'll find that the portrait mills and wedding people use
645 for film economy.

But there's a reason every art or craft emphasizes the quality of
the input of a process. Why we teach visualization as the first
step in photography. Why we have sayings like:

Measure twice, cut once.

Garbage in, garbage out

You can't inspect quality into a product.

I don't think these sayings originated discussing preferences in
film formats...

They're panacea, applicable to any art, craft, or science.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

n.t.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

Even better, get a circular or hexagonal sensor. That way you can
use all (or nearly all) of the "image circle" and dynamically
switch between square, 4:3, 3:2, or full circular output.

Probably about a year ago, I made a post to this site where I calculated the percentage utilization for round, square, hexagonal, and rectangular sensors of different ratios.

The hexagonal wins, as fitting the largest number of images with the least average utilization.

Circular sensors have to be cut from the wafer as hexagons, anyway. It's the optimal packing shape.

The circular output will be limited to RAW files, because jpegs and
tiffs must be rectangular.

Unless you give the JPEG and TIFF files black borders (or alpha transparent borders).

My sugar and cafine intake is definitely way too high today.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[tony martie]

My Kodak Pro Back Plus has a square sensor. It's a nice change when using on the 4x5 camera, no more rotating the back. When on the Hasselblad it's, um.... still just like using a Hasselblad!

 

[Leon Wittwer]

I do a lot of panoramas. For a given number of pixels, I prefer the 3:2 ratio as this maximizes the number of pixels in one direction. This maximizes the resolution that I can get in a 1-D panorama (2-D panoramas are a lot more hassle). The slight penalty is that I have to do more stitches for a given long dimension.
--
Leon
http://pws.prserv.net/lees_pics/landscapes.htm

 

[BillDrew]

Good points and calculations, however the underlying assumption is that sensor pixels are expensive. True today, but not likely to be true in a few years. If Moore's law holds, a very few years.

When sensor pixels become cheap so camera manufacturers can no longer use pixel count as "Mine is Bigger Than Yours" marketing, I expect a square format will emerge. It will be sold as variable format. There are a couple of cameras doing a bit of that now giving the choice of 4:3 or 6:4 aspect ratios. That choice will become very wide with a square format.

...
And that's exactly why you should try to match the sensor
orientation to the picture format. Unless you're final output is
square, a square always gets cropped. 20% loss on an 8x10 is the
best case. 23% loss on a full page. 35% loss on a tabloid. THe
"average loss" is going to be somewhere around 28%.

A rectangular format cuts your losses dramatically. The 8x12 only
loses 17% on the worst case 8x10, 3% on the tabloid. On the averge,
you lose about 8% of your expensive pixels.

The 4:3 loses about 8% for either 8x10 or tabloid. And the average
loss is something on the order of 2%. (OK, maybe those "ideal
format" and 4:3 folks are on to something after all).

Of course, the losses are higher on the worst case of "swapping
orientations" from horizongal to vertical. But when you consider
the small percentage of shots on which you attempt this in the
first place, and the even smaller percentage of shots where it will
actually yield something useful, you're talkign about severely
compromising a camera (25% wasted pixels on just about every shot,
for that square shooter) to protect for a 1 in a thousand emergency.

 

[toughluck]

1. Camera bodies have to cost some. Fortunately, the price is simply a treshold that cannot be overcome, but does not change much with time and we can place it around 100?, applying to consumer p&s cameras only (for body+flash+electronics).
Having set that aside...

2. I've had a chance to catch a glimpse of a 1997/98 catalogue, featuring digicams among other cameras. All the cameras there had 0.3 mpix resolution. They cost 500? (? was not estabilished then yet).

Moore's law states that every 18 months (1.5 years), IC technology increases its performance by a factor of 2x, and its price drops.

Assuming a very conservative (and, it should be noted, incorrect) approach to the subject by increasing planar resolution by a factor of 2x (linear resolution increases only by 40%, hardly an increase in actual resolution). Noise levels drop, colour rendition and overall image quality increases.
1997: 0.3 mpix cameras
1998/99: 0.6 mpix
2000: 1.2 mpix
2001/02: 2.4 mpix

2003: 4.8; OK, so Moore's law seems to hold so far, right? Wrong. Truth is, before 2000, there was hardly a 0.6 mpix camera, all of them (consumer ones) were 0.3 mpix.
2004/05: 9.6 mpix
2006: 19.2 mpix
And quantum mechanics want to take their stance. It will not work.

Then there's the matter of price. Having set the body manner aside ( 100?), we are left with a 1997 camera with a sensor costing 400?. Assuming a (very) safe 25% consistent and constant drop in relative pricing, the 1998/99 camera should have cost 400?.
2000: 325?
2001/02: 269?
2003: 227?
Do 5 mpix cameras cost that little? No.

Going further, we should expect that 9.2 mpix consumer cameras will cost 195?, and 19.6 mpix ones (in 2006) - 172?.

Assuming an expotential increase in resolution (4x spatial factor (linear^2)), we should have had:

1998/99 cameras having 1.2 mpix. 2000 cameras: 4.8 mpix; 2001/02: 19,2 mpix; 2003: 76.8 mpix; 2004/05: 307,2 mpix (wow, right?). All those megapixels in around a year! Of course, quantum mechanics would start raising their voice in 2001, but what the heck.

Of course, we should have a virtually noiseless images up to ISO 400, maximum ISO of 6400 (now), and noise at lower level than cameras had at ISO 400 in 1997.

Let's draw an analogy. Vacuum Tubes. They cost a lot, but are available. As soon as a new VT design hits the market, its predecessors cost less for a while, and then disappear, completely replaced by the new series, while their price stays relatively the same. Same happens to digital cameras. Older, lower resolution, sensors are discontinued, and cameras based on them stop selling after having been rebated already more than 10 times.

Another valid analogy: a sensor (or rather the silicon wafer from which one is cut) follows the pricing pattern similar to gold's. Increasing yields is impossible. One bad pixel is bad, and would be unacceptable, even in consumer digital cameras. Smaller pixels don't help yield at all. To drop prices of sensors, somebody would have to invent a process that ignores any flaws in silicon, and always makes the perfect sensor.

Why do CPUs adhere to Moore's law? Because they can be scaled proportionally. Smaller nanometer paths are now easily achieved, they allow for higher processing power, better heat dispersion, which helps yields if there is a critical flaw. Still, yields stay at an ominous level of 40%. But each CPU is extremely small, around 1cmx1cm. This is close to 1/1.8" sensors, but in sensors, there may not be any critical error.

Then, wafer manufacturing business is growing (no pun intended*), but most of the wafers supply CPU makers. Larger chips will always be expensive. Wafers are expensive, and only a limited number of sensors can be cut from them. It's around 6 of FF sensor per wafer, but 20 for APS-sized ones. Yields of the latter are much greater not only in percents, but also absolute numbers.

) silicon wafers are made by 'growing' them.

When sensor pixels become cheap so camera manufacturers can no
longer use pixel count as "Mine is Bigger Than Yours" marketing, I
expect a square format will emerge. It will be sold as variable
format. There are a couple of cameras doing a bit of that now
giving the choice of 4:3 or 6:4 aspect ratios. That choice will
become very wide with a square format.

...
And that's exactly why you should try to match the sensor
orientation to the picture format. Unless you're final output is
square, a square always gets cropped. 20% loss on an 8x10 is the
best case. 23% loss on a full page. 35% loss on a tabloid. THe
"average loss" is going to be somewhere around 28%.

A rectangular format cuts your losses dramatically. The 8x12 only
loses 17% on the worst case 8x10, 3% on the tabloid. On the averge,
you lose about 8% of your expensive pixels.

The 4:3 loses about 8% for either 8x10 or tabloid. And the average
loss is something on the order of 2%. (OK, maybe those "ideal
format" and 4:3 folks are on to something after all).

Of course, the losses are higher on the worst case of "swapping
orientations" from horizongal to vertical. But when you consider
the small percentage of shots on which you attempt this in the
first place, and the even smaller percentage of shots where it will
actually yield something useful, you're talkign about severely
compromising a camera (25% wasted pixels on just about every shot,
for that square shooter) to protect for a 1 in a thousand emergency.

 

[Phillip Bonfiglio]

The medium format world has done well with square image capture:

1) You never have to tilt the camera sideways, just crop as needed.
2) A rectangle wastes more useable lens image projection than a square

But this won't work for consumer digicams, because you can't just
plug it in and print it without sitting down and thinking about how
you want to crop the picture.

That leaves DSLRs, most of which already suffer from sensors that
don't utilize enough of the image projected by the existing lens.

So if fullf rame DSLRs are to become the digital version of
mainstream medium format, there would be a place for square
sensors. Who would want one?

--

My Gallery:
http://www.pbase.com/thabear

 

[BillDrew]

I agree that applying Moore's law to digicams is a bit of a stretch. For sure I don't expect pixel count to increase at that rate for a whole bunch of reasons. However my 3 1/2 year old camera sold for $1,000. Comparable cameras, or better cameras, now cost about $250-300. Pretty much in line with the price dropping in half about every 18 months. The drop in price for the whole camera isn't likely to continue to follow that path, but the electronic parts will drop in price faster than the lens and body.

I expect that pixels will continue to get cheaper much faster than they get more numerous. With cheap pixels, the argument against a square format goes away. It is likely to be implemented as a "Multi Format Camera" with square as one of the "Formats". The other "Formats" would result from simply throwing away cheap pixels.

 

[Dominic Groß]

All cost/wasting issues aside (you explained it perfectly) I don't think that less wide formats than 3:2 are that usable. One would have to crop very often as oposed to cropping 3:2 rarely to get square picture when one needs it. This is by the way what I did not like about four thirds, the aspect ration. I would rather prefer a wider sensor like 16:9 or even 1.85:1. The sensor height could stay as it is now in the 1.5x - 1.7x DSLRs and the sensor could be wider so that one can select some sort of a landscape/panorama mode that is wider. The viewfinder could be masked off like Sigma's but with a switchable mask...

--
Regards from Old Europe,
Dominic

http://www.pbase.com/sigmasd9/dominic_gross_sd10