I still don't get the crop debate

[dlleno]

ok, yes I get that in today's market, the crop sensors are less expensive, and that it is cheaper to produce many, smaller sensors (more revenue from the same silicon substrate). for that reason I don't see the crops going away soon.

but the advantages of the smaller sensors (reach) are (currently at least) overshadowed by the consequences of pixel density -- large sensors put more pixels on the image with lower pixel density, which is why the IQ is higher. captain obvious at your service

1d4 is a good example, of "best of both worlds". but why constrain the camera's capabilities to the geometric size of the silicon under the mirror? I see no technical reason (yet I do see a market reason) why a full frame sensor could not be asked to behave like any of the other crop sensors, 1.3 or 1.6, and still produce images just as good if not better as the native crop technologies they would mimic -- after all, the trade-offs of pixel density, noise, and pixel count are the same no matter what the substrate size is.

so -- other than to drive the consumer and prosumer crop markets, and perhaps experiment there with high pixel densities, there's no reason why the best FF sensor could not produce the same IQ and reach combination as the best crop sensor -- by simply cropping the FF image "in camera" -- or even out of camera, for that matter.

am I out in the weeds? Why arn't the Canon pro bodies available with selectable crop configurations? or maybe the 1dx will do that and I'm behind the times .

the only reason I can think of NOT to do that, would be that the higher pixel density would compromise high ISO work and would come up just short of matching the abilities of a good crop sensor. case in point: cropped 5D3 images are very close to the same resolution as native 7D images at the same FOV. but it seems Canon could solve this problem with a sensor that would, for example, apply a lower pixel density when needed, and a higher pixel density when required. yea, more like the D800 only do it better.

 

[JR]

I could be wrong but in terms of cropping a FF sensor, patents may be involved. Nikon is doing it with its FF sensor, but again I believe patents may be involved here.

 

[Mt Spokane Photography]

Nikon had a compelling reason to offer it. Their pro series DSLR's were crop cameras, and so were the lenses. Photographers with a ton of money invested in pro grade lenses wanted to be able to use them when Nikon went to FF. The crop mode is offered for those who bought those expensive DX lenses.

Canon did not sell DX lenses for their Pro series cameras, just FF lenses. When the Rebel Series came out, Canon offered EF-S lenses as a low cost option for a Standard Zoom and wide zoom. They will not physically fir on a FF body. So why would Canon crop a image for lenses that would not fit??

If you just want a cropped image, its a trivial thing, it can be done automatically as you import into photoshop, and, if you decide to change the crop a little, you can.

 

[K-amps]

Nikon had a compelling reason to offer it. Their pro series DSLR's were crop cameras, and so were the lenses. Photographers with a ton of money invested in pro grade lenses wanted to be able to use them when Nikon went to FF. The crop mode is offered for those who bought those expensive DX lenses.

Canon did not sell DX lenses for their Pro series cameras, just FF lenses. When the Rebel Series came out, Canon offered EF-S lenses as a low cost option for a Standard Zoom and wide zoom. They will not physically fir on a FF body. So why would Canon crop a image for lenses that would not fit??

If you just want a cropped image, its a trivial thing, it can be done automatically as you import into photoshop, and, if you decide to change the crop a little, you can.

Except that the UWA EF-S lenses outperform similarly priced UWA FF lenses in the crop area. For those who have invested in a crop body with a couple of lenses (especially hobbyists) allowing FF bodies to take EF-S lenses gives them options. Unless Canon thinks that the users are too naiive to see the EF-S Vignette on FF sensor and complain about it.

 

[mitchell3417]

It's my understanding that the mirror on a FF camera would hit the back of an ef-s lens. The only way you can shoot with an ef-s lens on a FF (assuming it would mount on a FF) is to have the mirror locked up permanently i.e. live view mode.

 

[JR]

Nikon had a compelling reason to offer it. Their pro series DSLR's were crop cameras, and so were the lenses. Photographers with a ton of money invested in pro grade lenses wanted to be able to use them when Nikon went to FF. The crop mode is offered for those who bought those expensive DX lenses.

Canon did not sell DX lenses for their Pro series cameras, just FF lenses. When the Rebel Series came out, Canon offered EF-S lenses as a low cost option for a Standard Zoom and wide zoom. They will not physically fir on a FF body. So why would Canon crop a image for lenses that would not fit??

If you just want a cropped image, its a trivial thing, it can be done automatically as you import into photoshop, and, if you decide to change the crop a little, you can.

Thanks I did not know that. Does this mean it is not related to the fact Nikon would have patents to crop a FF sensor then that Canon is not offering the same thing? I guess it would be more a marketing decision then on Canon part?

Interesting...

 

[unfocused]

I don't get the crop debate either, but for entirely different reasons.

1) You essentially answered your own question when you said that the cost of producing an APS-C sensor is less than a full frame sensor.

2) It's not correct that the advantages of a smaller sensor are overshadowed by the consequences of pixel density. At most normal ISO speeds, the differences between crop sensor and full frame are virtually undetectable in prints, on the printed page or on a website. (hence the term "pixel peeping")

3) Are you suggesting that manufacturers make full frame sensors that are of the same or similar pixel density as APS-C sensors? If so, then any benefit that might exist with a full frame sensor would disappear. You can't have it both ways – dense pixels on a full frame good, dense pixels on APS-C bad?

4) While in-camera cropping sounds good in theory, the technology isn't there yet. To accurately frame an image with in-camera cropping requires some sort of electronic viewfinder and the current crop of electronic viewfinders are a poor substitute for an optical viewfinder.

5) After spending a lifetime learning to frame images through the viewfinder, I have no interest in "guesstimating" a crop by using a full-frame viewfinder to frame a cropped image.

6) I don't understand this statement: "cropped 5D3 images are very close to the same resolution as native 7D images at the same FOV." I'm not particularly knowledgeable about the calculations (and I'm sure I'll be corrected if wrong) but I believe a 22mp full frame image cropped to APS-C is 8.6 megapixels. Which is clearly not very close to the 18 mp of the 7D.

What I cannot understand is why so many full frame users seem so upset by the very existence of APS-C bodies. There are two primary DSLR formats today and it is unlikely that either Canon or Nikon are going to abandon either one in the foreseeable future.

 

[LetTheRightLensIn]

I could be wrong but in terms of cropping a FF sensor, patents may be involved. Nikon is doing it with its FF sensor, but again I believe patents may be involved here.

That would be the height of absurd patents. How in the world can you patent cropping?? (then again some patents are pretty absurd, but still, this would really take the cake)

 

[dlleno]

ok all good points -- patents may be involved here, and there is the "crop lens" factor which I'll comment on later. As for just cropping the result of a FF image, yes of course that it trivial. after more thought:

Unfortunately, cropping the FF image does not produce equivalent results to the native crop sensor, at least today for Canon, and there's a good reason for that (at least for Canon!). I suspect there is some FF pixel density, larger than today's 5D3, that would do this at the 1.6x FOV, but Canon chose not to do that -- they chose to advantage the 5D performance in low light/high ISO, and its low-density sensor does that very well. So well, in fact, that it can't (quite) be cropped to 1.6x with equivalent performance to the 7D, which itself is aging. The 5D3 pixel density is approximately the same as the 40D of five years ago -- which means the state of the art has advanced considerably and, should they want to, Canon could produce a 10mp 1.6x crop camera with rather stunning low-light performance. Of course, no one would buy it because of the low pixel count, but it would be doable, and in fact they have a 10mp crop camera today in the 5D3 -- by cropping in post! My point is that at any particular pixel density state-of-the-art, the FFs can always be cropped in post to mimic the crop FOV , but the IQ is just not quite there yet because the crop sensors are designed to maximize resolution in favorable light conditions and will win 'by a neck' when compared with (Canon) FF sensors, which have too low of a pixel density to compete head-to-head when throwing over half of the image information away . For example, Canon could have produced a 38mp FF which (when cropped to 1.6) would have the same pixel count as a 50D, but its low-light performance would not have been as good as the 22mp they chose for the 5D3.

I'm not faulting Canon for this at all; in fact the low light performance of the 5D3 means more keepers and more corner case shots in more difficult situations, greater ability to bring out shadow detail in post, etc. . halleluiah go team go I like it.

Yea I also get that it would be silly to operate a camera body in crop mode when the "crop lenses" won't fit, and I understand Nikon's marketing choices there. But those who would appreciate in-camera crop (and FOV preview in the viewfinder) are the most likely to understand that cropping the image to to change the FOV is not about the ability to fit cheaper lenses on there, it is about changing the FOV (for more "reach") and still maintaining acceptable IQ. I'm talking here about the camera that could truly combine the reach advantages of the crop body with the IQ advantages of the FF. One FF body that, when cropped, would absolutely not be rivaled by any 1.3 or a 1.6 body, when the subject distance and lens is the same.

I do realize that producing such a camera would knock the 1D series out of the lineup because the 1.3 crop bodies would offer no advantage over the FFs of the same pixel density. and it would probably be a $12K body...

I'm not a physicist but would it not be possible to produce a sensor that would operate in more than one pixel density mode (this is a Canon forum so don't fault me for missing something about the D800 here...). I don't mean crop I mean pixel density -- for example, operate the sensor at FF 22mp with all the low light advantages thereunto appertaining, and also at FF 38mp to advantage the situations that the crop bodies depend on -- with favorable light, you get higher resolution when you don't need high-ISO, AND you can crop down to 1.6x for more "reach" and still have a 15mp image that no other natively-cropped body would be able to rival.

By the way, if we are expecting the 1Ds and 1D lines to converge, the only way I can see keeping the birders happy ,for example, is to allow the 1Dx to operate in either FF or 1.3 crop mode -- as long as that offers some advantages the 1D4 crowd will value. But I'm not up on the latest 1dx rumors so apologies if I'm missing a market clue from this design.

 

[LetTheRightLensIn]

If you just want a cropped image, its a trivial thing, it can be done automatically as you import into photoshop, and, if you decide to change the crop a little, you can.

There are a good reasons to do it in cam.

1. [WASTE OF SPACE] What if you are shooting wildlife that is distant? Why waste storage on gartbage you will trim away? Why save a full FF RAW file when the subject doesn't even fill a 2.56x smaller APS-C cropped area?

2. [SPEED] - D800 can do only a meager 4fps at FF but has the throughput to handle 25MP at 5fps and 16MP at 6fps. It would stink to have some 38MP 5Ds stuck at 4fps no matter what, no? When you might have had 6fps at APS-C crop? The latter is what makes the D800 somewhat of an all rounder and why it doesn't get trashed by the 5D3 in that regard. I mean if they can drive it 6fps 38MP then awesome and maybe they need that to counter exmor dynamic range, but if not, a crop mode sure would be awesome for speed boost.

Canon came up with sRAW and they sadly tend to only want to do what they did first no matter how great the reasons for adding additional ways to do things so I doubt they will offer crop mode to save space (and for wildlife/reach, sRAW-type space savings are worse than nothing).

 

[LetTheRightLensIn]

but the advantages of the smaller sensors (reach) are (currently at least) overshadowed by the consequences of pixel density -- large sensors put more pixels on the image with lower pixel density, which is why the IQ is higher. captain obvious at your service

it's little to do with pixel density as the D800 should finally have made clear, it's about the total surface area that makes SNR better

so -- other than to drive the consumer and prosumer crop markets, and perhaps experiment there with high pixel densities, there's no reason why the best FF sensor could not produce the same IQ and reach combination as the best crop sensor -- by simply cropping the FF image "in camera" -- or even out of camera, for that matter.

You don't gain actual reach for distant subjects by cropping, only the photosite density matters for how many pixels per duck you get.

Cropping modes on FF can no way magically add more pixels per duck.

What they can do is spare you wasting tons of storage space on giant areas you'd crop away anyway and, if the mirror box allows it, more frames per second.

but it seems Canon could solve this problem with a sensor that would, for example, apply a lower pixel density when needed, and a higher pixel density when required. yea, more like the D800 only do it better.

There is no way for a sensor to magically change photosite density. Photosites are physical things and they can't just multiply or join at will. (well i suppose they could bin them down on chip to save storage space, if you had a really, really high density sensor for the times you didn;t care about reach or detail)

 

[woollybear]

I did some back of the envelope math...and I'm not buying the cost of sensor reasoning.

If Canon is using a 300mm wafer, they will get roughly 80 sensor dies at full frame and 215 sensors at APC-C. (Don't shoot me here, I didn't account for edges, targets, etc. I just divided the area of the wafer by the sensor area). If yields drop for the FF (yields decline more for larger die) and the yield works out to even 4 to 1, then you get about 50 FF and 200 Crop sensors per wafer. If you account for pixel density I would expect the yields to even out (smaller feature sizes would see lower yields). So I think I am being somewhat biased in favor of crops here.

How much does a sensor cost? Even Intel Microprocessors (far more complex to manufacture) would cost at most 100 bucks to make. A minimum feature size of 1 micron is very crude by today's standards. Chips like that might cost a dollar or less to make. The pixel size on a 50D is a little under 5 microns.

So, given the cost of manufacture of a camera, I think the sensor itself is a minor percentage. My guess is that the full frame jacks up the cost of the camera because of the other components, or because of the complexity of assembling the larger parts in essentially the same package.

 

[dlleno]

2) It's not correct that the advantages of a smaller sensor are overshadowed by the consequences of pixel density. At most normal ISO speeds, the differences between crop sensor and full frame are virtually undetectable in prints, on the printed page or on a website. (hence the term "pixel peeping")

yes, what I mean here is that the consequences of pixel density do come into play at normal ISO speeds at the same FOV when pixel peeping. This is evidenced by the fact that a 5D image, when cropped to 1.6x, is a lot closer in actual, realizable IQ to the 7D native image than the pixel calculations would imply. So overall system resolution (I should have used the term IQ instead of resolution) is not just about pixel count. I do think it is astonishing that an 8mp (cropped) image from the 5D3 compares very well with the native 18mp image from the 7D. I wish I had the reference (sorry) but in another thread this topic was discussed, and examples posted showing this.

3) Are you suggesting that manufacturers make full frame sensors that are of the same or similar pixel density as APS-C sensors? If so, then any benefit that might exist with a full frame sensor would disappear. You can't have it both ways – dense pixels on a full frame good, dense pixels on APS-C bad?

precisely, and this is why the advantages of a smaller sensor are overshadowed by the consequences of pixel density.

4) While in-camera cropping sounds good in theory, the technology isn't there yet. To accurately frame an image with in-camera cropping requires some sort of electronic viewfinder and the current crop of electronic viewfinders are a poor substitute for an optical viewfinder.

I hear that

5) After spending a lifetime learning to frame images through the viewfinder, I have no interest in "guesstimating" a crop by using a full-frame viewfinder to frame a cropped image.

+1 to that.

6) I don't understand this statement: "cropped 5D3 images are very close to the same resolution as native 7D images at the same FOV." I'm not particularly knowledgeable about the calculations (and I'm sure I'll be corrected if wrong) but I believe a 22mp full frame image cropped to APS-C is 8.6 megapixels. Which is clearly not very close to the 18 mp of the 7D.

right, and this is the point -- actual IQ comparisons between the two show that the realizable IQ of the 8.6mp crop of the 5D are remarkably similar in actual pixel-peeping results to the 18mp image produced by the 7D. so there is something else going on here besides pixel count (i.e. consequences of high pixel density)

What I cannot understand is why so many full frame users seem so upset by the very existence of APS-C bodies. There are two primary DSLR formats today and it is unlikely that either Canon or Nikon are going to abandon either one in the foreseeable future.

+1 on that too. Especially if Canon continues to produce FF bodies with low pixel densities to advantage low noise, shadow detail, and high ISO, the cropped results will always fall short of the native crop bodies and there will always be an advantage, in normal lighting and ISO conditions, in the ability of the crop body to put "more pixels on the image" and produce a better image for the same distance-limited subject using the same lens.

I guess what this discussion has emphasized is that it would take a true dual-mode sensor -- one that could operate at more than one true density (not just crop) -- to combine the advantages of the FF and the best crop bodies. such a beast would probably not sell because the market would probably not accept the price. Yet, I remain astonished at how close the 5D3 has come to achieving equivalent IQ to the 7D at the same FOV. It also illustrates why the pro body 1D4 has been successful as a crop body -- low pixel density. It will be interesting to see what Canon does with the 5D3-era sensor technology, i.e. what the next generation of crop bodies will offer.

also, to keep the discussion on target - I am talking about distance-limited situations where the subject distance and lens is the same, i.e. where the crop bodies put more pixels on the image and the FF bodies have to throw pixels away to get the same FOV. since overall IQ is not just about pixel count, the difference in actual IQ is of interest, not just the factor of 2.56 (or 1.69) which we can all type into our calculators.

 

[dlleno]

I did some back of the envelope math...and I'm not buying the cost of sensor reasoning.

If Canon is using a 300mm wafer, they will get roughly 80 sensor dies at full frame and 215 sensors at APC-C. (Don't shoot me here, I didn't account for edges, targets, etc. I just divided the area of the wafer by the sensor area). If yields drop for the FF (yields decline more for larger die) and the yield works out to even 4 to 1, then you get about 50 FF and 200 Crop sensors per wafer. If you account for pixel density I would expect the yields to even out (smaller feature sizes would see lower yields). So I think I am being somewhat biased in favor of crops here.

How much does a sensor cost? Even Intel Microprocessors (far more complex to manufacture) would cost at most 100 bucks to make. A minimum feature size of 1 micron is very crude by today's standards. Chips like that might cost a dollar or less to make. The pixel size on a 50D is a little under 5 microns.

So, given the cost of manufacture of a camera, I think the sensor itself is a minor percentage. My guess is that the full frame jacks up the cost of the camera because of the other components, or because of the complexity of assembling the larger parts in essentially the same package.

good points -- it may be more than just yields here too. what I'm seeing is that the FF sensors have the more expensive technologies applied as well ,beyond pixel counts. There is also the economies of scale -- there may be 100 times as many 300mm wafers yielding 215 crop sensors each, compared to the number of 300mm wafers producing FF sensors. But your point is well taken that there are other consequential mfg costs to consider as well.

 

[pwp]

I shoot 1D4 bodies and will miss the x1.3 crop of the APS-H sensor. It's going to COST me.
My 300 f/2.8 will lose an effective 90mm on the FF 1DX bodies which I have on pre-order.

The pricey required solution is a new 400 f/2.8II. I like APS-H.

Paul Wright

 

[Mt Spokane Photography]

Nikon had a compelling reason to offer it. Their pro series DSLR's were crop cameras, and so were the lenses. Photographers with a ton of money invested in pro grade lenses wanted to be able to use them when Nikon went to FF. The crop mode is offered for those who bought those expensive DX lenses.

Canon did not sell DX lenses for their Pro series cameras, just FF lenses. When the Rebel Series came out, Canon offered EF-S lenses as a low cost option for a Standard Zoom and wide zoom. They will not physically fir on a FF body. So why would Canon crop a image for lenses that would not fit??

If you just want a cropped image, its a trivial thing, it can be done automatically as you import into photoshop, and, if you decide to change the crop a little, you can.

Except that the UWA EF-S lenses outperform similarly priced UWA FF lenses in the crop area. For those who have invested in a crop body with a couple of lenses (especially hobbyists) allowing FF bodies to take EF-S lenses gives them options. Unless Canon thinks that the users are too naiive to see the EF-S Vignette on FF sensor and complain about it.

But, they Do not fit on a FF, they hit the mirror and cannot be mounted for that purpose. Why have a crop mode for lenses that cannot be mounted?

As far as 3rd party lenses that will mount, I do not see Canon doing anything to support or encourage buyers of the competition.

 

[dlleno]

I shoot 1D4 bodies and will miss the x1.3 crop of the APS-H sensor. It's going to COST me.
My 300 f/2.8 will lose an effective 90mm on the FF 1DX bodies which I have on pre-order.

The pricey required solution is a new 400 f/2.8II. I like APS-H.

Paul Wright

I hear you... I may be out of touch with 1DX rumors here, so is it confirmed that "H" is going away? Interestingly enough, if the 1dx is a 27mp FF then it will equal the pixel density of the 1D4 and you can crop the FF images in post to mimic the 1.3x crop and still get 16mp "on the bird".

 

[elflord]

I did some back of the envelope math...and I'm not buying the cost of sensor reasoning.

If Canon is using a 300mm wafer, they will get roughly 80 sensor dies at full frame and 215 sensors at APC-C. (Don't shoot me here, I didn't account for edges, targets, etc. I just divided the area of the wafer by the sensor area). If yields drop for the FF (yields decline more for larger die) and the yield works out to even 4 to 1, then you get about 50 FF and 200 Crop sensors per wafer. If you account for pixel density I would expect the yields to even out (smaller feature sizes would see lower yields). So I think I am being somewhat biased in favor of crops here.

How much does a sensor cost? Even Intel Microprocessors (far more complex to manufacture) would cost at most 100 bucks to make. A minimum feature size of 1 micron is very crude by today's standards. Chips like that might cost a dollar or less to make. The pixel size on a 50D is a little under 5 microns.

So, given the cost of manufacture of a camera, I think the sensor itself is a minor percentage. My guess is that the full frame jacks up the cost of the camera because of the other components, or because of the complexity of assembling the larger parts in essentially the same package.

According to the "overnight flight next to a canon engineer" thread, the sensor is the most expensive component. http://www.canonrumors.com/forum/index.php?topic=1606.0

If you're trying to make sense of prices using a pricing model along the lines of Retail Price = Manufacturing Cost * Markup (or total production cost including amortized R&D, if you like), it won't necessarily make sense. The price is ultimately determined by supply and demand. There is markup because Canon (or any other company) will try to avoid releasing a product that doesn't enable them to make money, and competition between manufacturers constrains the amount of markup but while it's constrained, there is absolutely no guarantee that it will be a fixed multiplier.

 

[dlleno]

I shoot 1D4 bodies and will miss the x1.3 crop of the APS-H sensor. It's going to COST me.
My 300 f/2.8 will lose an effective 90mm on the FF 1DX bodies which I have on pre-order.

The pricey required solution is a new 400 f/2.8II. I like APS-H.

Paul Wright

I hear you... I may be out of touch with 1DX rumors here, so is it confirmed that "H" is going away? Interestingly enough, if the 1dx is a 27mp FF then it will equal the pixel density of the 1D4 and you can crop the FF images in post to mimic the 1.3x crop and still get 16mp "on the bird".

its confirmed I am out of touch. my apologies I could have looked that one up to learn that the 1Dx is 18.1mp. cropping to 1.3x would result in about 10.1 mp "on the bird", or about what the 40D does.

Its clear that Canon is really reluctant to push the pixel density on these FF bodies. just look at the whopping difference in pixel density between the 18mp 1DX and the 18mp 7D!

 

[LetTheRightLensIn]

yes, what I mean here is that the consequences of pixel density do come into play at normal ISO speeds at the same FOV when pixel peeping. This is evidenced by the fact that a 5D image, when cropped to 1.6x, is a lot closer in actual, realizable IQ to the 7D native image than the pixel calculations would imply. So overall system resolution (I should have used the term IQ instead of resolution) is not just about pixel count. I do think it is astonishing that an 8mp (cropped) image from the 5D3 compares very well with the native 18mp image from the 7D. I wish I had the reference (sorry) but in another thread this topic was discussed, and examples posted showing this.

If you look at the better test carried out by Romy, the 7D actually showed noticeably better detail than the 5D2 when distance limited. One guy's test didn't show as much difference, but his results were the outlier.