Full Frame or APS-C

[Craig Gerard]

Full Frame or APS-C

If you had 8,500+ individual museum specimens (some irreplaceable) to photograph for archival and educational purposes, would there be any advantage using a full frame DSLR such as the Canon 5D Mark II as opposed to using the Canon 7D?

Would the higher resolution FF have significant benefits for the intended purpose over the APS-C with subjects ranging in size from approximately 2mm to 100mm?

I use Canon as an example here because I am somewhat familiar with their attributes; but this question is not necessarily limited to Canon gear.

This setup would be used indoors under appropriate conditions.

I have some other questions relevant to this project, but will start with this one first.


Craig

[Mitch640]

I owned a 7D for awhile. The full size RAW file sizes are huge, 27Mb +/- each. Also, the 7D sensor is new technology, and uses gapless sensor sites. Also, the 7D has a new technology sensor filter, two of them actually, and is the first Canon I owned that could reproduce a good Red. I have owned and shot with all the Canon D bodies in the 1.3x and 1.6x range, except for the Dxxx series and the 10D and the new MkIV.

I think the enormous pixel size of the 7D is overkill, if all you shoot for is to show the images on the web. Anything beyond 10Mp is a waste. But for archival purposes, I think the 7D would be a good choice, since the files could print bigger and I assume, show better in projections. But again, I have no experience with the full frame 5D's or the 1Ds series bodies.

Another option you might want to investigate, are cameras with a CCD sensor. The only Canon model I know of that used CCD is the old 1D classic. I had one for years, and the images and color were on a different order from the CMOS type sensor images. Worth some investigation, if any maker ever even made a high Mp CCD sensor. Just saying, cause the CCD images are truly better.

[soldevilla]

I am working on a similar project in the end I've decided to work with my good old 350D. 8 Mpx are sufficient for most applications, publications, exhibitions ... and increase Mpx increases the processing time and storage size and if you multiply by the number of shooting images can make the project unaffordable.
And anyway, whatever you do, within two years the technology you use will be obsolete ...

[enricosavazzi]

Full Frame or APS-C

If you had 8,500+ individual museum specimens (some irreplaceable) to photograph for archival and educational purposes, would there be any advantage using a full frame DSLR such as the Canon 5D Mark II as opposed to using the Canon 7D?

Would the higher resolution FF have significant benefits for the intended purpose over the APS-C with subjects ranging in size from approximately 2mm to 100mm?

I use Canon as an example here because I am somewhat familiar with their attributes; but this question is not necessarily limited to Canon gear.

This setup would be used indoors under appropriate conditions.

I have some other questions relevant to this project, but will start with this one first.


Craig

With regards to the first two questions, my short-short answer is No.

A more extended answer is that 6-8 MB is enough for the large majority of applications, but not necessarily all. I published many pictures taken with a 3.5 Megapixel camera, and when reduced in print to 5-6 cm and placed in composite illustrations printed in scientific journals, it is impossible to distinguish them from pictures taken with a higher-resolution camera. If you need to publish a full-page cover picture of just one specimen, 8-10 Megapixels may be necessary. Anything above about 12 Megapixels is probably overkill for practical uses.

As for whether a full frame sensor gives better results than an APS-C one, it does, but the advantage is not visible in the conditions normally used for macrophotography and photomacrography of collection specimens. Also here there are caveats. If you use microscope objectives or industrial photomacrographic zoom lenses for direct projection, it is easier to cover the APS-C format without image degradation in the peripheral regions, than the full frame format.

In spite of occasional statements to the contrary, the sensor size does not affect the depth of field and loss of resolution by diffraction in macrophotography and photomacrography. If you compare pictures of the same specimen magnified and printed/displayed at the same absolute size, taken with different sensor sizes and at the same lens aperture, there is no difference in the above respects.

[NikonUser]

I can see that a small sensor and a 5 megapixel would be sufficient when the subject fills most of the frame.
But what about when the subject fills less than 50% of the frame?
Would it not be better, in such a case, to have a large sensor and more pixesls?

In this image I am interested in only what is within the red rectangle.

NU10113

[Rusty]

enricosavassi wrote:

In spite of occasional statements to the contrary, the sensor size does not affect the depth of field and loss of resolution by diffraction in macrophotography and photomacrography. If you compare pictures of the same specimen magnified and printed/displayed at the same absolute size, taken with different sensor sizes and at the same lens aperture, there is no difference in the above respects.

Am i missing something here ?
Let us assume 2 cameras 10 mp full frame and 10 mp crop , both equipped with the same 100mm macro lens both at the same aperture
the final full frame image is a box of matches
To get the same framing ( final 10 mp image) you will have more distance between the focal plane and the subject on the crop camera giving you more dof ?

[Mitch640]

Daniel, I think that's assuming you are using a lens with an aperture iris. Would DOF matter, or even be observable using a microscope lens? And it would not matter with either lens if you stack.

[enricosavazzi]

Am i missing something here ?
Let us assume 2 cameras 10 mp full frame and 10 mp crop , both equipped with the same 100mm macro lens both at the same aperture
the final full frame image is a box of matches
To get the same framing ( final 10 mp image) you will have more distance between the focal plane and the subject on the crop camera

Yes.

giving you more dof ?

No.
Regardless of working distance, focal length, crop factor (=sensor size), shooting at the same lens aperture and with the subject framed in the same way in the images, once you magnify the images to the same absolute size and compare them, DoF is the same.

[enricosavazzi]

I can see that a small sensor and a 5 megapixel would be sufficient when the subject fills most of the frame.
But what about when the subject fills less than 50% of the frame?
Would it not be better, in such a case, to have a large sensor and more pixesls?

In this image I am interested in only what is within the red rectangle.

NU10113

There is a correct answer and a right answer

The correct answer is, a larger number of pixels gives you a higher resolution (regardless of sensor size), unless the resolution is limited by the lens.

The right answer is, it is best to avoid cropping, and instead to shoot from a closer distance or with a lens of longer focal length.

A further thing to remember is that, with the same lens and at the same magnification, a full-frame camera gives you roughly twice the subject area as an APS-C. So, if you cannot get any closer and/or increase the magnification, a full frame camera makes things worse (or at least gives you a larger area to crop away in post-processing). Let's say that the useful area in your picture above is 40% of the surface of a picture shot with a 10 Mp APS-c sensor. So 60% of the picture area is wasted, and the cropped picture has a resolution of 4 Mp. If you shoot the same subject at the same magnification with a 10 Mp full-frame sensor, the subject area is twice the one shot with the smaller sensor. Cropping in the same way means that only 20% of the sensor area is used, which produces a picture of only 2 Mp. In these conditions, you would need a 20 Mp full-frame sensor to give you the same resolution (in the cropped area) as a 10 Mp APS-C sensor.

[mgoodm3]

As long as the lens can adequately cover ther larger detector, the larger detector will slightly outperform the smaller detector in theoretical resolution even at the same megapixel rating with the same field of view. this advantage will be mainly at low magnification and will disappear as the magnification rises.

As an example:

DX detector at m=1 f/4. The effective aperture is 8. The Airy disc size is proportional to the effective aprture. We'll call the pixel size as 1. So the ratio is 8:1

FX detector same FOV m=1.5, f/4. the effective aperture is 10. The pixel size is 1.5. The ratio is slightly less at 6.7:1. that means slightly improves resolution.

That means that the FX detector will stay detector limited longer than the DX detector will as the FOV gets smaller.

[rjlittlefield]

In spite of occasional statements to the contrary, the sensor size does not affect the depth of field and loss of resolution by diffraction in macrophotography and photomacrography. If you compare pictures of the same specimen magnified and printed/displayed at the same absolute size, taken with different sensor sizes and at the same lens aperture, there is no difference in the above respects.

The conclusion is correct, but the explanation may be misleading.

Rather than "at the same lens aperture", it should say "with the same entrance cone".

When this is correct, the effective aperture expressed as an f-number will scale in proportion to sensor size. For example effective f/16 on full-frame 35 mm is equivalent to about f/10 on a 1.6 crop sensor and equivalent to about f/114 on 8"x10" film. In each case, the cone of light accepted by the lens will be the same.

This relationship is simple only when it is expressed in terms of the effective f-numbers. The corresponding nominal f-numbers -- the ones that you would set on the ring of a lens -- change in a complicated way depending on magnification and sensor size.

--Rik

[rjlittlefield]

As long as the lens can adequately cover ther larger detector, the larger detector will slightly outperform the smaller detector in theoretical resolution even at the same megapixel rating with the same field of view. this advantage will be mainly at low magnification and will disappear as the magnification rises.

As an example:

DX detector at m=1 f/4. The effective aperture is 8. The Airy disc size is proportional to the effective aprture. We'll call the pixel size as 1. So the ratio is 8:1

FX detector same FOV m=1.5, f/4. the effective aperture is 10. The pixel size is 1.5. The ratio is slightly less at 6.7:1. that means slightly improves resolution.

That means that the FX detector will stay detector limited longer than the DX detector will as the FOV gets smaller.

But notice that in this case the effective f-number did not scale in proportion to the sensor size. The higher resolution for the larger sensor is accompanied by less DOF.

This analysis is appropriate under some conditions. Mark has essentially assumed that the same lens is used at the same aperture setting but at two different magnifications, and that the lens is diffraction-limited in both cases. This may or not be true for any particular real lens.

--Rik

[rjlittlefield]

Cutting back to the core question, I agree with Enrico's analysis. Most cameras have more pixels than you need, and the advantages of a larger sensor are weak at best.

Operational features of the body are far more important. You want EFSC (electronic first shutter curtain, currently found only on Canon cameras), accompanied by a good live view and long shutter life. I have not researched these issues between the 5D Mark II and 7D.

In terms of the overall project, perhaps the most important aspect is to be sure that it can be accomplished within budget and schedule. 8500 specimens are a lot to shoot and process.

Careful thought should be given to what resolution is needed (desired). Once that decision is made, then the whole process can be optimized in terms of what pixel counts and aperture settings should be used, to produce the smallest and shortest stacks. These decisions will strongly affect the time needed to shoot and to process the stacks.

It may be that the first cut produces an estimated time or budget that is too large, in which case one would need to reconsider where the requirements might be reduced or the resources increased.

--Rik

[Chris S.]

I'd think long-term here--not about what is needed now, but what may ever be needed. Subject-handling--removal from storage, positioning, lighting, and returning to storage--will be the biggest investment (assuming limited human intervention in post processing); the hope should be to do the initial capture so well that it never has to be done again. So I'd capture and archive images containing the most information possible. You may use a batch job to make smaller conversions of these images for stacking, but if the full information exists in archive, the widest set of options exists for the future.

I'm not, of course, suggesting that larger files necessarily contain more information than smaller ones. Just that file size concerns, for a project like this, should be treated as irrelevant; data storage is cheap and getting cheaper.

As for maximizing information capture, others here know much more than I do, but I'd expect it to depend substantially on the optics used. Since some of the theory gets beyond me, and I'm an empiricist at heart, I've played around with my optics and come away with the rule of thumb to choose a sensor that uses most of the best portion of my image circle. I have both full-frame and crop-frame cameras, but for most of my optics (microscope objectives), a crop sensor fits the image circle better than a FF one, so I standardize around that.

Another step I'd seek to do only once is any substantial, non-automated post processing. For so many images, it seems likely that most of them will receive only automated stacking. For these, I see little downside in stacking smaller versions of the archive files, as this can be redone at full resolution with little expense other than CPU cycles. But for any image that receives much human attention in PP, I'd first go back and restack with full-res images, then do the retouching.

Like some others, I also prefer a CCD imager to CMOS under studio conditions. Not sure that EFSC would be important in this situation, as it's hard to image continuous light being advisable for such a high-throughput project. For high-throughput applications, a fault-tolerant workflow is advisable, and using flash buys a lot of fault tolerance. With flash lighting, EFSC is not important. In fact, for such a high-throughput scenario, I'd question whether live view might heat up the sensor enough to degrade the images.

--Chris

[rjlittlefield]

Not intending to get personal here, but just making sure all the issues are on the table... As far as I can remember, Chris S. is speaking from a position of having never used EFSC, having expressed preference for looking through an angle finder as opposed to watching a monitor, and using finite objectives on bellows as opposed to infinity objectives on telephoto. It's important to keep that background in mind when considering issues about a new project to be performed by other people.

Regarding EFSC and liveview, there is certainly no need to use those all the time. But in my personal experience, which includes initially not having those capabilities and then acquiring them, it would be a serious loss of flexibility and convenience to omit them from a new system. In the system I took on travel HERE, I did framing and focus limits using liveview with continuous illumination, then turned it off while actually shooting stacks using flash. This was sufficiently fault-tolerant that I was able to get a good 10X stack even when somebody sat wiggling on the other end of the table. But when I'm using that same equipment in my regular environment, I routinely shoot with continuous illumination and EFSC in order to get more flexibility with the lighting while still getting images that are sharp even with a 40X objective. So from my standpoint, being sure to have those capabilities available is a simple decision.

Coverage of an objective is a big issue with direct projection, much less so with infinity on telephoto. All that's needed is to choose the telephoto to match the sensor size. 200 mm on a 1.6 crop sensor gives rated magnification with Nikon CFI60 objectives. 300 mm on FF covers almost the same size subject at the same resolution, just scaling up the magnification by a factor of 1.5. Bottom line is that using modern optics, I don't see that there's much advantage one way or the other on sensor size for this issue.

The one clear advantage of a larger sensor is that in principle it's quieter because it can count more photons. Assuming the same technology and pixel counts, a 1.6 crop sensor also has 1.6X higher noise levels. But on their lowest ISO setting, I suspect that all these cameras are more than quiet enough for the intended purpose.

On the issue of how much resolution to acquire, I lean in the same direction as Chris. Certainly for all of my own stacks, I shoot at highest resolution -- which also implies highest frame count -- and then downsample in processing if appropriate. But in my entire life I've shot only a few hundred stacks. Shooting 8,500+ stacks is a different ball game. So it's worth explicitly thinking about the tradeoffs.

--Rik