Sharper Macro than Canon 5d2 and 100mm Macro?

[smahn]

Hi,

Other than throwing some cash at it and get better quality lenses, which only hurts the wallet, there seems to be an issue with Raw conversion/sharpening as well. The reason for wanting an AA-less camera is not clear to me, because the aliasing artifacts that produces will also complicate Capture sharpening. Maybe throwing money at improving the hardware won't solve your problem entirely?

Well I'm asking about the lens/chip combination. I don't know how explore the hardware beyond that. But in chip sample comparisons I tend to see better sharpness/detail/resolving power from non AA chips. But I'd be happy to see examples that demonstrate otherwise. And if you can demonstrate that the 5d2 is as good as it gets, I'd be happier still. 

Now this may be because you are attempting to wring more performance out of the lens than there is, but the postprocessing tools/skills may also need to be improved. Default sharpening settings are notoriously stupid, they do not take the camera settings into consideration, while the aperture and magnification factor usually pretty much dictate the required settings and are therefore relatively simple to set up correctly.

Cheers,
Bart

I don't use default sharpening settings, I use, to varying degrees: LR at max amount and zero detail, and sometimes high/max detail and low amount. Photo Ninja. Topaz Focus. Topaz Detail. Smart Sharpen at high and low radii. USM at high and low radii. Which is simply to say, I'm not ignoring the PP.

I'm not trying to put one piece of hardware over another, or hardware over technique. I'm looking to refine and evolve all of these. I just thought it would be easier to start at the root: the hardware at the capture itself (which this thread was about). Then the PP would come after and be tailored to it.

[John Koerner]

Your new Sigma is definitely a nice lens.

This being said, initial statement of the OP: "I'm not doing true 1:1 macro, more like 1:4".

Cheers,
Bernard

True. Which actually makes a 180mm even more ideal, because you have so much more ability to isolate the background when you don't have to be close enough for 1:1.



_____________________

Please elaborate.

* Working distance is an intimate dance. It has to be just right. Far enough back that my lens isn't inside my set/tent, but not so far back that I can't reach the set from the side of the camera. Point being, I find the 100mm range ideal, I think 180-200 would put me too far away.

This doesn't make any sense. If you're cropping, then why not just frame your shot correctly by getting the correct distance to begin with?

The difference in minimum focus distance, between the original Canon 100mm and the Sigma 180, is about 6".
In other words, what is "not close enough" in the 100mm (causing you to crop closer-in) would be about the perfect distance with the 180mm, requiring no crop at all.

So it sounds like the exact reverse is true: a 180 mm would put you at exactly the right distance to frame your shot properly, no-crop-needed.

Jack

[Bart_van_der_Wolf]

Well I'm asking about the lens/chip combination. I don't know how explore the hardware beyond that. But in chip sample comparisons I tend to see better sharpness/detail/resolving power from non AA chips. But I'd be happy to see examples that demonstrate otherwise. And if you can demonstrate that the 5d2 is as good as it gets, I'd be happier still. 

There is only a small difference in resolution between cameras with and without AA-filters. The difference that is there, is caused by higher micro-detail contrast of the AA-less versions, but that also comes with a higher risk of (color) moiré and other aliasing artifacts which will make sharpening more difficult.

The biggest jump in resolution will come from upgrading the sensor to a higher MP version. That will immediately pull more resolution out of the current lenses. Upgrading the lens quality then will add additional resolution. The sensor resolution boost is the bigger factor in an improved system resolution.

I don't use default sharpening settings, I use, to varying degrees: LR at max amount and zero detail, and sometimes high/max detail and low amount. Photo Ninja. Topaz Focus. Topaz Detail. Smart Sharpen at high and low radii. USM at high and low radii. Which is simply to say, I'm not ignoring the PP.

The aperture you used has a significant effect on the Radius you should use in all sharpening tools, regardless of the camera/lens used, and regardless of the subject you shoot. There is a direct relationship between the two, so with different apertures you should use different radii. The better sharpening tools use deconvolution sharpening for Capture sharpening. Then you can add subject dependent Creative sharpening as a second and separate operation.

I'm not trying to put one piece of hardware over another, or hardware over technique. I'm looking to refine and evolve all of these. I just thought it would be easier to start at the root: the hardware at the capture itself (which this thread was about). Then the PP would come after and be tailored to it.

Yes, that's clearer now.

As said above; Sensor, Lens, post-processing, in that order, will have the largest impact on your results. But of course they all matter for the final result.

Cheers,
Bart

[smahn]

So it sounds like the exact reverse is true: a 180 mm would put you at exactly the right distance to frame your shot properly, no-crop-needed.

You like the 180, perhaps I'll try it. I've been doing this for some years though and all I can say is I'm comfortable at the distance the 100 puts me. Again, though, I may try it.

But I want to be clear about something, I'm not trying to add resolution via megapixels. Most of my deliverables are 2000x2000 px or less. So I'm looking to increase the sharpness at a given image size.

Downsampling is a whole another issue I need to explore. I get caught in a trap where the client wants most images for exploded view web (about 1600x1600 px) and for some shots wants high res for ads. I generally work at about 2,200-2,500 px on the long edge for my master image and down sample to 1600 for jpegs. The problem is the downsample (and any post downsample sharpening) can really mess with the fine details. So I'm really not looking to start with larger MP images. In fact, with larger MP sensors I might crop more.

FWIW, much of the fine detail I'm talking about is tight lines of diamond facets. If a non AA filter would cause those lines to be jagged that would be a problem.

[John Koerner]

You like the 180, perhaps I'll try it. I've been doing this for some years though and all I can say is I'm comfortable at the distance the 100 puts me. Again, though, I may try it.

I've used a 100 mm for years also, but prefer the 180mm (Canon's at first, now Sigma) for a number of reasons already mentioned.

I've been doing this for some years though and all I can say is I'm comfortable at the distance the 100 puts me. Again, though, I may try it.

You've not given the 180mm a chance, so really you're "comfortable" doing what you're doing, only because you've not tried an alternative.

The very fact you admit to cropping means you're not operating at the correct distance. The distance may be comfortable for your hands/convenience, but by virtue of the fact you're cropping this means you're not close enough to maximize a full image; you're throwing away part of what you take.

If you had a 180mm at the same distance you keep your 100mm, then you would be zoomed-in more and (1) would not have to crop and (2) would have a bigger/better image than a cropped equivalent.

The Sigma 180mm is a better lens in every way the word "better" can be applied to a lens (resolving power, contrast, color, bokeh, etc.) ... and its reach would allow you to frame your shot from the same position you use the current old 100, but where you could now utilize the full image, thus allowing you more detail and printing size options than using the lens you now have.

That's my $0.02 ...

Jack

[smahn]

John, using the whole image is not advantageous in my situation. But thanks.

[bjanes]

Now this may be because you are attempting to wring more performance out of the lens than there is, but the postprocessing tools/skills may also need to be improved. Default sharpening settings are notoriously stupid, they do not take the camera settings into consideration, while the aperture and magnification factor usually pretty much dictate the required settings and are therefore relatively simple to set up correctly.

Bart,

I am familiar with your recommendations regarding adjusting the radius in the sharpening tool and have used your slanted edge tool with a couple of my lenses. However, I don't know how to factor in the magnification. The effective aperture does decrease (larger f/number) when one focuses closer and the magnification increases, but with the newer lenses with a built in computer chip, the effective f/number is reported in the exif and is automatically set by the camera. What further needs to be done?

Cheers,

Bill

[John Koerner]

John, using the whole image is not advantageous in my situation. But thanks.

Sure. But a crop is a compromise, never an advantage.

[smahn]

Sure. But a crop is a compromise, never an advantage.

Sure, until FF becomes a disadvantage, which can happen with heavy downsampling.

But lets not argue. You gave some good advice, I hope to have the opportunity to see if it suits my use.

[Bart_van_der_Wolf]

Bart,

I am familiar with your recommendations regarding adjusting the radius in the sharpening tool and have used your slanted edge tool with a couple of my lenses. However, I don't know how to factor in the magnification. The effective aperture does decrease (larger f/number) when one focuses closer and the magnification increases, but with the newer lenses with a built in computer chip, the effective f/number is reported in the exif and is automatically set by the camera. What further needs to be done?

Hi Bill,

At normal focusing distances of say 10 feet or more the differences between nominal and effective aperture are small. And as you say, it may not be immediately clear from the Exif data (unless one knows what it is that is reported) what the effective aperture is. It depends on the specific camera system, but I suppose one could compare infinity focus with close-up focus and Aperture priority metering to see what is recorded.

At close-up focusing of e.g. a 1:4 magnification ratio, the required exposure time increases by approx. a factor of (1+1/4)^2, or with the same exposure time the effective aperture number increases with a factor of (1+1/4).

A more precise figure depends on the lens design, and for the aperture it is:
Effective Aperture = Aperture x (1 + Magnification/pupilFactor) .
The pupil factor is the Exit/Entrance pupil diameter ratio, which can be estimated by looking through the lens from both ends and measuring the apparent diameter of the pupils.

This effective aperture is what determines the diffraction that dominates the differences in required sharpening radius at narrower apertures. I've shown it before, but attached is the measured blur, and required radius to restore sharpness, for my EF 100mm f/2.8L Macro IS lens. It might be that other lenses show similar characteristics and resulting settings, but one can only be sure by measuring it, e.g. with the slanted edge method.

Cheers,
Bart

[smahn]

There is only a small difference in resolution between cameras with and without AA-filters. The difference that is there, is caused by higher micro-detail contrast of the AA-less versions, but that also comes with a higher risk of (color) moiré and other aliasing artifacts which will make sharpening more difficult.

Bart, I would appreciate if you could expand on this. Much of my work is, if there is a such a thing, high end web based. I don't need huge megapixel resolution, I need really sharp and micro detailed smaller/medium size images.

Here's a google image search for pave diamond:

http://tinyurl.com/on9nna6

Imagine one of those bracelets with a thousand diamonds and the client wanted every facet of every diamond distinct. it takes a high degree of resolving power even if it's ultimately not a high res image. (I fear the term resolution is getting confusing. I'm more talking about being able to discern (resolve) the finest lines through the lens and at the chip, not how many megapixels the resulting file yields.)

Now you maae the point that an non AA chip will give better micro contrast at the outset, but that an AA chip may yield a sharper final by allowing more aggressive sharpening. Other than me buying/renting two cameras to test side by side (which I may ultimately do, but which two?) do you have examples or other means to illustrate which might serve me better?

The best comparison tool I've found so far is here (though I've not searched far and wide):

http://www.dpreview.com/previews/nikon-d810/9

And the Nikon d810 just looks better than the competition, though I haven't looked into the lenses and processing used.

[Bart_van_der_Wolf]

This effective aperture is what determines the diffraction that dominates the differences in required sharpening radius at narrower apertures. I've shown it before, but attached is the measured blur, and required radius to restore sharpness, for my EF 100mm f/2.8L Macro IS lens. It might be that other lenses show similar characteristics and resulting settings, but one can only be sure by measuring it, e.g. with the slanted edge method.

One more thing to add.

These specialized lenses can use different methods to achieve the close-up focusing capability, and also other lenses use internal focusing groups that may change the focal length and for that matter the entire optical performance. It then becomes a bit tricky to blindly apply corrected measurements that were made at a vastly different distance/magnification.

The difficulty with calibration of the close-up focusing performance of a lens is that the quality of the test target also becomes much more important. I'll try to do a test today if I get a chance, and determine the blur characteristics in close-up scenario's. For that I'll use a slanted edge made from self-adhesive pvc material (white on black, to reduce the contrast a bit), cut with a sharp knife for a clean sharp edge. Other possibilities are to use the edge of a razor blade, although then there are some issues with reflection.

Cheers,
Bart

[Bart_van_der_Wolf]

Bart, I would appreciate if you could expand on this. Much of my work is, if there is a such a thing, high end web based. I don't need huge megapixel resolution, I need really sharp and micro detailed smaller/medium size images.

Yes, your objective is becoming clearer now. The added number of pixels are not the real reason for the higher resolution though, although that can be achieved by down-sampling (but you seem to have some issues with that, which we can also solve). It also has to do with how dense the sensels are packed, and denser sampling will pull more resolution out of any lens.

Here's a google image search for pave diamond:

http://tinyurl.com/on9nna6

For those I'd assume that DOF restrictions are your main cause for loss of resolution, assuming that the entire object must be in focus. For such objects it is almost unavoidable to do focus stacking at the lens' best aperture. However, in the plane of best focus (which is much thinner than the DOF range), the previous sharpening observations still stand, i.e. aperture determines resolution and therefore the sharpening radius that's required to restore resolution. Focus magic or Topaz InFocus are dedicated deconvolution sharpening tools, and Topaz Detail also offers deconvolution, and other tools that address the different levels of detail.

Imagine one of those bracelets with a thousand diamonds and the client wanted every facet of every diamond distinct. it takes a high degree of resolving power even if it's ultimately not a high res image. (I fear the term resolution is getting confusing. I'm more talking about being able to discern (resolve) the finest lines through the lens and at the chip, not how many megapixels the resulting file yields.)

Here down-sampling can be your friend, not your enemy, so I wouldn't eliminate high MP cameras from your shortlist. Not only will the DOF limited resolution improve up to the limits of pixel size, but also (photon shot- and read-)noise will be reduced which would make sharpening easier because of a higher signal to noise in the image.

Now you make the point that an non AA chip will give better micro contrast at the outset, but that an AA chip may yield a sharper final by allowing more aggressive sharpening. Other than me buying/renting two cameras to test side by side (which I may ultimately do, but which two?) do you have examples or other means to illustrate which might serve me better?

It's hard to find direct comparisons, but it's a basic Digital Signal Processing (DSP) fact of life; discrete sampling introduces aliasing artifacts if there is finer detail than the sampling interval. Now in the case of the pavé seting of diamonds, there is an added complexity that may even help, and that's the DOF requirements. Since it will be hard to get sufficient DOF at the maximum resolution, the defocus will act as a kind of AA-filter. But that still leaves the in-focus edges susceptible for aliasing, so it remains a tricky approach (especially with a Bayer CFA that can lead to false color artifacts caused by aliasing).

The best comparison tool I've found so far is here (though I've not searched far and wide):

http://www.dpreview.com/previews/nikon-d810/9

And the Nikon d810 just looks better than the competition, though I haven't looked into the lenses and processing used.

From the specs the D810 certainly looks capable, and there are probably quite decent close-up or macro/micro lenses available. I have no experience with how well e.g. tethering works with the Nikons. My personal experience is mostly with Canons and they have good tethering capabilities, and lots of tools that support e.g. focus stacking like Helicon Remote. The D810 is apparently also supported by Helicon Remote, so that leaves those options open.

Cheers,
Bart

[NancyP]

This is for studio shooting. The upcoming Canon 50 MP camera might give you the best detail. Canon sensors lag on dynamic resolution, but that's a non-issue for studio work. I can also see that 180mm is overkill - how blurred do you need a white background? Do you currently focus stack? If so, how do you move your lens or sensor?

[ErikKaffehr]

Hi,

As Bart said, with macro work DoF and Diffraction always fight each other. The best way to achieve fine detail may be to use focus stacking.

This very dead fly photographed using a Photar 25/2.5 and it has been stacked from something like 30 exposures:


Full size image: http://echophoto.dnsalias.net/ekr/Articles/Stacking/2015-01-14-22_24_18_ZS_PMax_small.jpg

I guess the Canon 100/2.8LII macro is quite good, but any lens will be killed by diffraction when stopped down to much, and shooting close up normally adds extension so real aperture may be much smaller than the aperture shown on the camera.

The Sigma 150 and 180 mm tele macros are very well corrected for axial chroma, the stuff that yields magenta/green out of focus areas, at least according to Photozone tests.

Best regards
Erik

This is for studio shooting. The upcoming Canon 50 MP camera might give you the best detail. Canon sensors lag on dynamic resolution, but that's a non-issue for studio work. I can also see that 180mm is overkill - how blurred do you need a white background? Do you currently focus stack? If so, how do you move your lens or sensor?

[smahn]

Do you currently focus stack? If so, how do you move your lens or sensor?

Ok, I'm glad you asked. Bart's responses have convinced me I shouldn't really improve my techniques with what I have before I shell out big bucks. And whatever I pick up in the process can be utilized with the better eq too.

So my dirty little secret is that I typically shoot at f16 on a lens that goes to f32, so two stops from max. Working in a production environment, I have to work quickly. For each item, I shoot it from a few angles with lighting tweaks. Then the client chooses which he likes best. It's very cumbersome to focus stack all these "working shots", not knowing which will be the final. If the item is a ring or bracelet, it's often okay (if not attractive) to have the front in focus and the back soft. So at f16, and not using the whole sensor, I can often get the DOF in one pop. So there is, I confess, a degree of diffraction I'm working against.

But for each shoot there's always a couple I do have to stack, in which case I do about 3 focus slices at f16, focusing using the lens, and combining in LR with the plugin Enfuse.

Bart, you are the sharpening maestro, but I have to admit sometimes this sharpening stuff gets a little too mathy for my little brain. I love technique, but in the pursuit of art not science. I'm more like a chef trying a recipe than an immunologist looking for a cure. All of which is to say, are you able to point me to a couple of concise tutorials that would benefit me for overcoming diffraction, and perhaps automating focus stacking? If you (or anyone) can point me to something that would be great. I just want the recipe though without the math and science. If it passes the eye test that's good enough for me.

I'm on a Mac, fwiw.

[Jack Hogan]

This very dead fly...

Ugh, Erik, a dr. Hagen's morbid dream 

[Jack Hogan]

Sensor, Lens, post-processing, in that order, will have the largest impact on your results. But of course they all matter for the final result.

Amen to that.

[ErikKaffehr]

Hi,

What I have seen, mostly, is that there is little difference between decent quality lenses once you stop down to f/8, or so. Aberrations have settled down and diffraction starts to dominate. Now, keep in mind that with close ups, real aperture may be much smaller than marked on the lens, in the old times if the lens was set at f/16 the real aperture would be f/32 at 1:1, now days with floatings element and internal focusing the effect may be much less.

With the good sharpening methods we have at hand, much of the contrast lost to diffraction can be regained. So it may be a good idea to focus on optimising post processing.

Best regards
Erik

Ok, I'm glad you asked. Bart's responses have convinced me I shouldn't really improve my techniques with what I have before I shell out big bucks. And whatever I pick up in the process can be utilized with the better eq too.

So my dirty little secret is that I typically shoot at f16 on a lens that goes to f32, so two stops from max. Working in a production environment, I have to work quickly. For each item, I shoot it from a few angles with lighting tweaks. Then the client chooses which he likes best. It's very cumbersome to focus stack all these "working shots", not knowing which will be the final. If the item is a ring or bracelet, it's often okay (if not attractive) to have the front in focus and the back soft. So at f16, and not using the whole sensor, I can often get the DOF in one pop. So there is, I confess, a degree of diffraction I'm working against.

But for each shoot there's always a couple I do have to stack, in which case I do about 3 focus slices at f16, focusing using the lens, and combining in LR with the plugin Enfuse.

Bart, you are the sharpening maestro, but I have to admit sometimes this sharpening stuff gets a little too mathy for my little brain. I love technique, but in the pursuit of art not science. I'm more like a chef trying a recipe than an immunologist looking for a cure. All of which is to say, are you able to point me to a couple of concise tutorials that would benefit me for overcoming diffraction, and perhaps automating focus stacking? If you (or anyone) can point me to something that would be great. I just want the recipe though without the math and science. If it passes the eye test that's good enough for me.

I'm on a Mac, fwiw.

[Bart_van_der_Wolf]

Ok, I'm glad you asked. Bart's responses have convinced me I shouldn't really improve my techniques with what I have before I shell out big bucks. And whatever I pick up in the process can be utilized with the better eq too.

So my dirty little secret is that I typically shoot at f16 on a lens that goes to f32, so two stops from max. Working in a production environment, I have to work quickly. For each item, I shoot it from a few angles with lighting tweaks. Then the client chooses which he likes best. It's very cumbersome to focus stack all these "working shots", not knowing which will be the final. If the item is a ring or bracelet, it's often okay (if not attractive) to have the front in focus and the back soft. So at f16, and not using the whole sensor, I can often get the DOF in one pop. So there is, I confess, a degree of diffraction I'm working against.

Hi,

I suspected something like that. Diffraction will spoil a lot of the fun, but narrow apertures may be required to get enough DOF in a single shot. If that's the case, then you need a good deconvolution sharpening tool, like e.g. FocusMagic. It works best, like all deconvolution tools, when noise is modest, so make sure the exposure level is high enough that you don't need to push in postprocessing.

But for each shoot there's always a couple I do have to stack, in which case I do about 3 focus slices at f16, focusing using the lens, and combining in LR with the plugin Enfuse.

Bart, you are the sharpening maestro, but I have to admit sometimes this sharpening stuff gets a little too mathy for my little brain. I love technique, but in the pursuit of art not science. I'm more like a chef trying a recipe than an immunologist looking for a cure. All of which is to say, are you able to point me to a couple of concise tutorials that would benefit me for overcoming diffraction, and perhaps automating focus stacking? If you (or anyone) can point me to something that would be great. I just want the recipe though without the math and science. If it passes the eye test that's good enough for me.

Diffraction is a fact of physics/life. the best you can do is avoid it by using wider apertures and/or by using something like FocusMagic to restore some of the seemingly lost resolution. Topaz InFocus can also restore resolution, but it is more finicky and very quickly produces artifacts when the settings are sub-optimal. FocusMagic is a bit more robust for this restoration stuff.

I understand the workspeed requirements, although they will work against getting the best quality. It also means that you'll have to automate as much as possible. Helicon Focus with Helicon Remote can help with the stacking, 'Remote' can calculate and automatically take the required number of shots when you set the front and rear positions that need to be in focus, and then Helicon Focus will stitch the results, if necessary from Raws.

Your hesitation towards down-sampling can be solved if you e.g. use Lightroom which does a decent job . The downsampling will also allow to shoot with more detail where the subject is in focus, and it will 'sharpen' the out-of-focus parts a bit, which were caused by the DOF limitations.

Cheers,
Bart