February 27, 2015, 10:45:10 AM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - torger

Pages: 1 ... 13 14 [15] 16
EOS Bodies / Re: What do you want from the 5D mk III
« on: April 15, 2011, 03:09:52 PM »
Well, actually I did mean that EF-S glass is poor :-). I did not however mean that EF-S is not good enough for 8 megapixels. Much of the glass show resolution limits on the 18 megapixel 7D though, such that if you put in a lens with known high resolving power you see a clear improvement.

EF-S glass is not poor because there's some law of physics that makes it so (oh well the smaller image circle and the need for shorter focal lengths may make it a bit more difficult), but because in general it is expensive to manufacture high quality optics, and EF-S is made for APS-C, and APS-C is in the low cost segment so the lenses should be cheap too. Of course, some of the EF-S lenses are not exactly cheap, such as the 17-55mm f/2.8, but still cheaper than the corresponding full-frame glass.

If only looking at resolving power there are indeed low cost lenses that are very sharp, such as the 50mm. Some focal lengths are easier to make sharp than others, just because you there's a cheap 50mm it does not mean you can make a sharp cheap 24mm (it's about the distance to the image plane and other factors).

With the 50mm, the cheap ones are actually sharper than the L version on small apertures, so they are great landscape lenses. There are a fair amount of less-than-sharp L lenses, but those are not so much optimized for maximum resolution at f/5.6-f/8 but some other aspect such as ok sharpness and very nice bokeh at largest apertures. Afterall, really high resolving power is somewhat of a niche inhabitated by landscape photographers and others that don't shoot a single photograph without a tripod and a remote shutter release :-).

The EF-S lenses are all zooms (except the macro 60mm), and as far as I know none of them is as sharp as the good old 24-70 f/2.8 L zoom even at f/8.

What you do if you want to get full use of the 7D's 18 megapixels is that you use some of the high resolution full-frame lenses. The larger image circle gives you the advantage of less vignetting and better corner performance. I'm quite sure that the "pro" APS-C models are intended to be used together with pro full-frame glass, so therefore it is ok with 18 megapixels.

However, 18 megapixels on an entry-level camera where the users will most likely use EF-S zooms is somewhat overkill, but probably unavoidable for marketing reasons -- resolution-as-a-number probably sells in the low end.

If an EF-S lenses, including those that cost >U.S.$1,000, can't resolve 8MP (40% at the center of a 5Dmk2's sensor), it certainly can't resolve neither 10MP on an EOS 1000D nor 18MP on an EOS 7D. If this is true, it sounds bad for the Canon brand - either Canon spends too much on megapixels or it doesn't spend enough on low end glass.

Even the worst lenses can resolve without any problem any current SLR cameras and can resolve even higher detail with a new higher mp sensor. There are some myths out there which unfortunately confuse people which could lead them take wrong decisions on their purchase.

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 15, 2011, 05:04:01 AM »
On the megapixel issue I think many people debate from the viewpoint of low end cameras. The megapixel issue in compacts and APS-C is not the same as in fullframe and medium format.

Even the good EF-S glass has fairly low resolving power - the optics is designed for good price/performance. For fullframe, you have professional glass with considerably better resolving power, and you have ~2.5 times larger sensor area. Getting the same pixel pitch as a 7D on fullframe would mean 45 megapixels.

Considering that EF-S glass generally has low resolving power and the pixel pitch is much smaller, the megapixel count has been pushed much farther on APS-C than on fullframe. Compact cameras even more so.

Fullframe is still rather conservative on the megapixel count. The reason why we have not yet seen 40+ megapixels on fullframe is probably more due to issues handling the file sizes (tough requirements on processing and storage) rather than limits in sensor technology. There's been some limits in glass too, but some important upgrades have been made.

If a photographer doesn't need high end resolution, there's APS-C. I see little reason to make a fullframe camera that does not aim at maximizing performance with high end glass, since fullframe sensors are due to their size a lot more expensive to manufacture than APS-C, and you could satisfy lower resolution requirements with APS-C.

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 13, 2011, 12:48:13 PM »
Here is why you see the difference: the 1D4 sensor has a 1.3 crop, which means that its surface is 1.69 times smaller than that of D3s, which means that it gathers that much less light, which means that you should see more noise worth about log2( 1.69) = 0.75 stops per photo.

Ahh... did not think about sensor size, you're right that does have some impact. I'll look more into this in the future, really interesting for us technical nerds :-). But wait a minute, that would only count for photon shot noise, right?

Emil Martinec's noise discussion sheds some light on the issue:


Just as you say he says that light collection efficiency is largly independent of pixel size, but that there can be read noise advantages of larger pixels at high ISOs.

EOS Bodies / Re: DIGIC V [CR1]
« on: April 13, 2011, 11:13:35 AM »
I was wondering (for a long time now)
Introducing Digic V, is Canon considering going away from the jpg file format and introduce a newer, more improved format like png?

PNG is not designed for photographic images. "Next generation" format would probably be JPEG2000 or something, but it is computationally intensive and there's poor software support etc. I think JPEG is still the least bad alternative to RAW for some years to come. Most professionals use RAW anyway. It is unfortunate that Canon, Nikon etc refuses to make their RAW formats public though. In practice they are thanks to reverse engineering in dcraw, but it would be a nice gesture from the manufacturers to actually publish the specs.

EOS Bodies / Re: DIGIC V [CR1]
« on: April 13, 2011, 11:08:50 AM »

<li>Dynamic range improvements, unknown to what extent</li>

I tought DR was more a sensor thing? What's the link between a new processor and that?

My guess is that it has something to do with the RAW->JPEG conversion, perhaps some tone-mapping feature. That is no real DR improvement, but some more or less refined way to tone-map more range from the RAW data into the JPEG. Even if so it is good news since it probably means that the sensors is expected to have higher dynamic range than the currently available, and thus creating a need for automatic tone-mapping for those photographers that don't shoot RAW.

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 13, 2011, 08:25:40 AM »
I guess you are referring to this comparison:


I see clear differences to the D3s's advantage, I'd say it is about 2/3 stops in that comparison, the reviewer sees 1. D3s ISO12800 looks only slightly worse than 1D4 ISO6400. If you say that it is no difference, I understand your conclusions, but I cannot draw the same since I think the D3s shows clearly better performance.

Afterall, it's all subjective. I'm not saying you're wrong, it could just as well be me that is more picky than the average user. It is a very interesting discussion.

(There's one thing that testers often miss, I'm not sure if done here, the light meters can make different decisions, typcially low contrast scenes are underexposed if auto-exposed, so I prefer tests when there is manual exposure according to the expose-to-the-right principle so I get to see the raw sensor performance. Light meter behavior is of course relevant to action photography though, but much harder to test... anyway, when shooting from a tripod and you have the time one should always expose manually, it is not uncommon that the camera's auto-exposure algorithm throws away 1-2 stops in dynamic range just like that, at least according to my experience.)

yet there is no visible difference in the noise levels (between the two) at ISO 12800.

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 13, 2011, 06:46:57 AM »
You cannot be categorical on this issue. Yes, the noise advantage of large pixels/sensels is generally over-estimated, but saying it is not a factor to take into account is going too far.

The noise advantage of a larger pixel is that you gather more light (more signal), and if the read noise is the same, you get a higher signal/noise ratio (you get some advantage of less photon shot noise too with more signal). So if you make a smaller pixel you need also to lower the read noise to keep the same signal/noise ratio per pixel. It is also easier to gather the photons in a specific area with one pixel than if you have pixel borders in it - photons can hit in-between pixels (micro lenses developments has improved this collection problem though).

Whether or not the advantage of large pixels will compensate or supersede the advantage of averaging noise between several pixels you need to subjectively compare on a case per case basis. If the small pixel sensor has great electronics and great micro lenses, and the large pixel sensor has poor the small pixel sensor will certainly win. If the quality of the electronics and sensor construction is exactly the same, which sensor will win? I don't know, I'm not sure if anyone knows, there's too many variables.

The high ISO advantage of the D3s I think is due to both larger pixel size and quality electronics. However, if you make a camera body specifically aimed at high ISO you don't really need many megapixels - high ISO typically means hand-held and action and then it is generally not possible to get the extreme sharpness you need to max out 20+ megapixels, plus the noise levels will be so high in any case that the pixels are not carrying useful info at the pixel level. So you have more than one reason that makes a high ISO sensor favor fewer megapixels.

But perhaps I'm wrong not believing in combining top performance in high ISO and high pixel count. I would be glad if I am, because I do favor high pixel count since my main interest is landscape photography...

The D3S is currently the best ISO handling DSLR out there.

Yes, but that's because of its sensor size (and its top technology), not because of the fewer pixels.

The D3S is currently the best ISO handling DSLR out there. But you are saying that even if it had 30mp, that wouldn't affect its performance and the image at higher ISO's would infact be better?

Exactly. (But only in theory because Nikon / Sony may not have the actual sensor technology which would allow them to increase the pixel density while maintaining the level of noise. As the comparison between the 1D4 and D3s shows, Canon does have such technology.)

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 12, 2011, 05:03:01 PM »
Both images are underexposed, the D3s 0.5 stops more than the D3X, so the D3S is less favoured.

When you shrink an image, the pixels are averaged out which reduces impact of the noise. So in some circumstances a high MP sensor can with more noise per pixel can win over a lower MP sensor. On my screen the D3s wins with a little though, but not much.

When I discussed quality above I was discussing "high end" prints and limits of human vision. A bit early to dream about that perhaps, I admit. On normal viewing distance I see individual pixels on my computer screen, which is 1920x1200 @ 100 ppi. That is not satisfactory, but is what current technology can do. When the screen has 300+ ppi (like the iPhone4 has, but that screen is a bit small for a workstation :-) ) the quality is approaching the vision limit (for photos, not for line art). For the best prints, we're already there. For the screens I think we will be there within 5-8 years or so. I want to produce "future proof" photos as soon as possible...

I don't really believe in compromise of high MP and high ISO, I really like Nikon's approach with D3x (high MP) and D3s (high ISO)

Here is a little trick that anyone can do.

EOS Bodies / Re: What do you want from the 5D mk III
« on: April 12, 2011, 11:17:58 AM »
I guess the 5D must be some sort of all-around camera. I don't really believe in compromise of high MP and high ISO, I really like Nikon's approach with D3x (high MP) and D3s (high ISO), but for the non-flagship model I guess you need that compromise, and I guess the rumoured 28 megapixel could provide both an upgrade in MP and ISO performance.

In general I think the value of resolution is a bit underestimated, at least for tripod-mounted photography. 400 ppi for book/screen distance viewing distance is a good target, and this is what you can get with high quality printing technology. My personal experience is that there is a clear visual difference between 200 and 400 ppi in printed material. You don't really see individual pixels at 200 ppi, but the lack of micro detail leads to an unnatural pastel-like look, especially highlights are suffering. At 400 ppi at book viewing distance the image looks perfectly natural, and you get the sense that the image has resolution beyond what the eye can perceive, which is what you should strive for. Being satisfied with 200 ppi images is like being satisfied with 8 bit sound.

Also worth noting is that film images have a much more pleasing look if blown up too large than a digital image, so I would say that the resolution requirement is higher for digital than film. There's a charm to film artifacts, but there's no charm in digital ones. The ideas of what resolution is required for a certain size/viewing distance come from the film era, and it certainly needs upgrading to fit the digital era.

You cannot get high ppi count on computer screens yet (they are currently around 100 ppi), but it will come. To fill a 24 inch screen with 300 ppi or both pages of a high quality photo book you'll need around 40 megapixels, and I think that is a reasonable resolution for full-frame, but you will probably have to sacrifice ISO performance then, so every photographer not using a tripod and often shooting at high ISOs would of course want to trade megapixels for better ISO performance.

Of course you could say that high MP count should be reserved for medium format, but I think 40 - 50 megapixels is sort of ideal pixel count for unlimited time into the future (its related to the human eye limitations and normal picture sizes), and since that pixel count is technically possible with good quality for 35mm I think we should eventually get there. At least with some camera body... entry level full-frame might not be the right one.

EOS Bodies / Re: Resolving Power on the 5D Mk II / III
« on: March 29, 2011, 04:01:15 AM »
When discussing resolution limits in a camera system (which we landscape photographers like :-) ) you should *not* see it as a chain where the resolution cannot be higher than the weakest link allows. Instead, the total resolving power is the weighted sum of all parts in the system.

Like it or not, the science behind MTF says that the final MTF is the product of the MTF of all the components.  Since MTF is always less than one, the product is always smaller than the weakest link.

I've never heard that MTF is the sum of all the parts.  MTF is basically resolving power.


Sure. The saying "weighted sum of all parts in the system" is not scientific, it is just a way to think which relate better to what you see in practice than the thinking "not sharper than the weakest link" which is more common.

MTF is measured or probably more often estimated by theoretical calculations. Airy disc diameter is based on some subjective criterion on where the response is "low enough" compared to the peak. And on it goes -- that is no limit is on/off sharp, what you see as you stress the system with finer and finer detail is gradual reduction of contrast.

When you chain together several systems all which have soft limits, soft limits that also behave differently and are differently defined, it is hard to predict the end result. You cannot say that diffraction airy disc diameter defines a resolution limit the same way a sensor sensel diameter does. Still I see people make those assumptions all the time, that is if the airy disc is 6 microns, then 4 micron sensels give no better fine detail contrast than 6 micron sensels, which is not true since those both limits are soft and behave differently.

Since limits are soft (gradual reduction in contrast, where some subjective definition puts a "limit" at some place along that gradual reduction), you most often see improvements in the end result if you improve one part in the system, even if another part is past its "limit". Of course, if a part in the system is very far past its limit then you will see no gain, but you cannot really theoretically calculate exactly when "no visible gain" happens, you need to test in practice and see what happens.

EOS Bodies / Re: Resolving Power on the 5D Mk II / III
« on: March 28, 2011, 08:32:43 AM »
When discussing resolution limits in a camera system (which we landscape photographers like :-) ) you should *not* see it as a chain where the resolution cannot be higher than the weakest link allows. Instead, the total resolving power is the weighted sum of all parts in the system.

The reason for this is that most limitations are soft limits, take diffraction for example - it is not 100% contrast, reaching airy disc diameter and then 0% contrast. Instead a limit leads to graceful fine detail contrast degradation. This means that if you improve the resolving power of one component in the system, you will in practice most likely see some improvement even though you have "passed the limit" on some other component/factor. Of course, it will pay off less to improve a strong link than a weak, but sometimes it can be worth it anyway.

A typical example when it can be useful is outresolving the diffraction since it really is a soft onset limit.

When it comes to bayer array sensor you have the whole anti-aliasing filter and demosaicing factor to take into account, making the actual resolution limit of the sensor itself fuzzy. Pixel peep a foveon sensor (no AA filter, RGB for each pixel) and compare to a unsharpened output from a bayer sensor in Canon or Nikon and you'll see clearly that you don't get 100% contrast down to the pixel level.

(Actually, I tend to think that it can be an advantage to have slight contrast loss at the pixel level that bayer/aa has, this makes jaggies and other pixel artifacts less likely to become visible in a cropped image, even after sharpening has been applied)

Canon General / Re: Canon Officially Acknowledges Lens ID Issue
« on: February 25, 2011, 01:27:24 AM »
Sigh... Canon (and all other camera manufacturers) really like to treat customers like crap. They keep their RAW formats secret so they have to be reverse engineered (with potential risk to get sued), or even worse encrypt them (Nikon). And they more or less deliberately mess up camera-lens communication to make things hard for third-party lens manufacturers. Oh, camera-computer communication is a secret too. In all, it's a scandal (the RAW format thing is the worst part though if you ask me), they act like coward companies that gives a damn in their customers' interests as long as they sell.

Still, there are lots of more or less fanatic fans out there... I don't understand it, Canon and Nikon certainly don't deserve it.

But with film there is a quality that will never be produced from digital, it is so natural, the images of a good medium format camera feels like you could fall into them, they are beautiful.

I think many photographers that shoot film do so to stand out, to get a touch of uniqueness in their work. Large format film is a lot more romantic than 35mm digital that anyone can easily handle (technically). 35mm digital is crowded to say the least. However, I personally get a bit suspicious when a photographer makes a big thing out of the format he/she uses. In a way I think it's more "honest" to just go for the format that gives the best possible technical quality, and make pictures with that. A "unique look" if any is then a product of the content rather than the production process.

Here is a Canon white paper about manufacturing sensors.

I've always wondered how large part of the cost of the camera that is the sensor. Does anyone know? It seems to me that if you buy a 5D a much larger part of the money goes to finance the sensor than if you buy a 7D.

The white paper says that a full-frame sensor can be 20 times more expensive than APS-C. The 5D is about $2500 and the 7D $1500. Is the 5D sensor as much as $1000, or even more?

I guess the square sensor would be 31x31 mm to fit within the image circle of 44mm (about 10% more area than 24x36), not 33x33 as suggested in the initial post. I guess there would also be some technical difficulties to fit the mirror in there.

I guess you also could argue that 24x36 actually is better use of the image circle than 31x31. It certainly will be if your compositions most often is in 3:2 format. If I would use square format I think I would crop them to some rectangular format most of the time, but perhaps that's only because I'm not used to square composition?

Lenses / Re: Your lenses wishlist for 2011.
« on: February 13, 2011, 04:01:00 AM »
As a landscape photographer I'm kind of obsessed with sharpness. Most of the time there are of course other factors limiting sharpness but I don't like the lens to be too much of a limiting factor when the other conditions are good.

I'd like to see a 24-70/2,8 IS. The current non-IS is really sharp, but with IS it would be a better all-around lens. As it seems we will get a new 24-70, but perhaps without IS and I don't think it is likely that a new lens can be considerably sharper than the old... we'll see.

Then I'd like to see upgrade of the lower cost 35mm and 24mm, those lenses are not really that sharp. I think primes should outperform zooms in sharpness, but the 24-70 is actually somewhat sharper than those at f/8. The L-lenses at 35 and 24 are slightly sharper than the zoom, but really expensive and somewhat heavy - not a great buy (for an amateur that doesn't need the superior build quality) if you intend to shoot at around f/8 most of the time.

The tilt-shift TS-E 24mm II is a great lens, if you're serious about landscapes and have the money - it is one to own. The old 90mm TS-E performs really well too, but the 45mm is not performing so well on a modern high megapixel body, so a new TS-E 45mm would be great. Actually I think TS-E at 35mm would be more useful than 45mm though.

Pages: 1 ... 13 14 [15] 16