Big Pixels and Lens Quality

[xpatUSA]

We often discuss the subject of "matching" lenses to sensors, often resulting in quite fierce debate with copious mentions of the Real World ;-)

Some time back, I bought a camera noted for taking sharp pictures. Naively, I also bought the worlds best, i.e. sharpest, 70mm true macro lens for my table-top photography. Later, I bought a 17-70mm zoom lens in case I ever ventured out into the said Real World.

Lately, I've been wondering just how cost-effective the macro lens is? The point being that, with large pixels, you can almost stick a pin-hole in front of the sensor and still take good pics, so the theory goes. After much intensive research and thought, I concluded that my zoom should be just as "good" as the relatively expensive macro lens, at least for Real World shots. And here we are:



Perhaps unfair, the macro lens being designed for close-up work, flat field, etc. but the basic point is made by the [100% crop] result above, eh?

This post might be of interest to D3/D700 drivers, whose pixels are only slightly smaller than mine . . .

[chauncey]

I have difficulty in believing that a 17-70 zoom lens would hold its own against a dedicated macro lens.
My own 180 macro, mounted on my Canon 1Ds3, can out-resolve any other white lens that I own,
with the possible exception of my 300mm 2.8...even that is a close call.

In your example, it appears that both images are somewhat soft...are they AF or MF, might explain the difference.

[rpcrowe]

It is often quite difficult for me to ascertain sharpness of images posted on the web. I cannot really tell the difference between the two images regarding sharpness. The images both look a little soft on my monitor but, that doesn't really mean anything...

However, that is why I will often say that the photographer's choice of camera and lens can be predicated on the ultimate use of an image. IMO, images destined for posting on the web, on social networks or for sending by email do not necessarily need to be shot with the top-most cameras or lenses...

[RustBeltRaw]

I agree that sharpness comparisons should not be made on anything but unprocessed full-res versions. While we're at it, were these photos taken at the same aperture, ISO, and shutter speed? Was the camera tripod-mounted?

As a final point, a "real-world" application for the macro lens is focusing extremely close for higher magnification. How does the 17-70mm hold up then? I realize Sigma calls it a Macro, but they define that as any glass with 1:3 magnification or above. Not exactly ideal, to my mind.

"Matching" the lens to the sensor implies that the lens is sending pixel-sized chunks of light through its rear element. A poor sensor will not be worse with higher-quality glass, and a high-quality sensor is likely to reveal the weaknesses in lower-end glass.

[rpcrowe]

Lex,

You are right about the various companies (such as Sigma) using the term "macro" as a sales incentive. The rational behind this is that after being enlarged to a size such as 4" x 6" (standard discount store size print), the image is then life size.

You are also correct about the true macro lenses providing their optimum performance in close-up and macro photography...

Even my ancient Tamron 90mm f/2.8 SF SD Macro lens (which I purchased almost ten years ago for $100 USD on eBay) provides excellent imagery in close-up and macro photography and also works quite well with an extension tube added to increase the image ratio over 1:1...

[xpatUSA]

Sorry Gents,

The originals are long gone - it was a quick hand-held test. The images I posted are full-res 100% crops, however, with no post-processing (I should have made that clear). Same f-number, same ISO and same shutter speed IIRC. AF/MF "I don't remember", as Reagan used to say.

. . mounted on my Canon 1Ds3 . .

The post is about big pixels. The pixel pitch of the 1Ds3 is what, exactly?

Might repeat the test on my wobbly 'National Geographic' tripod and also indoors just for grins.

So, no comments on the use of less-than-perfect lenses with big-pixel cameras, chaps?

Ted

[chauncey]

The post is about big pixels. The pixel pitch of the 1Ds3 is what, exactly?

Hell, I don't even know what pixel pitch means but...it does do this, 800 px crop


.

[xpatUSA]

Might repeat the test on my wobbly 'National Geographic' tripod and also indoors just for grins.

Outside test on tripod went fairly well apart from variable lighting due to clouds. Same 10" x 8" target at about 10 yards, same softness to the images. Used 10 sec delay, hands off because the tripod sucks. f/8 and 1/250 sec for the macro and 1/500 sec for the zoom. Since it is so hard to compare images on the web, how about showing the edge spread instead?

Zoom lens:



Macro lens:



The rise times are not that much different, considering that the zoom was at full extension. And that was the point of the post . . not how soft my images are but how alike the lenses are on a camera with big pixels.

My close-up test failed: it showed the zoom lens to be far superior to the macro . . duh . . I had inadvertently moved the zoom to about 35mm while adjusting the focus . . 73, diabetic, incipient Parkinson's, what can I say?

P.S. Great bird shot, Chauncey. Any post-processing? Does your RAW converter apply any default sharpening?

BTW, the pixel pitch on your APS-H sensor is 7.2um which is quite big by modern standards; ergo, you could have bought much cheaper lenses . . .

Ted

[jcuknz]

As I live in the real world ... well I think I do ...am ... I have always been impressed by how my original 3.3Mp Canon P&S produced great images .... later dpreview educated me to the factor found by dividing the pixels into sensor size.
Here the DSLR always comes out with a single whole number and perhaps some decimal places.... but the P&S/bridge is often up in the mid thirties to mid fourties and the good intermediate cameras such as Canon's G fixed lens models was/are in the mid twenties like my Panasonic FZ of the time.

I then worked out the figure for my Canon P&S and it too was in the single figures like a DSLR.

Don't know why but dpreview gave up the practice ... I suspect it was too revealing

On the subject of macro lenses ... I don't have one and probably never will because while without doubt they are great for their flat field if you shoot stamps and coins et.al. I normally find that the edges are of little importance in the big close-ups I take and a long lens with a moderate close-up lens does all I need ... though it does need to be a long lens as I moved from the 430mm of my bridge camera to the 280 of MFT ... now I need a 4 dioptre rather than my trusted 2 dioptre of years gone by. It is a pointless exercise with a short lens.

[xpatUSA]

... later dpreview educated me to the factor found by dividing the pixels into sensor size.

Interesting. What were the units used for "pixels" and "sensor size"?

Genuinely interested, not extracting the Michael.

[jcuknz]

Sorry Ted but I have forgotten the details but I think you divide the pixels [ 3.3Mp etc ] into the sensor size [ area/mm ] ... or maybe the other way around but the DSLR comes out to under ten while the good smaller camera is around 25 and the rest 35<45. In comparing cameras it does ignore developments in sensors so that is perhaps another reason it was dropped.

Just checked my original s20 which has a sensor 7.144 mm x5.358 mm = 35.10 / 3.3Mp = 10.64 [ so my memory was a bit off.]
But when I double check a couple of DSLRs of the period the numbers just do not match my memory so I cannot really be sure of how it was calculated. I remember that a full frame worked out as under 5 and APS-C were between 6 & 8.

Sorry

[RustBeltRaw]

A good approximation of pixel pitch is dividing the sensor's width (landscape orientation) by its horizontal resolution, and its height by its vertical resolution. The numbers should come out fairly close, and you can average them for a good approximation. For instance, my Canon 60D.

Sensor: 22.3mm x 14.9mm
Resolution: 5184px x 3456px

Width pitch = 22.3mm / 5184px = 4.30µm
Height pitch = 14.9mm / 3456px = 4.31µm

Note that the pixel pitch will be larger than the actual outside dimensions of a given photosite (the sensor for a single pixel), thanks to some stuff that needs to fit between pixels. Micro-reflectors, circuitry, etc. The biggest advantage of a big pixel is a very high signal-to-noise ratio, making it less noise-sensitive. That's one of the major reasons for full-frame cameras.

"But, ho," you say, "then why aren't medium-format digital cameras superb high-ISO performers?"

They could be. Their signal-to-noise ratios are awesome. But medium format manufacturers are designing their cameras for studio environments, where high ISO performance is less critical than detail reproduction. So they focus on the sensor, not the amplification circuits and noise-reduction algorithms. This fact makes me very interested in rumors regarding a Canon medium-format system, which will probably cost as much as a fully-optioned compact car, but might well be a world-beater and bring medium format into the world of action and low-light work.

The lens/sensor resolution argument leads one to believe that a Nokia Lumia 1020, with its ridiculous 41MP sensor (1.12µm pixel pitch ), is the perfect match for a Canon 85mm f1.2L, and other stuff-of-legend, crazy-sharp lenses. But sensor technology is a lot more complicated than one number. At the end of the day, good glass never hurts.

[chauncey]

I'm not sure I'm getting the point of the post unless your trying to suggest that some bodies cannot take advantage of high quality glass, I would not argue that point as it has been proven ad nauseam. Any semi-serious photographer is capable of running searches concerning "do lenses out resolve sensors". The point where quality corners are cut because of financial considerations is a personal decision and to suggest otherwise could be construed as being rather bigoted.

The "need/want" discussions is another one of those ad nauseam threads that is often raised.
I generally buy my gear based solely on what it "could do" rather that what it "can do". Camera bodies are selected by sensor size/quality and glass is selected based on MTF characteristics. I want the viewing distance of my prints to be governed solely by the length of one's nose.

The year was 2008 when I decided to get more heavily involved in photography and plunked down an obscene 8 grand on my 1Ds3. The "you don't deserve that camera"/"you aren't good enough"/"you don't need that camera" outpourings boggled the mind. My basic response was that I was once young and fiscally embarrassed, as they are, but I've aged and done well and it was none of their GD business how I spent my money.

[Manfred M]

One way I have always looked at large versus small sensors is the parallel from the film day. A low ISO film had a fine and tight grain (small sensor pitch) while a fast film had large grain (large sensor pitch). In theory, a well lit scene shot with a high quality lens should have produced a superior result and you should have been able to get away with a lower quality lens on faster film.

In real life, a mediocre lens gave poor results on both fine grain film as well as coarse grain film. I find the same is true in the digital world.

[Brev00]

I am confused about one thing. Well, more than one, but I will let all the math stuff slide. The one thing: are these supposed to be real world shots? If so, your world looks a lot like a Lenstip office space. I have both the 70 and the 17-70. Using them on a D90. I suppose my pixels are much smaller than yours. Always been my problem. But, on that body, the 70 is sharper than the 17-70 starting even at 2.8. These pics look pretty dark to me which might account for performance issues. Plus, you say the shots were hand held. Tsk. Tsk. Go outside and shoot a tree in good light if not the notorious brick wall. Then, if hand held, at least you could do justice to the lenses with some nice shutter speeds and a mix of f stops. I am pretty certain from my experience that the 70 will crush the 17-70 on a D3 especially since the 17-70 is a dx lens and will put the body in dx mode.

[xpatUSA]

Larry, your post is verging on rude but I will assume that the condescending tone is not deliberate.

I am confused about one thing. Well, more than one, but I will let all the math stuff slide. The one thing: are these supposed to be real world shots? If so, your world looks a lot like a Lenstip office space.

I am sick and tired of people playing the "real world" card as if it changes the laws of physics. If I choose to paste a couple of targets to my well-house window that is what you get.

I have both the 70 and the 17-70. Using them on a D90. I suppose my pixels are much smaller than yours. Always been my problem. But, on that body, the 70 is sharper than the 17-70 starting even at 2.8.

I solved that problem by returning a D90 to B&H within a week of ownership :-) Pixels were way too small ;-)

These pics look pretty dark to me which might account for performance issues. Plus, you say the shots were hand held. Tsk. Tsk. Go outside and shoot a tree in good light if not the notorious brick wall. Then, if hand held, at least you could do justice to the lenses with some nice shutter speeds and a mix of f stops.

Looks like you missed my post where I went back outside and took shots on a tripod. The earlier images look dull because I use QuickMTF to compare stuff (it works better with a low-ish contrast), as opposed to your suggestion of using the notoriously inaccurate human eye to view a nature shot on my cheap monitor screen. (Classic photographical obfuscation by introducing as many variables as possible into a discussion about a single subject. BTW, you forgot to throw in 'ISO')

I am pretty certain from my experience that the 70 will crush the 17-70 on a D3 especially since the 17-70 is a dx lens and will put the body in dx mode.

Could you explain the relevance of this comment to our discussion? I mentioned the D3 earlier only because it has big pixels, not expecting that someone would put a 'DX' lens on.

My post was not about these two lenses. It was about lens quality as a general property as it relates to pixel size. The lenses I used were the only ones I had to make a quick test of my thought. Yes folks, I've only got two lens for the Sigma DSLR and those are it.

Today's task is to make a "real world" close-up comparison - just for completeness as someone suggested earlier. I'm thinking of a "real world" Ingersoll stop-watch shot in LO res. mode, which, on my SD9, bins pixels on-sensor to a whopping 18.24um (0.8MP 'resolution', if you must).

I apologize to members for not making the purpose of my post more clear.

Still waiting for an answer from Chauncey to my questions about the bird shot.

[xpatUSA]

Today's task is to make a "real world" close-up comparison - just for completeness as someone suggested earlier. I'm thinking of a "real world" Ingersoll stop-watch shot in LO res. mode, which, on my SD9, bins pixels on-sensor to a whopping 18.24um (0.8MP 'resolution', if you must).

Here we go, no graphs, just real pictures of type PNG to avoid any JPEG mentions ;-)

17-70mm zoom:




70mm macro:



Camera on Giottos tripod with ball-head. Shots taken in mirror-up mode, LO res, f/8. Manual focus - best of four shots selected from each lens. Neutral sharpness, images brightness and whiteness adjusted to be similar, 100% crop - not re-sized.

P.S. I've got a bright shiny Sigma 50mm EX DG f/2.8 macro lens waiting at the Post Office as we speak.

[Brev00]

Not my intention at all. Really confused. I thought you were about to show real world examples so your actual examples threw me for a loop. I did not know your dark, low contrasty images were that way on purpose. Again, that was a red flag for me as a person with only the images provided on my monitor. Some of what I said was an attempt at lightness that was obviously not successful.

[xpatUSA]

No problem, Larry,

CiC is one of the more gentlemanly communities and we all get on fairly well most of the time. I myself have been known to a bit 'sarcy' and, upon occasion, slightly pedantic . But even we pale into insignificance compared to some posters elsewhere (dpReview springs to mind).

Gotta go, just picked up the 50mm and eager to see how it performs.

[edit] As this thread sinks slowly out of sight, I'll just report that the 50mm f/2.8 macro performed no "better" than either the zoom or the 70mm. I'll also report that the GH1 with a 45mm Leica Macro-Elmarit produced a slightly better image (SOOC jpeg) than any of the above [/edit]

Thanks to all for their input.