Prime lenses with DSLR a waste of time?

Discussion in 'Mirrorless Digital Cameras' started by martin_blais, Dec 10, 2005.

  1. Hi

    I have been having an argument with a friend of mine
    about using prime lenses with digital SLR cameras and
    I'd like to know what the "pros" on this forum think.

    Here is the situation: for a number of years I have
    been doing photography with zoom lenses, I use a
    Nikon D70 and a Sigma 18-125mm (equiv. 27mm-188mm,
    f3.5-5.5) and noticed that many of the shots on
    photo.net have much greater clarity that I've ever
    been able to capture with my camera/lens combination
    (basically, I've come to think I'm using a relatively
    crappy lens). After much reading and speaking with
    film photographers, I have become convinced to buy
    some prime lenses and experiment.

    Essentially, my friend claims that using prime lenses
    with a digital camera won't help, for the following
    technical reasons (note that he also has little
    experience with film photography, but he's using
    reasoning from signal processing and I must admit
    that DSLR do bring a significant number of
    differences in the process, so he might be right),
    that I would do just as well with a faster zoom lens
    (such as the expensive new AF-S 17-55m/f2.8 Nikkor
    lens):

    (Quoting him)

    - "On digital cameras, a lens of too high quality
    introduces aliasing artifacts. Consider that the
    sensor is sampling incoming light. The Nyquist
    theorem states that if a signal is sampled below
    twice its highest frequency, then aliasing can
    occur. The lens is filter for the signal which is
    incoming light. If the lens is too good, then the
    signal that passes through it may have more
    frequency than the sensor can capture. A slightly
    lower quality lens reduces the frequency of
    incoming like and therefore the sensor has more
    chances of sampling it without introducing
    aliasing. Note here the term frequency refers to
    the spatial domain, not the wavelength domain which
    affects our perception of colors."

    "The CCD is performing sampling. Because it has
    finite resolution it can introduce aliasing. One
    proof of the existence of the problem is to point
    out that most SLR manufactuers (specially
    full-frame) actually put an antialias filter in
    front of the CCD. That filter serves to blur the
    incomming light in order to prevent the CCD from
    producing an aliased signal. Because of such a
    filter, optics of too high quality will neither
    produce aliasing nor give you any more detail than
    sufficiently good optics. Obviously, sufficiently
    good optics are hard to define and will depend on
    the sensor characteristics."

    To this, I countered that without a perfectly
    matching filter size the bad lens might be blurring
    excessively, and thus not provide the optimal
    resolution on the CCD. Besides, there is something
    that just seems dead wrong to me about using a crappy
    lens as a filter hoping to get better images. Also,
    if the CCD does have some kind of AA filter on it, it
    should take care of the higher-frequencies reaching
    it (if that's indeed the case).

    Also, I indicated to him that I observe higher
    contrast in photos from others taken with prime
    lenses, to which he adds:

    - "any blurring of incoming light would cause loss in
    contrast between adjacent pixels only if the
    scattering of light were to be larger than a
    photosite. Otherwise, I do not see why adjacent
    contrast would be reduced. Now we know that a prime
    lens is made of much less pieces of glass than a
    zoom lens. From that it stands to say that a prime
    lens will blur light less than a zoom lens, if each
    piece of glass is equal quality.

    However, there is nothing to say that the total
    blurring introduced by more pieces of quality will
    be sufficient to reduce contrast in adjacent
    pixels. Now add to that the anti-alias filter
    commonly found on DSLR and it is easy to understand
    that there is a limit where improved resolution
    will not be seen. As a matter of fact, if the CCD
    were to see such resolution it would produce an
    aliased image."

    I feel that there might be other reasons why a prime
    lens would still produce better images, even with the
    higher-frequencies reaching the CCD. I pointed out
    that that different lenses yield varying qualities of
    blurred regions (boke). And of course, I can't get a
    f/1.4 zoom lens (but forget that for argument's
    sake).

    So what do you think? Does the traditional film
    advice of using prime lenses for producing high
    quality images also apply to digital SLRs? About how
    about some first-hand experience? (Note that the
    question is whether I will get better image quality
    between a higher-quality zoom lens and a prime lens,
    at equivalent apertures.)
     
  2. You should always use prime lenses whether they are of fixed focal length or zoom. Ask your friend if they have ever heard of another theory "Garbage in, Garbage out".
     
  3. "... I can't get a f/1.4 zoom lens (but forget that for argument's sake..."....the rest of your post is way too technical for me to think about, but the faster speed (ie f/2 and faster) is the main reason I use primes with my digital SLR. If you shoot a lot of low light, there is no other way to do it. Even f/2.8 at ISO1600 with a Canon 20D just don't cut it sometimes. that extra bit of light with an f/2 or f/1.8 prime does the trick.

    That doesn't answer your question.......directly, that is.....but it is the only reason I personally use primes on my DSLR.
     
  4. Jake, I think Martin is referring to the traditional use of the word "prime" in photography to mean single focal length lenses........not prime as in superior.
     
  5. Thomas, I'm sure you're right. To have a little fun, I used the word with yet another meaning "of the best possible quality; excellent". I think you can see how this play on words is appropriate as there are some very high quality zoom lenses as well as prime (single focal length).
     
  6. <<If the lens is too good, then the signal that passes through it may have more frequency than the sensor can capture.>>

    I believe there is an error in the definitions... the concept of "frequency" is not easily translated to the concept of "resolution" for purposes of making the aliasing theory hold here.

    Be well,
     
  7. Your friend clearly is thinking only in terms of sharpness, which is only one advantage of a prime lens (others including greater light-gathering capability, shallower depth of field when desired, and minimal or eliminated vignetting)
     
  8. Martin, another thought:

    This is an example of a theoretical discussion that can be tested with remarkable ease and speed.

    Find some well-lit subject matter than offers more detail than can be captured by the camera... high-contrast and low-contrast stuff in various colors... you know, a potpourri of detail-intensive stuff including fine print and itty bitty points of interest.

    Fix DSLR to tripod. Set your zoom lens at 50mm, focus as accurately as possible on your test subject stuff, and take a picture of the subject matter. Then mount a 50mm f/1.8 lens (change nothing else about the test conditions) and focus and take a picture.

    Which picture is better?

    The answer to that question is worth a lot, and may be worth more than any answers to the theorietical discussion.

    Use lenses of fixed focal length if you tend to like the results better than the results with your zoom. Sometimes, the subjective result is the only result that matters.

    Be well,
     
  9. Martin,

    Your story is too long to answer averything...and some already answered. e.g."On digital cameras, a lens of too high quality introduces aliasing artifacts", and pretty much everything else that you quoted.

    Where did you callect nonsens like that ?
     
  10. Does the traditional film advice of using prime lenses for producing high quality images also apply to digital SLRs?
    Yes.
     
  11. "On digital cameras, a lens of too high quality introduces aliasing artifacts."

    Not true. A high resolution lens at its optimum aperture and magnification does not produce any artifacts on the D70 or D1X.
     
  12. On digital cameras, a lens of too high quality introduces aliasing artifacts
    Your friend has been sniffing airplane glue. Tell him to either switch to something more biodegradable like Elmers, or switch to smoking crack. In either case please inform Mister Wizard that he's has no clue.
    Modern sub-frame dSLRs have several characteristics that make them different than 35mm film SLRs. The biggest is that the image sensor/size is much smaller than a 35mm frame. While this results in extra magnification for tele work like wildlife and turns your 50mm into an 85mm, it also means you are cropping the image, and hence being a lot harder on the optics.
    The next difference is dSLR's by default produce very soft images which aren't enhanced by fake contrast gradients that many high contrast films like Velvia and classic B&W films have. Add on very aggresive AA filtering and other processing, along with a perfectly linear contrast curve, and USM becomes your best buddy in Photoshop. Anybody who tells you they aren't post processing those ultra sharp dSLR images you see posted are lying.
    I've shot extensively with both Canon and Nikon film SLR systems before moving to digital, and I've found many of the lenses I thought were acceptable with film SLRs *are not* acceptable with non full frame dSLRs for the reason I've mentioned above.
    If somebody were to tell me I could only use Canon amatuer zoom kits with my 10D, I'd throw it in a river and go back to my FE2 and Sigma zooms because the results would be that much better. As it is, I stick to mostly my 50mm 1.4 and 100mm 2.8 on my 10D because by and large the half dozen other Canon lenses I've tried were too soft in comparison. They are literally like shooting 35mm film that's constantly/slightly out of focus.
    Zooms lenses have always been handicapped next to quality primes, but with 35mm SLRs and their larger image circle you could make a quality zoom that was close enough in performance to a prime to not complain to much. However, APS sensor based dSLR and their image circles magnify any miniscule optical problem and defect. The solution is either larger digital sensos like the 5D and much higher prices, -or- re-engineer your lens line-up to be optimal with the smaller and more demanding dSLR sensor frames. Unfortunatley Nikon and Olympus have taken this more seriously than Canon.
     
  13. It is obvious that whatever the light sensitive medium is going to capture the image (digtial or analog) it is the lens that must render the image. The better the lens then the better the rendered image. If the lens has poor edge to edge sharpness or vignettes or has other lens defects...how is that going to help the image?
    So, as others have said, the lens matters because the light matters.
     
  14. Consider that the film is sampling the incoming light....

    The real flaw in this argument is that sensor sites are not points. They have area, and more or less average the light over that area--supersampling. The antialiasing filter compensates for the dead space between sites, enlarging the sampling areas of the pixels until the meet and partly overlap.
     
  15. ZZZZZzzzzzzzzzzzzzzzzzzzzzzzzzzzz...........

    What Scott said.

    ZZZZZZZZZZZzzzzzzzzzzzzzzzzzzzzzzzzzzzzz.............
     
  16. Theories are so much fun. Ya gotta love' em. Tell your scientist friend to head on down to the local camera shop with a blank CF card. Put the card into a camera and format it. Then put a 50mm lens on the camera, put the camera on a tripod and take a picture. Swap the 50mm lens for a zoom with 50mm in the range. Take another picture at 50mm. Take out the CF card and bring it home and look at the jpegs on his computer.

    Have a good laugh.
     
  17. jbq

    jbq

    It takes extreme cases to see minor cases of artifacts. On my 5D with a 50/1.4 I can get minor moire when shooting around f/6-f/7 with the right kind of subject.

    Primes get measurably and visibly sharper pictures, better control over depth-of-field, and lower distortion than cheap zooms.
     
  18. As to wondering if you've got a relatively crappy lens, yes, you do. There are two things at work: First, what seperates good lenses from poor lenses is more than just design, it's tiht tolerances and quality materials - both of which are expensive. At the price range of the lens in question, you're just not going to find lenses that can match more expensive lenses in tolerance or materials, no matter how good the design is. Of course, there are a lot of aspects to lens quality, and those are just two - but in general, you get what you pay for.

    As to your friend's idea, he's crazy. First, if he's worried about the lens out-resolving the sensor, point out to him that the nyquist theorem says that you only need to sample twice along a cycle, so two pixels can (theoretically) sample an entire cycle. That means that a Digital Rebel XT's sensor could theoretically sample up to 70 lp/mm, and a D2x could sample 90 lp/mm. Now challenge him to count how many 35mm lenses he can find that will actually resolve 90 lp/mm with any useful contrast. Once he's found just how few there are, and how expensive they are, ask him just how steady you have to hold the camera during an exposure to keep from shifting the image just the 5 or 6 microns that it takes before you've shifted an entire pixel, and effectively halved your resolution.

    When he says that the problem only occurs if the blurring is enough to spill light to adjacent photosites, ask him just how large he thinks these photosites are - with high-resolution dSLR cameras, just the diffraction from stopping down your lens is enough to cause that to happen as early as f/8 to f/11.

    If his statement were true, the easy answer would be to increase the number of megapixels. It wouldn't be *that* hard, manufacturers are already packing 8 megapixels into comparatively tiny sensors in point-and-shoots, they could do the same on larger sensors to get whoppingly large numbers - but they'd run into the same problems with small photosites that they run into with point-and-shoots, not the least of which is noise. And if you think that being diffraction-limitted at f/8 is bad, wait until it happens at f/5.6!

    Also, his idea that a prime lens will out-resolve a zoom simply because the prime has fewer pieces of glass doesn't really hold water. Have him look at the MTF graph for the 50mm f/1.4, with just 7 elements, and a 70-200 2.8L with 18 elements.

    True, the more pieces of glass you have, the more you can accumulate errors from loose tolerances - but (to a point), tolerances are just a matter of cost. The problem with having too many pieces of glass *used* to be the reflection and resultant loss of contrast - you lost as much as 10% for each element in a non-coated lens. Fine, you say, don't use many elements. Then you have other technical deficiencies which you'll gripe about.

    The development of modern optical coatings proved to be the first great improvement, because they cut your loss down to as little as 1% - or less - per element. That allowed the use of more elements to correct various deficiencies. The second great boon was the advent of computer modeling - working out the optics of something like the 70-200 2.8 would have been simply impossible before the use of computers.

    I can certainly say that my 70-200 beats the pants off of my 50mm f/1.8 in every aspect related to image quality. Just Saturday I did a studio session with the 70-200, and at the end, switched to the 50 f/1.8, and it's a no-brainer to judge which images were from which lens. The only areas where the 70-200 loses are light-gathering and weight. Of course, it also costs 15 times more, but it does illustrate the point that a zoom is not *necessarily* worse than a prime.

    That being said, it's much easier and cheaper to design and produce a high-quality prime lens than a high-quality zoom lens. A prime lens of half of the cost of the 70-200 would likely do just as good of a job, or possibly even better. And where a $250 prime can be a somewhat respectable performer, a $250 zoom lens is going to be a study in trade-offs and compromises.
     
  19. Another thing to be aware of is just because the new lenses have Nikon stamped on them doesn't mean they're necessarily good. Case in point: my 18-55mm that came with my D50 is a piece of junk in my opinion.

    Furthermore, Nikon should not desecrate the ED (extra low dispersion glass) symbol by using it so freely on cruddy lenses. It makes it very difficult for us long-time Nikon fans to figure out which lenses live up to the Nikkor quality we have come to expect.

    That said, there are some great zoom lenses. For example, I have borrowed a friend's 17-35mm and it's quite good, though expensive.
     
  20. OK, look, you're drowning all of us in tech talk.

    Start from the very first issue. You're using a variable aperture, third-party zoom with an
    incredibly long range of focal lengths.

    Why not pick up a good zoom in order to improve your sharpness and clarity? I've got a
    Nikkor 17-35 and a something-to-200, both 2.8 and both fantastic on my D70.

    I've also got a Tokina 28-70 2.8 that I use, but it kinda sucks. It's pretty sharp in the
    middle, but soft around the edges. Unusable at smaller apertures, really.

    The other thing to do is look at your technique. If you're shooting really wide with your
    Sigma wide opened, you're going to have less sharpness than shooting at middle
    apertures. Also, what kinds of shutter speeds are you using? Are you applying sharpening?

    The point is, there are a million things you can do to improve image quality. None of them
    necessarily involve a debate over fixed focal length vs. zoom lenses.
     
  21. ..."So what do you think? Does the traditional film advice of using prime lenses for producing high quality images also apply to digital SLRs? About how about some first- hand experience? (Note that the question is whether I will get better image quality between a higher-quality zoom lens and a prime lens, at equivalent apertures.)"

    I have both Canon and Pentax DSLR systems, and use both prime and zoom lenses (Canon L series zooms and primes, their top line USM primes; Pentax A, FA and DA series lenses).

    Comparison of captures with matching focal lengths in both lines (between zoom and prime of the same focal length and at the same aperture) demonstrate that prime lenses continue to perform better than zooms with only rare exception. I'm sure all the same reasons that they perform better with film SLRs continue to apply.

    Godfrey
     
  22. Jim Gifford wrote:
    I believe there is an error in the definitions... the concept of "frequency" is not easily translated to the concept of "resolution" for purposes of making the aliasing theory hold here.
    Actually it does, if frequency refers to spatial frequency. Wavenumber and position are, of course, Fourier transform conjugates. But I do agree that the original objections are bogus.
     
  23. What Scott and Beau said.

    DSLRs, especially high-end ones, are brutally harsh about revealing optical defects in lenses.
    They benefit from the best quality lenses you can use. In most cases, that will mean primes
    instead of zooms.
     
  24. Start hanging out with artists instead of engineers and
    see if your photos dont improve.
     
  25. many of the shots on photo.net have much greater clarity that I've ever been able to capture

    God often lies in the post-processing.

    - Because of such a filter, optics of too high quality will neither produce aliasing nor give you any more detail than sufficiently good optics. "

    More detail? Maybe not. but higher MTF on the detail you have? Absolutely.

    "any blurring of incoming light would cause loss in contrast between adjacent pixels only if the scattering of light were to be larger than a photosite.?

    Thoroughly wrong. All you have to do is make light that should've been delivered to one pixel show up at its neighbor instead and you've lost local contrast. Of course, up to a point this effect is swamped by the scattering from the AA filter, but it's not zero. His theorem is like saying if you don't swerve out of your lane by at least the width of your car you won't hit the car next to you.

    Now add to that the anti-alias filter commonly found on DSLR and it is easy to understand that there is a limit where improved resolution will not be seen.

    Again, extinction resolution is far from the most important stat for judging sharpness, it's just the easiest to understand. Resolution at 50% MTF (modulation transfer function, that is, what percentage of the source's contrast is preserved) and resolution at 90% MTF are far more useful predictors of apparent sharpness. The vast majority of lenses cross these thresholds far below the AA filter frequency.

    So what do you think? Does the traditional film advice of using prime lenses for producing high quality images also apply to digital SLRs?

    Depends on the aperture you're shooting at. By f5.6 the great zooms (24-70/2.8L, 70 -200/2.8L) are very close to prime quality, better in some respects. The 24-70 at f5.6 beats all the canon primes in its focal range except the 35/1.4 and maybe the 50/1.4.

    But you mostly get what you pay for. "L" quality 3x zooms are a whole different universe from bargain 6x+ zooms.
     
  26. Rob Murray , dec 11, 2005; 11:39 p.m. hit the nail on the head when he wrote:

    "Start hanging out with artists instead of engineers and see if your photos dont improve."

    You mean we should stop taking pictures of test-charts? And instead focus on composition? On getting the right moment? What's the fun in that? ;)
     
  27. I'm with Rob and Peter above. Forget about it. Your mind and your eyes are the first "lens elements" in the optical/signal chain. Garbage in, Garbage out. If you clutter your head worrying about this trivia you won't be looking and seeing. If you can't love any of your pictures without them demonstrating lens/sensor resolution, become an optician and forget photography.
     
  28. Okie, Jim, I got the lens and ran the comparison you suggest. I did some rummaging in the kitchen and put together a bunch of objects on a table like dpreview does, near a window and with an overhead light. I setup my tripod. Some fixed parameters: - Camera: Nikon D70 - ISO 200 - Quality: fine jpeg (L) - WB fixed to Incandescent. - M mode. - Using AF (focusing in center on badardi bottle). - Lenses: Nikkor 50mm f/1.4, Sigma 18-125 f/3.5-5.6 I did four shots with both lenses: 1. F5 (5.6) @ 1/5s 2. F8 @ 1/2.5s 3. F11 @ 1/1.3s 4. F16 @ 1.6" To my surprise, the 50mm f/1.4 lens was not *amazingly* clearer than my cheap zoom lens. It does render a little bit more detail everywhere (especially at F11), but not by an insane amount, like I expected. I mean, you really have to magnify and compare to notice it. It's a bit sharper everywhere. I expected the difference to be much larger. Of course, I made some portrait photographs of my girlfriend last night that I could never have made with the other lens under the given lighting conditions... no regrets about getting this lens, it's fabulous. But I'm not so sure about the "clearer lens" part of the argument. See for yourselves: http://public.fotki.com/blais/primezoom_lens_comp/ If I made some silly oversight in my parameters, please let me know! cheers,
    00EZvl-27067684.jpg
     

Share This Page

1111