I often see comments that run along the lines of "Since digital cameras
don't have the resolution of film, you don't get any benefit from premium lenses
since the digital sensor limits image quality". I also see statements to the
effect that digital sensors need the very best lenses to give good results. True?
Well from a theoretical viewpoint I can make a convincing argument based on
convolution of MTF curves that better lenses should give better images even on
small sensor, limited pixel count digital SLR camera. But what do real tests
Well, I took some digital shots using a Canon EOS 10D 6MP DSLR and both a
Canon EF 75-300/4-5.6 IS lens (IS off) and a Canon EF 300/4L lens. Both lenses
were used at f5.6 and ISO 100 sensitivity was used on the camera. Shutter
speed was around 1/750s and the camera (or lens) was mounted on a sturdy tripod.
Note that these images are 400% blowups from the original digital file. We are
looking in close detail at the very limit of resolution so everything looks
Here, in the center of the image, you can see that the better lens not
only resulted in higher contrast, but also greater perceived sharpness.
Resolution is limited by the sensor, not the lens, but the better lens does give
a better image. The difference is less than you'd see on film under optimum
conditions. Still, you can score one for the better lens.
At the corner of the image the difference is more obvious. Again contrast with
the better lens is higher, but in this case the cheaper lens is clearly showing
much more chromatic aberration (green/purple fringes on the tangential
lines). Score another one for the better lens.
So it's certainly true that better lenses give better results even on a 6.3 MP
10D. Does that mean you should only use "L" lenses - of course not. There are
many excellent non-L lenses. For example the 50/1.8, the 24/2.8 and the
28-105/3.5-4.5 are all good lenses. In fact the 75-300 is a very good lens at the
shorter end, but as you can see above does suffer somewhat when zoomed to 300mm
and used wide open. It's still not bad, just not as good as the 300/4L. What is
equally true is that you can't just say "Well, the 10D doesn't have the
resolution of film so it really doesn't matter what lens I use". Just like with
film, choice of lens is still a factor in image quality, though probably not
quite as big a factor.
Just as an aside, one very interesting aspect of using a digital sensor is
that, to a large extent, the image resolution is independent of the ISO setting.
Above are resolution patches shot at ISO 100 and ISO 800. While the ISO 100 patch
might not be quite as good as an equivalent film test and high resolution scan,
the ISO 800 patch looks just as good! There's a bit more noise, but not enough to
have any effect on resolution. You can't say that about film!
What about film. Won't I get better results shooting film and scanning?
In theory, yes. It's not difficult to show that film with 80 lp/mm resolution
scanned at 4000dpi should give higher resolution than a 6.3MP digital image. 100
lp/mm film scanned on a drum scanner should be even better. But those are just
numbers. Also, resolution is not sharpness. Sharpness is a subjective quality
that depends on contrast, acutance (edge sharpness) and resolution, moreover the
perception of sharpness is related to some spatial frequency filtering that the
eye/brain system does. Our perception of sharpness depends more on how well
certain spatial frequency bands are reproduced and these aren't always the
highest spatial frequencies. Grain and noise probably come into the equation too.
There's much more to sharpness than just high lp/mm numbers. The only real way to
judge which of two images is sharper is to look at them under the conditions that
they will actually be looked at!
Test Targets: Here's a comparison of a two digital and
one scanned film shot of a cropped section of an image of a resolution test
target (all corrected to the same size on screen). The rightmost image is a
section of a shot of a test target taken with a 300/4L lens at f5.6 on Kodachome
25 film and scanned at 4000dpi with a Canoscan FS4000US scanner. The leftmost
image is a shot taken with the EOS 10D 6.3MP camera with the same lens from the
same distance. Since there's a 1.6x "digital multiplier", the field of view of
the 35mm image was cropped. The leftmost image compares the native resolution of
the 10D sensor with Kodachrome 25 scanned at 4000dpi. As you can see, there's
really not a lot of difference. Note that the 6.3 MP sensor on the 10D is about
15 x 22mm. If we had a full frame 24 x 36 mm sensor with the same pixel density
it would be a 16 MP sensor, so you can think of the 10D as having a cropped 16 MP
full frame sensor! That's better than any current DSLR, better than the 11 MP
Canon 1Ds and even better the 14MP Kodak 14n. Yes, it's true, the 10D has a
higher pixel density than either the 1Ds or the 14n - but of course the sensor is
1.6x smaller so the total pixel count is less. Given this and all the hype (some
deserved, some not deserved) about the 1Ds and 14n being better than 35mm film it
shouldn't be surprising that the left and rightmost images are quite similar.
However comparing the native resolution of the sensor to scanned film isn't
really a fair comparison in real world applications. To get the same digital shot
with the same magnification and field of view as the film shot we need to use a
lens 1.6x shorter in focal length, i.e. 187.5mm, so the center image was shot
with a 70-200/4L zoom set to 187.5mm and f5.6. Now the field of view of the
digital camera is exactly the same as the field of view of the film camera. As
you can see, under these conditions the resolution from the 10D is clearly lower
than the scanned film
A second way to look at these images is to say the left and rightmost images
represent what you might see if the made equal prints from the 10D and from a 15
x 22 mm cropped section of the 35mm negative (the 10D sensor is 15 x 22
mm). In this case resolution is very similar. The center and rightmost
images represent what you might see if you made the same sized print from the 10D
and from the full frame 35mm slide. The 10D image needs 1.6x more magnification
to get to the same print size and thus resolution suffers
But we don't spend our time taking pictures of test targets and looking at the
scans under high magnification - well, most of us don't anyway. We shoot real
objects and look at prints. In the real world, differences you can easily see in
scientific tests may not be so obvious. Nor may these differences be as
significant as they appear to be from scientific tests. Most real objects don't
have detail in the form of high contrast black and white bars with sub-mm
Real Images: It we want to look at real images we have
the problem of what to compare with what, but for this test I chose to shoot 35mm
film with a 500/4.5L lens on Sensia 100 and scan at 4000dpi with a Canoscan
FS4000US scanner, since that's a pretty typical example of what I use. The
comparison digital shot was made with an EOS 10D set to ISO 100 and a 300/4L lens
at f5.6. I know from previous testing that the 300/4L and 500/4.5L give very
similar - and very high - image quality. Both are capable of putting over 80
lp/mm on film in the center of the frame. With the "digital multiplier" of
1.6x the "effective focal length", or more correctly angular coverage, of
the 300/4L on the 10D sensor was equivalent to that of a 480mm lens on 35mm film,
so I didn't need to change my shooting position much to get the same image as
with the 500mm lens. The full frame shot is shown below (it's the digital shot)
and the red box shows the area enlarged,
Below is the first enlarged image. It's a 100% crop from the scanned film. On my
17" monitor with a screen resolution of 1280x1024 this would represent a "real
life size" section from a 40" x 60" print. It's approximately a 3.25mm square
section on the slide. Your monitor and screen resolution will give a different
scaling factor of course (unless it's a 17" using 1280x1024).
Now here (below) is the same section reproduced to the same scale from the 10D
image it had to be upsized by about 200% to match the scale of the 4000 dpi scan
since it started out around 2000 x 3000 pixels whereas the 4000 dpi scan started
out around 4000 x 6000 pixels. Actual lp/mm resolution is lower, as evidenced by
the less smooth diagonal lines, but viewed from a distance it's hard to tell
(remember, on my monitor this is equivalent to a 40" x 60"
print!). Also, the shadow detail is better AND the highlight detail
is better, even the color balance is better. Now maybe I could do a more
optimized scan (maybe not). Maybe I could have shot on Velvia rather than Sensia
100 - but I wouldn't normally do that. Maybe my exposure could have been tweaked
a little, but this is what I got in real life and represents what I'd be likely
to get under typical shooting condiions in the real world. This scan probably
took 5-10 minutes including loading the slide, running the preview, running the
final scan and tweaking it slightly to try to color correct. The digital image is
straight out of the camera, saved as a JPEG (not even as a RAW file) with the
default camera parameters and aperture priority autoexposure with multizone
metering and aufofocus. I pushed the shutter and this was what was recorded.
Again remember that though this image shows "jaggies" on some of the diagonal
lines, this image represents a section of a print much larger than anyone would
attempt to make from a 6MP DSLR file. Below is a representation of a 20" x 30"
print and a 10" x 15" print (as displayed on my
17" 1280x1024 monitor). These are just approximate sizes of course, but I don't
want people to get a false impression of image quality (or lack of it) by looking
at greatly enlarged images without realizing just how enlarged they are.
20" x 30"
10" x 15"
Based on MY particular film and digital workflow:
Even on a 6MP Digital SLR like the EOS 10D, better lenses give better
Digital scans of high resolution film (Kodachrome 25) at 4000dpi on a
FS4000US scanner yields higher resolution images than those shot directly with an
EOS 10D. Not surprising, but confirmed by experiment.
While 10D resolution is somewhat lower than ISO 100 film, it's very little
affected by ISO setting so it's quite possible that high ISO digital resolution
may be better than scanned high ISO film.
From a practical viewpoint, 10D images printed on an inkjet from digital
files are probably equal film up to maybe 11x14. They are certainly good enough
to be very hard to distinguish from film.
As an aside, if you want the ultimate in sharpness do it the right way. Get a
large format camera and shoot film!
Of course it's possible that if I shot everything on Velvia and had
professional drum scans made of my slides, film would look better than it does
when I shoot Sensia 100 and scan on an FS4000US. In fact I'm sure it would. If I
regularly made 20x30 prints, maybe drum scanned Velvia would be significantly
better than 10D digital images. Again, I'm pretty sure they would. However the
point is that I don't get professional drum scans done, and for
most of my work I don't use Velvia and I don't think I've
ever had a 20x30 print made from a 35mm slide, so such
comparisons, while valid in the abstract, aren't really valid for me and my
workflow. I'm just not going to use nothing but Velvia at EI 40, I can't afford
to get everything drum scanned and I'm not in the business of making 20x30
prints. So while film may be better in the abstract case, for me it's not.
For me I think digital has now replaced film. Not 100% but certainly 90%. I'll
still shoot some film, but the first camera I'll reach for is digital and I'll
only use film when I think it can do something digital can't or if I really need
the higher resolution that scanned fine grain high resolution film can give. Most
of the time I think the digital images will be good enough and for most of my
applications they won't look much different than optimized film scans.
Note I'm not saying "digital is better than film" or even
"digital is as good as film". I'm just saying that for me, most of the time
digital (from an EOS 10D) will likely meet my needs for image quality.