Resolution
5D vs 5D Mark II ... thoughts on resolution
Monday, April 13, 2009
Before going any further let me first state this isn’t an attempt at any type of definitive resolution test. I’ve been working with the 1Ds Mark 3 long enough to know it has more resolution than my 5D which is visually apparent in daily work, so the 5D has been relegated to backup duty, i.e. never used. The 5D Mark 2 has the same detail threshold as the 1Ds Mark 3 ... nearly identical sensors ... so there isn’t much need to test it.
My idea was to look at sensor resolution from a little different perspective. The idea is simple ... an object of the same size can be moved further away, but you have the same number of pixels available to record the detail.
Before going into this, a brief discussion regarding resolution. There is talk that current sensors out resolve lenses and further improvements in resolution are somewhat meaningless. Along with this, there is discussion how most photographers don’t need that much resolution. This evolves into discussion of appropriate viewing distances, and how only photographers examine prints closely to inspect detail.
I’ll admit that many photographers don’t need the resolution ... at least they don’t need it most of the time. Be that as it may, that doesn’t mean all. There are those where perfection is the goal, and the ability to record and reproduce as much detail as possible is one of the objectives. The other challenge is how do you know when you do and when you don’t ... and are you willing to live with the lack of resolution when you do need it? Personally I’m not there ... I’ve got way too many great digital captures that aren’t even close to resolving the detail I would like to see. There are fewer now with 21mp DSLR’s and my 39mp PhaseOne back ... but even with that I have images that disappoint.
As far as the other two thoughts ... I’m not buying them.
Most people don’t examine large prints up close because they offer nothing to examine ... they fall apart. But stand around in any gallery with great images which are large and offer great detail and you will see them examined by many. Great images can draw you in to examine subtleties that can’t be seen from a distance ... as long as there is something to examine.
And what about out resolving lenses? This has become the one of the key discussion points on many sites ... end this megapixel war, this megapixel madness because it’s now pointless. I’ve been on the fence with this idea, but after reading The Online Photographer: Why 80 Megapixels Just Won't Be Enough, I’m not buying that one either. There is no direct correlation between the detail in the analog world and the sensor sites. The reason digital data can even represent real information is the ability to sample sizes extremely small. Until we can perfectly record all of this data we haven’t reached the optical ability of the lens.
Yes, we have to be able to perfectly record even the imperfections of the lens, and we aren’t there yet. In fact, we have a long way to go. Many thought the 1Ds Mark II would out resolve many lenses ... and were surprised when it didn’t seem to make a difference. The same claims were raised with the 1Ds Mark III and the D3x - they would definitely out resolve most lenses ... again most are just fine. I’m not sure where sensor technology is headed, but consider this. The anti-alias filter of DSLR’s is an achilles heel for resolution. With enough pixel density this filter becomes unnecessary, and until that point the lens quality discussion is almost moot. In addition, as long as we have to use a bayer sensor, we need a massive increase in pixels to maximize detail ... where each of those 4 pixels of the bayer array are equivalent to a single pixel. Finally, until now there hasn’t been much of reason to really push optics since they were designed based on the resolution of silver halide particles. How much better can lenses be designed if we have a device which can actually take advantage of improvements?
Ok enough of that. My little test was pretty simple. Set up a scene that contains some small detail, and duplicate that detail further away. Then compare the closer objects shot with a 5D against the further objects with the 5D Mark II.
Here is the scene ...

A lovely spring snow has left my front yard a huge reflector. You can see 4 soda cans sitting in the driveway. I’m about 90 feet from the 1st two cans, it’s about 30 feet further to the other cans. They’re pretty small. Here is a 100% crop from each camera.

Note they are nearly identical in size. To me the one on the left seems every so slightly better. The key here is they aren’t the same cans. The one on the left is from the 5d Mark II .... and they are the cans that are 30 feet further from the camera. So I have the same number of pixels to record an object that is now 30 feet further away.
How does this relate? I see myself shooting a landscape of aspen trees ... and now my ability to resolve the leaves extends further away. How about when doing a group picture ... I can move 30% further away and add more people (in fact quite a few more) , yet resolve eye detail the same. I’m not sure how to measure this ... but then I don’t care how much ... I just know it isn’t insignificant.
In my meanderings on the internet I’ve read discussions about the anti-alias filter , and it appears from the discussion once there are about 30-35 mp’s on a full frame DSLR sensor the pixel density should be fine enough that moiré would become almost a non issue. At that point we wouldn’t have reached the maximum resolution to record analog detail perfectly with our digital device, but we might be at the practical limit - further gains may be pointless because of other limitations (such as human vision, printer/display technology, etc.)
I know this is contrary to what many preach in forums ... but in those discussions they really don’t analyze the challenge of converting our analog world to digital bits of 1’s and 0’s, there are way too many assumptions about the limitation of optics, and usually ignore the resolution killing use of AA filters ... not to mention the issues of using a bayer sensor.
“Spanish Moss”
Canon 1Ds Mark 3
24-105mm f/4L IS at 105mm
f/9 at 1/125th
ISO 200
A quick shot while playing at Hilton Head last year ... some nice textures which the 1Ds Mark 3 captured very well.