I don’t know about you, but I’ve always taken it for granted that my lenses out-resolve my camera. I’ve spent a lot of money over the years acquiring Canon L glass and, as I’ve moved up the ladder from older APS-C bodies to full frame, it just didn’t enter my head that my new cameras were starting to out-strip my lenses. Perhaps this was just denial – who needs to be told that they have to start the lens acquisition process all over again! However, I’ve started to contemplate a lens refresh program. Why? Read on…
DxOMark has just come up with a very interesting new concept for measuring lens resolution. It is called the Perceptual Megapixel method and it distills a whole series of measurements into a single number that is supposed to represent the actual number of megapixels that the camera body and lens combination delivers to the final file. Now, at this point, you may rebel and ask how a single number can represent the resolution of a lens across its entire range of apertures and, in many cases, its whole zoom range. But, suspend your disbelief for a moment because we’re going to use it as a method of comparison between lenses and draw some interesting results. The number itself may be controversial, but if it is consistently calculated across lenses, then we should be able to
use it for comparison purposes.
Here are a couple of interesting examples: let’s say you’ve shelled nearly three grand for a Nikon D800 with 36 beautiful megapixels. You decide to scrimp a little on a lens to give your wallet time to heal, so you buy a Tamron 28-300 full-range lens. According to DxOMark, you get a lowly 8 virtual megapixels out of that combo! All that money for the big Nikon and you get resolution right out of the previous decade! Here’s another example. Let’s say that you’ve got a Canon 5d mk II and you’re trying to decide on a 50 mm prime lens. It’s a crowded market out there, but you’ve got 3 lenses in your hot little hands: a Canon 50mm F1.4, a Sigma 50mm F1.4 and a Zeiss 50mm F1.4. The Perceptual Megapixel score for the Canon is 15, giving you just over 70% of your camera resolution in your prints. The other two lenses deliver a P-Mpix reading of 13, or just over 60% of your sensor’s resolution. Even better, the Canon is the cheapest of the three at $385 versus $500 for the Sigma and $800 for the Zeiss.
How many Perceptual Megapixals Do You Need?
Before we go on an all-out hunt for the lenses with the highest P-Mpix score, let’s talk about resolution requirements. First, if you are like millions and millions of people out there who never print their shots, then don’t bother reading any further. You need 2 megapixels max. If you crop out half your image, you can get by with 4 perceived megapixels. Go buy an old Canon Rebel and the cheapest lenses you can find.
If you make your 6×4 prints at the corner drug store or a lab and occasionally print up to 8×10, let’s calculate what you need. Lab printers average around 360 dots per inch, so for an 8×10 print, you’ll need somewhere around 10 perceived megapixels, depending on how much you crop. This is quite do-able with nearly any lens mounted on a full-frame camera. If you buy an APS-C camera, as we’ll see later on, your perceived resolution will fall a little short of the mark.
For those of us who print on inkjet printers, things get a little trickier. Some folks argue that you can’t tell the difference between a print resolution of 240 pixels per inch and anything higher. Others (e.g. Ctein), maintain that you should probably print at higher resolution (for an excellent article on the subject, click here). For simplicity’s sake, let’s assume resolution in the middle of the range and use 360 pixels per inch. For an 18×12 print, you will need 28 million pixels. For a 16×20 print, that number becomes 41 million pixels. In either case, your printer software will be up-rezzing unless you have a Nikon D800. It becomes pretty important to start off with the best quality image in the first place.
How Many Perceptual Megapixels Do I Get Out of My Kit?
Let’s take some typical camera and lens combinations and see how they do. Unfortunately, DxOMark is a bit out of date in the camera testing department and some of the more interesting camera bodies are too new to have a full range of tests done on them. I’ve picked the Nikon D3X (24 megapixels), the Canon 5d MKII (21 megapixels) and the Sony A900 (24 megapixels) for my comparison test of full-frame cameras. For lenses, I’ve selected the top of the line from the manufacturer in several categories: a wide-angle zoom (e.g. 16-35), a medium zoom (e.g. 24-70) and a telephoto zoom (e.g. 70-200). I’ve also selected a few primes at the 35mm, 50mm and 85 mm range. Here’s a table showing the results for Sony lenses:
As you can see, most of the Sony lenses are middling performers on a 24 megapixel sensor with the exception of the 24-70 F2.8 ZA. Here are the Canon numbers (yellow rows are lenses that exceed 70% of sensor resolution):
And, for you Nikon lovers, here are your results:
Here are some observations on these results:
- Wide angle zooms are generally disappointing, delivering 10-12 P-Mpix scores
- Mid-range zooms delvier decent results for Sony and Nikon at 14-16 P-Mpix. Canon’s entry is the older mark I model. The test results for the newer 24-70 F4 lens are not out yet.
- Canon has released new lenses in the 70-200 range that are stellar performers, reaching 18 P-Mpix. Nikon and Sony get middling performers at13 P-Mpix
- The 50mm primes are excellent performers. Nikon and Canon both make really cheap “nifty fifty” lenses that deliver good results in the 14-15 range.
- Sigma has delivered a winner with its 24-70 lens
How Do Lenses Compare Across Cameras?
Let’s take some good lenses and compare their results across cameras:
First up to bat is the “nifty fifty” 50mm F1.8 from Canon. As you can see, it delivers very good results on full frame cameras, with almost 70% of the sensor resolution being passed through. The news is not so good with APS-C cameras where the resolution drops considerably. Isn’t it interesting that the sensor resolution between the 20d and the 50d nearly doubled and yet the perceived resolution with the lens is nearly identical.
Let’s make it interesting now and choose a very good lens that is available on both Canon and Nikon mounts, the Sigma 24-70 F2.8. Here’s the chart:
This lens is capable of resolving about 70% of a 21 megapixel sensor. It achieves the same perceived resolution on a 24 megapixel D3X body. For some reason, it does better on a Canon 12 megapixel body than it does on the Nikon equivalent. I’m not sure I believe the Canon number.It sounds too good to be true. Again, it looks like there is a significant drop down from full-frame to APS-C
What About My APS-C Camera?
The news is not very good for APS-C cameras. Having a lot of megapixels concentrated in the middle 66% of the lens delivers about what you’d expect — 66% of the perceived resolution. Let’s take an example: the Canon 50mm F1.8 lens delivers a respectable P-Mpix score of 14 on a full-frame Canon 5d MK II. Even on a 12 megapixel 5d, the score is a respectable 11. On an 18 megapixel 7d, the score is 10. This is about what you’d expect — roughly 70% of the 5dMKII score because the camera is only able to use two thirds of the lens resolving power. The crackerjack lens in the Canon line-up is the new 70-200 F2.8 L II. This lens delivers a score of 18 P-Mpix on a 1dsMK III. On the 7d, you would expect 66% of that score and that’s exactly what you get — a score of 12.
My advice to owners of APS-C cameras is to trade up to full-frame. A used 5d MK II can be purchased for pretty close to $1,000 and will deliver more resolution per dollar than the best APS-C camera out there. This is probably true of mirrorless cameras as well. DxOMark hasn’t tested any of these little boxes yet, so it’s hard to know what they will deliver, but I’m guessing that the results will be similar to APS-C DSLR’s.
What About Medium Format?
Unfortunately, there are no camera/lens combos on DxOmark that I could find. However, I’m going to speculate that my observations for APS-C cameras work in reverse for medium format. If you take a large number of megapixels and spread them over a larger area and expose them through glass with a larger diameter, you will surely get better results. Any one of the current medium format digital cameras on the market is going to deliver better perceived resolution than a 36 megapixel Nikon D800. I’ll be watching the results in DxOmark to see if I’m right.
Should I Upgrade My Camera Body?
If your camera body is APS-C, then you should upgrade to full-frame. If you already have full-frame, then you may not need to upgrade. The P-Mpix results show that today’s newer lenses (e.g. Sigma 24-70, Canon 70-200) deliver excellent results with sensors in the 21-24 megapixel range. Upgrading beyond that (e.g. to a D800) may not deliver on the type of results you’re looking for. There are diminishing returns. The only lens tested across the whole Nikon range is the Tamron 28-300 unfortunately, but it illustrates my point. Upgrading from a D600 24 megapixel body to a D800E buys you a jump from 9 P-Mpix to 11 P-Mpix. Not the whopping big increase you’d expect!
IF the concept of a virtual megapixel can be used for meaningful comparison between camera/lense combinations (and that’s a big if), then there are several key conclusions that can be drawn:
- If you want or need super high resolution for printing large formal images, then you probably need to go for the larger sensor of medium format along with the matching large-scale lenses..
- In the full-frame 35 mm world, the current crop of very good lenses look like they peg out between 13 and 15 virtual megapixels. Buying a sensor over 24 megapixels looks like a waste of money until better lenses come along. This is a harsh message to D800 users.
- It looks like Canon’s strategy of refreshing its lens line-up is bang on. The 70-200 F2.8 is a very good example where the results are beyond anything else on the market and may have the potential to deliver even better results on a sensor in the 40 megapixel plus range. Some recently-released primes are showing similar results.
- APS-C cameras and micro 4/3rds cameras will not deliver high resolution images beyond 10 perceived megapixels with today’s lenses, no matter what size the sensor is.
This is probably not the message you were hoping for. It certainly puts me into a quandary. My main lens is Canon’s 24-105 F4. It delivers a mediocre 13 perceived megapixels of resolution. My 70-300 DO is even worse with only 8 perceived megapixels on my 5d MK II body. This is disappointing because I love the small size of the DO lens. I really don’t want to have to carry around some big white tube like the 70-200 L to get sharper pictures, but it sounds like I may have to. How are your lenses stacking up?