Modular Transfer Function (MTF)

Courtesy of Norman Korman (http://www.normankoren.com/Tutorials/MTF.html)

I was going to post this long time ago but today I'm motivated to finish it because a friend asked me "I have a 15MP Canon T1i camera, should I buy a Canon 18-200mm f/3.5-5.6 IS to replace my 18-55mm version?" Well, it depends. If the lens will give flexibility when shooting kids from wide to zoom, AND it's only going to be used on the internet (1600 pixel width), AND you don't care about optical quality, then by all means get the 18-200mm lens. On the other hand, the results may be disappointing when zoomed in to 1:1 resolution or printed at 8x10 or 11x14 and above because super zoom lenses like the 18-200mm get convenience and trade-off with inferior optical quality. So if one must use the 18-200mm lens, one might as well as shoot it on a 8-10MP camera because at full 15MP, pictures will have a tendency to look "soft" and "mushy"... they will look not much better than shooting with a lower megapixel sensor and then upsize it later in Photoshop.

Let me just say that when I need convenience, I'll simply use my wife's Canon SD750 point-and-shoot, which is a superb camera to shoot casual events... kids, pets, what not. I don't usually care about optical quality for casual events. In fact I don't expect the quality of the SD750 to remotely match that of modern DSLRs+pro-lenses. On the other hand if I'm serious about something, I'll get SLR equipments but not just any SLR equipments-- pro-lenses. There's no point of buying a toy consumer 18-200mm lens when a point-and-shoot will shoot just as well. Having that said, two years ago I bought the Canon equivalent-- back then a spanking brand new Nikkor 18-200mm f/3.5-5.6 VR. I had it for about 2 months. I sold it (on eBay) after only using it for 2 months because I was very disappointed with its optical quality. In fact, eBay today is full of used 18-200mm lenses-- people are now starting to discover that super zooms don't project good quality images, and are dumping them back to the used market.

So let's assume now that optical quality is in fact of high concern, how should one determine in general whether a lens is well matched for a particular MP sensor? You compare technical specs. However unlike looking at a single MP value, lenses are a bit more complicate. Lenses are usually characterized using the Modular Transfer Function (MTF).

Originally MTF started from USAF in the 50s when the air force wanted to know how much "lines PAIRS per millimeter" (lp/mm) a lens can resolve (this is different than l/mm!). The higher details, the more information they can get from the Soviet Union using high altitude spy planes and spy cameras. In the old film days, lp/mm was used commonly because there were only a few common formats: 35mm, medium format (45, 67, etc) and it made sense to talk about lines per millimeter projected on film. Today with a plethora of digital sensor sizes, people use lw/ph or lwph (line width per picture height) in favor of the older lpmm unit. Keep in mind, the two aren't equivalent and requires calculations to get from one to another, but for the sake of simplicity, let's just agree that the higher MTF value, the more details a lens can resolve. By the way there are a bunch of good technical MTF information online. My favorite is the following URL. Is this really necessary to read? Yes, absolutely, especially if you're a pixel peeping/engineering/math type of person:

http://www.luminous-landscape.com/tutorials/understanding-series/understanding-mtf.shtml

So going back to the question "I have a 15MP camera, should I get a 18-200mm f/3.5-5.6 IS lens to replace my 18-55mm vesion?" Unlike the camera body, we can't simply look at one single MTF value on a lens. What makes lens comparison complicated is that the center resolution will be different than the corner resolution because unlike a uniform sensor/film density, the projected optic quality isn't uniform throughout the area. To make things more complicated, lens resolution varies with aperture as well; subject to the laws of optics engineering and design, the higher the aperture, the higher resolution, and likewise the wider-open the aperture, the lower resolution. So we have two basic things to worry about now: 1) how much resolution in the different areas a lens projects 2) how much resolution using different apertures. In fact to make things even more complicated there are a bunch of other factors to consider (but we won't get into it here): contrast vs. resolution, color, chromatic aberration, vignette, distortion, and bokeh (BTW I hope I've made pixel peepers happy by mentioning all of these other issues).

Unfortunately, most manufacturers have their own MTF values derived from theoretical/calculated results. In fact, they don't make it easy for consumers to access that information. In addition, manufacturers' MTF is a lot lower than the theoretical values due to manufacturing variations, and subject to real world conditions (heat, humidity, calibration, irreproducible lighting levels, etc). Luckily however, there are numerous MTF charts published by independent testers online. Photozone is one very good European company that publishes optical results. Let's look at the MTF for the following two lenses:

18-55mm 55-250mm	18-200mm

Go to both pages, scroll to MTF for 18mm at f/3.5. You'll see that for center MTF value (middle of projection), both lenses resolve almost the same lp/ph. Now, look at border MTF value (edge of projection), and you see while the 18-55mm resolves at 2182 lw/ph, the 18-200mm resolves at an abysmal 1703 lw/ph. Extreme corner is another story to be told. Based on this MTF, one can say that the 18-200mm lens often yields mushy looking images. "This lens at f/3.5 and 18mm isn't very sharp" is what photographers would say. You can look at other apertures and other zoom ranges and see that aperture-for-aperture, zoom-for-zoom, the cheaper 18-55mm lens usually out-performs the much more expensive 18-200mm lens! (see footnote 1)

In short, lens engineering is about trade-offs, and one of the big trade-offs is convenience vs. optical quality. I understand that the 18-200mm lens is much more convenient for shooting pets and kids. It's certainly more convenient than having to switch between 18-55mm and 55-250mm lenses. So if convenience is the primary issue at hand, the 18-200mm is a fine lens. Just don't expect crystal sharp pictures.

If on the other hand optical quality is of importance (portrait, event, wedding, poster, etc), then don't get the 18-200mm. Instead get a better lens that can also resolve just as much resolution as the sensor has. MTF is your friend.

------------------------------------------------------------------

Footnotes:

1) One can make the argument that you can just close up the aperture to f/5.6 and beyond and both lenses will perform superbly. As good as that sounds, f/5.6 is 1.5 stop lower than f/3.5. That means the lens opening at f/5.6 is roughly 1/3 (1/(2^1.5)) than at f/3.5. Shooting f/5.6 under anything but bright day will yield a lot of disappointments.

Lynn & Ross' Wedding

Lynn and Ross are x-coworkers/friends from the Google days. They got married on August 22, 2009 in Malibu, Adamson House. What more could you ask for when you mix a beautiful day, a beautiful couple, and a beautiful location! Click www.lynnandross.com for more info.

I wanted to make sure this special couple had a good coverage throughout this once in a lifetime event and came in as a secondary/third shooter. Since we were not primary shooters, you will not see a single portrait of Lynn and Ross and no close-up of rings, shoes, etc on this web site. Enjoy!

I hope you enjoyed these pictures that Pam and I took. For other photographer's pictures, you can check out Michael+Anna's web site. They're very good, you have my endorsement!

It's all about the lenses!

I stumbled across a post on Craigslist. "High End Professional Photography Studio Sale (August 22&23) (1300 N. Wilton Pl., Los Angeles 90028)" I looked at the pictures in that post, and it brought back a lot of memories. All of the pictures here are from that Craigslist post.

Back to my story. The year was 1990. I was shopping for photographic equipments. I was armed with The Recycler (a print version of today's Craigslist/eBay-local). The rationale for shopping heavy duty photographic equipments was that they're good investments-- they would last a life time. After all, the basic photographic techniques had been the same since the early 1900s. You take a picture, develop on a canister, darkroom darkroom darkroom, print, done. It was done that way in 1900, in 1910, in 1920, 1930, 1940, 1950, 1960, 1970, 1980, and 1990 (black and white still done that way at the time). Why wouldn't it be any different in 2010, 2020, 2030? So I thought at the time.

I had an early version 1978 Pentax K1000 with a 55mm f/2 lens, and it was all I needed. The only other things I needed were darkroom equipments, and some black and white filters. I responded to a post on the Recycler. I showed up at this house in Granada Hills. This old man with white hair opened up the door. He said "oh, so you're Kevin?" He sounded a bit surprised, and I could hear the tone "Oh great, it's just a kid." He was a retired photographer selling off a bunch of old equipments, from drums, to easels, to enlarger, trays, chem bottles, camera lenses, everything. He turned out to be a nice and patient guy and explained some of the equipments that I didn't know how to use at the time. I ended up spending 4 hours looking at his arsenal of professional photo equipments. I came out with more photographic knowledge than before, more equipments, and an almost empty wallet. But I was ready to create my own photo lab, and I was excited!!! Many kids at the time just wanted to spend hundreds of dollars perfecting their Street Fighter techniques and such. I had my own darkroom... waaay cool, dude.

Even as digital photography became more prevalent in the late 90s, many professionals laughed at it. The resolution sucked. The color sucked. Everything sucked about it. Some experts estimated that it took 100 years for photography to reach the state of the art quality, and it would take at least a few more decades to do the same, digitally. But Moore's Law proved everyone to be wrong... doubling transistors every 18 months, and doubling computational power every 2 years or so... you continue this trend, and in 10 years, digital photography surpassed everyone's expectations. In less than 10 years, most of the pro photographers were already on digital (digital processing and digital capture).

When I finally parted with my darkroom equipments in 2008, I barely got back any money, after taking in consideration of shipping costs. You see these pictures of the old film equipments? They're no different than what I saw in 1990. The difference is utility-- they were useful in the 1990, and today, they're obsolete and worth very little. Hardly anyone wants to use an enlarger, let alone dealing with stinky chemicals and trays. Digital processing is not just "good enough", it is superior. With one mouse slider in Lightroom/Picasa/Photoshop, you can change contrast in 1 second. You can change the tone in 1 second. You can change the exposure in one second. You can burn and dodge in 10 seconds. A single print that may have taken masters like Ansel Adams hours to create, now takes minutes. In some instances, you can create amazing results in your digital darkroom, in seconds.

See the pictures on this post? Almost everything is worth... $0.00. They have little utility in today's world. The only things that are still worth something... are the lenses. Optics don't change like electronics. No matter what technology you're using, you need to capture light. Optical innovations isn't subject to Moore's Law. You see that old Nikkor AI 50mm f/1.4 above? It's still worth a lot. A good lens in the 70s, is still a good lens today.

It's all about the lenses!

Canon finally reversing their high megapixel trend

Today, Canon announced the Canon G11, a large sized point-and-shoot camera. The intended audience for the G-- series cameras include professionals and serious enthusiasts who use it in conjunction with their big SLR bodies. It is only 10MP. Interesting, the older Canon G10 had a whopping 14MP. It appears that Canon reversed the high-megapixel trend. What happened?

Well, remember the old post I made "High megapixel cameras preferred by lousy photographers"? For the past 3 years or so, Canon has making higher and higher megapixel cameras, esp. on small sensored point-and-shoot cameras. They have been doing so, because the laymen thinks that the higher the megapixel, the better it is. To Canon, the higher megapixels the cameras, the better they sell, so they kept going higher and higher. It is no different than a car maker enticing teenager drivers with meaningless specs like horsepower and torque. In the old days, the higher the horsepower and torque a car had, the better they sold. But then people got smarter, and realized that there was so much more that makes a car than horsepower/torque... things like comfort, mileage, ergonomics, reliability, cost of ownership, insurance rate, etc. As the car manufacturers mature, they started reversing the big horsepower trend.

IMHO, today is the day we can mark that the digital camera market is maturing. The fact that Canon reversed its high-megapixel trend, is a sign that digital camera shoppers are finally getting smarter, and more matured. They are now realizing that higher megapixel doesn't equate to a better camera. Let's take a quick look at images from the old G10. At the lowest ISO of 80, the images from the old Canon G10 (with a whopping 14MP) looks spectacular. But at the higher spectrum (1600-3200), it is completely unusable. It is grainy. It is artificial. It is ugly. As mentioned in my old blog previously, the trade-off for having high megapixel is 1) higher noise at higher ISO and 2) unusable pictures at less than ideal lighting situation. Granted, if the purpose of a camera is to use it during an ideal bright day or in a studio, a high megapixel is perfect for the job. But most people shoot in all conditions, day and night, indoors and outdoors, and most of the time in less than ideal lighting situation.

Let's take a look at a picture from a G10 (Courtesy of http://www.imaging-resource.com/):

100% crop shot at 100 ISO. It's got great tones, colors, and details:

100% crop shot at 1600 ISO. Look at the massive amount of in-camera noise reduction applied to the picture. It's very digital/artificial. It's ugly! We don't need to look at 3200 ISO even though the G10 can go up to 3200 ISO. That high ISO is marketing BS from Canon.

Not many people have had a chance to play with a G11 or look at sample pictures online, but the fact is clear-- Canon is reversing this high-megapixel non-sense. Personally, I think this is an indication that 1) the laymen are slowly recognizing that higher megapixel doesn't equate to better pictures 2) Canon is responding to the market demand, and making a camera that is more capable in less than ideal situations (vs. a high megapixel camera that sells well). In the past few years, numerous posts on photography forums indicate that people are now realizing that high megapixels means trading off resolution for image quality. Kudos to Canon for making a lower megapixel camera that shoots well, and kudos to all the buyers out there resisting buying a camera with high megapixels.

The digital camera market is maturing.

Model shoot - Jane Chiu

Jane Chiu is a former co-worker / model / friend who needs a few shots for her upcoming pageant. The location of this shoot is at Caltech. She's experienced so it's easy to shoot her, with minimal instructions and setup time.

At the time I had the luck of having FIVE assistants-- my beautiful wife Pam, her brother Ben, brother's wife Karla, her mom Yuen, and Jane's bf Rob. I designated Pam as the main light, Ben as the fill light, and Karla as the hair light-- the rest carried equipments. We had about an hour and a half to shoot (6:10 till sunset), so it was a very fast paced session. I like fast paced shooting sessions.

I prefer to shoot a client's head first-- the make up is still fresh, there isn't a lot of sweat, and the eyes and expressions are still sharp (model fatique kicks in after an hour or so). Therefore, I almost always start with head shots first. Below is a 3rd try. I kept this one because the face is clear, the lighting is soft (thanks Pam!) and the catch light in her eyes shows up clearly. Also I liked the blurred domes/arch that I placed in the back. Nikkor 105mm f/2.8G VR.

After a few usable head shots, I move to half body shots. The sun-ray hitting to the side of her hair and a little bit of the face-- flash magic. Nikkor 24-70mm f/2.8G.

After about 1/2 hour we moved to the hallway in Caltech. There were a bunch of people dressed up for their wedding shoot, and I decided to make them as the backdrop. No flash. No Photoshop magic, shot as is using a specialized "wedding lens" called Lensbaby Composer. Lensbaby Composer set on Manual exposure mode on f/4 aperture ring.

Every shooting session needs a darker, more serious shot. Here are two of my favorites using one of the dark buildings on Caltech. One is a soft dark shot, and the other one is a harsh cold-looking shot.

For some reason, many Asians like white-washed (hi-key like) shots. So here is one for the sake of variety:

Personally, I like vintage stuff. Below is an attempt at creating a 1960s film look. This could have been someone's mom in the 60s, or something:


In the end we had about 110 frames. 20 were duds (test exposures). I ended up picking only 30. All of these pictures were shot under an hour and a half, with five assistants. In the absence of assistants, I'd have to use umbrellas (instead of hand-hold diffusers), and it would have taken at least 4 hours. Thanks to Pam, Ben, Karla, Yuen, and Rob. That was pretty fun!

P.S. Feel free to send me more clients!

--------------

Technical data:

Nikkor 105mm f/2.8G VR

Nikkor 24-70mm f/2.8G

Lensbaby Composer f/4