Monday, March 3, 2008

Sigma 600mm f/8 Reflex Performance Addendum

Truthsayer asked the following question in a comment:

Why are you comparing a 200mm lens to the Sigma 600mm?

That's an excellent question, and probably an indicator that I rushed the last post (Part 1 of the Sigma 600mm f/8 Reflex review) before it had all the explanation it needed. So I'll try to fill in some blanks about the performance of the Sigma 600mm in this post.

Before I go on, I need to specify that I'll only be discussing the pure, best-case, performance of the Sigma 600mm. In real-life use, the Sigma 600mm will generally not perform as well as these 'ideal' test shots. So this is more of a theoretical discussion of how well the Sigma 600mm performs rather than what to expect out of it when you use it. In my next post, I'll discuss the real-life use of the lens and its many handling quirks.


So, why compare the Sigma to a Canon 70-200mm f/4 USM?

1. The short version is that I had the Canon 70-200mm handy. It is currently my only telephoto lens.

2. The slightly longer version is that the Canon 70-200mm is similar to the existing lens of someone interested in purchasing a Sigma 600mm (admittedly, the Sigma is a budget lens). A better comparison would probably be one of the 70-300mm consumer zooms, or even the Canon 70-300mm f/4-5.6 IS. Comparing the $100, 20 year old Sigma 600mm to the $7,000 Canon 600mm f/4L IS USM or even the $1,000 Canon 400mm f/4L USM would be just rude.

3. The real reason is that I wanted to validate my purchase. If my tests conclude the same thing as this guy (that a 600mm shot is no better than an upsized, high quality, 200mm shot), then there is zero reason to keep the lens around. I could get shots just as good (and with less struggle) using the 70-200mm. That conclusion would make me very sad.


And what did I learn?

Well, luckily, my tests showed that the Sigma 600mm f/8 Reflex can easily achieve more detail than an upsampled shot from the 200mm end of the Canon 70-200mm f/4 USM, as shown in the comparison shot below:

Left: Upsampled Canon, Right: Sigma 600mm 100% crop

Essentially, for shooting pictures of birds or other far away objects (like the moon) I should be able to achieve a substantially more detailed image using the Sigma. Obviously not super sharp, but better. That makes it worth keeping around until I can get a used 400mm lens.

After I read Truthsayer's comment about getting a 1200mm effective focal length, I started to wonder about the Olympus series of cameras. It turns out Olympus dSLRs use the four-thirds standard, meaning their sensors are smaller than Canon APS-C sensor and use a 4:3 aspect ratio instead of 3:2. The image at right (from Wikipedia) shows the relative sizes of the various digital sensors.

I dug a little further and found that those Olympus photographers have really put a lot of testing into the OM mount version of the Sigma 600mm. For instance, this thread provides a well-done review of the 600mm and concludes that the lens is pretty darn soft. And this thread concludes the same thing, along with a discussion of many of the other reflex lenses that are out there.

Some observations and comments:
  • The first review/thread actually gives MTF50 measurements for the lens, which is something I've seen nowhere else.
  • The second thread has a lot of handling comments that I'll echo in my next post.
  • There may be an adjustment method (screws under the center obstruction) to get the lens in better alignment. I'll check this out when I get home.
  • The second thread has a moon picture which is softer than mine. I think I got a good copy of the lens and it works better on a Canon than an Olympus due to the sensor size (more on this later).
  • Sigma lenses are prone to mold (I knew that!).
  • The Zuiko 500/8 Reflex is very well respected, but it goes for $300+ on eBay. Not worth it to me since it is still going to suffer from most of the handling drawbacks of the Sigma. Here's a good comparison including the Zuiko.
I think the main problem for Olympus/four-thirds owners is that their smaller sensor size (a little more than half the area of the APS-C, and a quarter of the size of a 35mm frame) is exposing the optical deficiencies of the Sigma 600mm mirror lens. I would urge Olympus users (or other 4/3 bodies) to stay away from the 600mm Sigma or at least go in with reasonable expectations.

Luckily, on my 20D, the Sigma performs a bit better just because the pixels aren't as close together!


MTF Discussion (Geeks Only!)

Ultimately, it comes down to spatial resolution, or line pairs per millimeter (lp/mm). In the business, this is called a Modulation Transfer Function (MTF). MTF is a trait of both the lens and the camera. I don't want to describe MTF50 in detail here, but if you are curious, look at Imatest's page on MTF or Bob Atkins' page on MTF.

Imagine what's happening in your camera: light enters the lens, gets refracted/reflected, and then is projected onto the image plane inside the camera. If the lens is good, then the image on this plane will be very sharp (very narrow lines can be distinguished). But to get a sharp image, you also need lots of pixels to resolve the image projection from the lens. MTF is thus a factor of both lens and camera, with the limiting factor (lens or camera) deciding the overall MTF.

So, imagine the Sigma 600mm f/8 projecting an image onto 35mm film. The lens designers did what needed to be done to get decent MTF (lw/mm) on film. Not perfect, of course, but most people who have used the Sigma 600mm f/8 on a film camera have been happy with its performance. Now, take the same lens, and stick it on a 4/3 camera. Since the 4/3 image sensor is half the height of film, even though the lw/mm is the same, we've got less total lines/image on our sensor. And, as a result, the overall image is going to look like it has half the spatial resolution and appear softer than the image on film. In First Light's post, he quotes an MTF50 on his E-1 of 703 lp/ph. But, on 35mm film, that would be around 1400 lp/ph, which is generally considered to be a good resolution.

Note that I switched to lp/ph (line pairs to picture height) for this argument. On my 20D, FL's figure equates to around 800 lp/ph (it is actually a little higher when you consider the difference in aspect ratio). Not great, but only a little on the poor side of good. That pretty much summarizes my experiences: all my images are soft and probably comparable to upsampling an image from a decent 400mm lens.


Conclusion:

It isn't that the Sigma 600mm f/8 is so bad, but it works better on larger sensors. And for my purposes, since it outperforms my other telephoto lens (a Canon 70-200mm f/4), it will sometimes find itself on my camera.

As an example, here's a shot which I could never have gotten with my Canon 70-200mm:

American Tree Sparrow (?), Sigma 600mm F/8, 1/2000s, ISO 800

Ok, well, I could have gotten it, but it wouldn't have as much detail. The image above is nearly full frame; click it to see it at 1600pix!

No comments: