Chasing Gains

My article about whether ISO is fake or not brought up a lot of discussion points. But there's one in particular I want to talk about today, because it plays into many of those discussions.

Are you chasing marginal gains?
Are you sure those gains are real?

Numerical gain conversations appear all the time now:

  • DxOMark numbers
  • Photonstophotos numbers
  • 14-bit versus 12-bit
  • 45mp versus 42mp (or 47mp, or even 36mp; any resolution number)
  • Dual gain versus Single gain
  • Low pass filter versus no filter
  • MTF line pairs values
  • Deep shadow banding
  • IS/VR actual capability (e.g. 4 stops CIPA versus 6 stops CIPA)

Now, my mantra has been and will continue to be: capture optimal data. To me that's the definition of the best practice of photography in the field. Optimal light, optimal composition, optimal exposure, optimal lens, optimal camera settings, optimal technique, optimal support, optimal composition, optimal everything (including me being awake! ;~).

That would make me a photophile. Much like audiophiles chased small and obscure gains so that they could hear the lute better over the bagpipes—that's a PDQ Bach reference, by the way—much of the discussion about "better" on the Internet now is about chasing very small gains

I mentioned whether or not you knew whether those gains were real or not. They might not be. Individual sensors can vary. Testing isn't perfect. Reporting of tests is often confusing or masking some assumption you didn't realize. 

One example of that I encounter all the time is the use of Imatest to produce MTF numbers. The actual number you get from a slant test—the one most of us use—is going to actually be dependent upon your test chart and the camera you use. 

If your test chart is a personally printed one on matte paper, that's going to produce different numbers than one of Imatest's expensive professionally produced charts. I happen to have three different-sized charts I can use with Imatest, and they all produce different numbers from the same camera and lens at the same settings. 

Thus, one thing I'd warn you about: do not attempt to compare numeric results from one site with another. You're going down a rabbit hole and will have an Alice in Wonderland experience if you do.

I'm not even sure you can compare MTF results within one site. Because those numbers often use different cameras, charts, and software for new tests than they did for older tests. You'll note that I periodically (about once a year), go through my posted lens reviews and update a few comments (and recommendations) when I start seeing that my previous results are no longer applicable to state of the art cameras, for instance.

Next, there's no precision in most discussions. One discussion that prompted me to write this article had a comment "clearly pulls ahead..." in it. Going back to the data that prompted that comment, the numbers were less than a third of a stop apart (and that's if you believed them to be accurate). Tell me how exactly are you going to use that less than a third of a stop gain when you're setting exposure in third-stop increments? ;~)

So many folk are buying now based upon fear of missing out (the FOMO acronym you might see on the Internet from time to time and wonder what it means). Welcome to paranoia if that's what's driving you. 

Here's a far better way to think about things: what's my budget for upgrading my optimal capture each year? And where is the best place to apply that budget (e.g. bang for the buck)? 

I'd argue that one place where we've gotten some clear useful change in the past five years is in lenses, not sensors. I'm seeing lenses these days that we would have killed for 10 years ago. Sensors? Not nearly so much. The last truly big and clearly useful changes in sensors occurred a bit more recently than a decade, say five to seven years ago depending upon type of camera.

Some of you may remember my words from 2003: "If you aren't getting good prints at the maximum size a desktop inkjet printer can produce, it isn't the camera." 

I recently was looking at some D100 images I took in that period—6mp DX crop camera—and they're pretty darned good. In fact, the primary things I wish were different in those images have nothing to do with more megapixels, better dynamic range, better lens, etc. (The dynamic range difference between the old D100 and the new D7500 at base ISO is almost three stops, by the way.) Nope, the primary thing I notice as I look back at old images like that is this: I wish my technique and composition were a little bit better. 

Maybe I missed exposure by a bit. Maybe I didn't see something in the composition I should have. Maybe I got the horizon line slightly off. Maybe I should have waited for the wind to die down a bit more. Maybe my timing was a little off. But it's rarely: gee, I wish I had another quarter stop of dynamic range or another 500 LW/PH increase in the lens MTF. 

I'm going to try to put this into some perspective soon. I've been fiddling with a series of very short videos showing how I approach processing an image. The image I chose to use? It comes from a 24mp camera/lens kit that currently sells for US$500. Shot in less than ideal conditions. And yet...the final results from that image can be pretty darned impressive. 

Just for my amusement, I went to photonstophotos to see how that camera stacked up against my Z7. Oh dear, at the shooting conditions I used the US$500 camera was about a half stop worse in dynamic range. Next, I went to a couple (of different ;~) Web sites to check on the MTF numbers: 2469 lines in the center on the US$500 camera versus 3280 in the center for the lens on my Z7. Oh dear, my photo is ruined! ;~)

Yeah. No. 

I once wrote about what to improve and upgrade first. It's time to revisit a simplified version of that. Violate this at your own expense. In order:

  1. Improve the photographer (e.g. your exposure, technique, and compositional skills)
  2. Improve your support (amazing how keeping the camera/lens steady improves results)
  3. Improve your lens (yes, many modern lenses can outperform older ones)
  4. Improve your camera

But I'm sure you want numbers, because that's what you continue to relentlessly pursue. So how about this: in the past 10 years I believe I've improved by two stops. Not the camera. Not the lens. Me. 

Chew on that stat for awhile. 

text and images © 2019 Thom Hogan
portions Copyright 1999-2019 Thom Hogan-- All Rights Reserved
Follow us on Twitter: @bythom, hashtags #bythom, #dslrbodies