I saw a post title on dpreview not too long ago that immediately registered with me, because I find myself today asking the same question: f/4 or f/2.8? Or I suppose given my recent reviews of the Sigma DX zooms, we could also label that f/2.8 or f/1.8 for the DX market.
The question came up for me recently because when I reviewed the Sony A9, I only had the f/4 zooms, and none of the f/2.8 ones. (I'll use some mirrorless examples here, but the same things apply to DSLRs.)
So let me be explicit: the difference between f/4 and f/2.8 for the situations I tend to shoot a camera like that in is often the difference between ISO 3200 and ISO 6400. Why? Because I'm always trying to keep the shutter speed at 1/1000 or above (e.g. sports and wildlife).
As good as today's sensors are, that can mean the difference between getting 6.5 versus 7.5 stops of very usable dynamic range on a full frame sensor. That first number is definitely pushing my limits and testing my patience in tweaking the noise reduction settings in post processing. That second number stresses me far less, though I really want even more range to work with if I can get it.
When I originally started using the Sony A7 models, it was really to have a smaller, more compact full frame system for travel. Thus the f/4 zooms and a couple of carefully chosen small primes were just the ticket. But the A9 would fill a different purpose for me—and I believe for anyone considering it—so I was back to looking at my Sony lens kit.
This is a common question we all face in choosing lenses, whether for mirrorless or DSLR cameras. Like lens designers, we're faced with making compromises, as there's no such thing as an optically perfect, fast, small, light, low-cost lens. And when you start talking about the more complex lenses, such as zooms, the compromises start piling up.
This is one of the reasons why we photographers tend to go into Gear Acquisition Syndrome: we discover our original generalized choices suddenly don't quite work for us as we target a particular subject or shooting session.
So a few things to consider:
- Noise — One reason why we want faster lenses is to keep noise propagation down in many situations. Remember, exposure is light filtered by aperture filtered by shutter speed. What usually starts us down the path of faster lenses is that we are dealing with very little light, or we are trying to keep our shutter speed above a certain high point (e.g. 1/1000+ for sports, and even above 1/60 for human subjects that might move in low light). Less light means more shot noise, period, so even a perfect sensor will capture noise, and we're still a ways away from a perfect sensor. The only way you can compensate for less light is faster aperture or longer shutter speed.
Many think that bumping the ISO up is fine, but it really isn't. A good rule of thumb is to think of bumping ISO up by a stop as lowering the dynamic range by a stop (many modern cameras do slightly better than this, using various strategies, but it's still a good rule of thumb). Thus, I'd rather use a stop faster lens than bump the ISO in most cases where I'm trying to maintain a high shutter speed. - Complexity — Not often talked about in still photography are t/stops. Simple primes with few elements often have actual transmission of light (t/stop) very near the theoretical transmission of light (f/stop). Zooms can be quite far off. So, for instance, the Nikkor 35mm f/2D prime transmitted more than 90% of the theoretical light, making it a t/2.1. Meanwhile, the Nikkor 200-400mm f/4G zoom transmitted only about two thirds of the light, making it about a t/5.1. So you can't always go by just the published f/stop. Lots of element groups, even coated to reduce reflections, still mean that you have air/glass transitions that are going to take a bit of light away. Put another way, some f/4 lenses are closer to their f/2.8 counterparts in terms of getting light to the sensor than the one stop marking might suggest. That's because those f/4 lenses (usually) have less complexity in the optical design.
- Auto Correction — One of the things that doesn't get talked about enough is automatic lens corrections. Typically three things get corrected: vignetting, linear distortion, and chromatic aberration. Chromatic aberration, amazingly, is the least offensive of these corrections, as it often just turns out to (mostly) be a color correction applied to small areas. Vignetting, on the other hand, can mean that the corners of your frame are boosted by as much as two stops. Oh, oh. If I'm at ISO 3200, that means my corners look as if they were shot at ISO 12,800 when correction is on. In other words, noise proliferation can happen in the corners if you're not careful.
Unfortunately, not all lens corrections can be turned off, even in raw converters. Adobe, for instance, is now applying automatic lens corrections for several camera makers, with no easy means of turning them off. I don't mind if they're automatically applied, but I want to be able to turn that off on demand, too.
Finally, linear distortion is a tricky change to image data. We actually move pixels and make up new ones computationally from the ones we have. Your composition changes, too, as what you thought were the corners are no longer the corners. Really dramatic corrections—often the case with the wide angle zooms—introduce all sorts of visual aberration to the corners of the image: acuity is lost, shape is lost, and textures lose their regularity.
What's all that have to do with f/2.8 versus f/4? Well, the f/4 lenses often need to apply less correction than the f/2.8 lenses, and they tend to have an actual t/stop closer to the theoretical f/stop. Not always true, but true enough to be a rule of thumb. On the flip side, the f/2.8 lens lets in more light, so noise can be tempered.
So: compromises. You make compromises when selecting lenses. If noise is the primary thing you fear, select f/2.8. If massive corrections are what you fear, probably pick f/4.
