Monday 15 October 2018

Full Frame (Sony A7) Adventures

For quite some years I shot 35mm film (which nowdays gets called Full Frame (FF)), and there was no digital. Then when the first D-SLR's came out I was abjectly disappointed to find that they were APS-C and none of my EOS lenses (and I had a few) worked well on them (or so I felt), so perhaps its better to say "worked the same".. my beloved 24mm became a sort of slightly wide normal (at effectively 38mm) and was neither wide (like 28mm or less) or "normal" (like 50mm) and wasn't even "fast" being f2.8

Eventually the 5D came out but by then I'd accepted the APS size and by 2009 had sold my EOS system and moved solidly into the Four Thirds (or specifically micro43) sensors with my G1 (and later GH1 and GF1). I have of course written *much* about that on this blog.

Back in 2014 I was looking with interest at the Sony A7 as being perhaps the perfect platform to use legacy 35mm lenses because "they were designed for that coverage" and in particular in "wide" and "normal" prime lenses were well priced (in a day when everyone wants a 24-105 f4 zoom), and (as the famous Ken Rockwell's discussion on wide angle lenses between digital cameras here observes, primes can be quite the advantage in some ways:

?

Both the above lenses are 28mm (the two zooms variously wider and less wide).

I know some people like to suffer for their art, but that's not me, I'd pick the middle one. Thus I've always preferred a selection of compact primes to a couple of huge zooms ... it brings other advantages with it too ... like DoF control. With a 28 - 70 f3.5~5.6 zoom you don't ever get "shallow DoF" ... thus you have abdicated control.

So knowing that I like shallow normals (which has been a kind of quest for me since I've been into m43) where the absence (at first, then dearth) them in m43 has been a problem. Over time this was solved with both the 20f1.7 and 25f1.7 (and yes I know I can spend big and buy the 25f1.4 or 24f.12 even) which were good and I was glad of but still somehow something was missing for me - low light performance of m43. Sure, f1.7 helped give better shutter speeds, but not enough, not if you wanted higher quality detail (IE: more than facebook or other social media image making demands).

Now just the other day I compared the Panasonic 25f1.7 on my GH1 with the Pentax SMC thread mount 50f1.4 on my A7 (photographing a packet of noodles) but found that test both "wanting" (I'm not really into noodle packets) and limiting. Not least because photographing at that distance isn't quite what I'm into with "normals". Plus I've wanted to use my Canon FD50f1.4 which I have used and tested before on my G1 (which of course using only the center portion) turns it into a crisp and sharp 100mm. Yet it was designed as a normal and I wanted to see how it "looked" ...

Today FD adapter arrived (from my favorite maker of FD adapters, I strongly recommend his products) so I could put my FD50f1.4 onto the Sony and have a look at photographing a person in a "normal" context. I took two shots, one at f1.4 (cos like why not) and the other at f2.8 (or maybe f2?) and they looked like this:


F1.4

and at f2.8


Which to me look (unsurprisingly) exactly like what I'm after, in contrast the P25f1.7 gave this wide open at f1.7



Its not as distinct from the background as the 2.8 (and indeed it should be closer if we assume the 2stops that exists between FF and m43). Its also a bit "wider aspect ratio because its 4:3 not 3:2 .. none the less its nice, but still, the books behind John are clearer and attract (unwanted) attention.

Overall (especially at this size) there is little rendering difference between this and a phone (less if you've got one of those that does the background Out Of Focus (OOF) in software now.

So what else is different?

Well for a starter I wanted to keep "shutter speed" above a minimum to reduce the possibility of subject blur (and yes I know IBIS will help with that, but won't help it if the subject moves ... as people are prone to). I also wanted to compare noise (which I also expected to show a 2 stop advantage to the bigger sensor, so I put the GH1 at 1600 ISO and the A7 at 6400 ISO (2 stops). This resulted in

  • the GH1 giving 1/250th at f1.7 (so yes its a dim room on a rainy day)
  • the A7 1/400th at f2.8 (and just over 1/1000th at f1.4 which really will freeze motion)
I don't want to pixel peep the shitter out of this because quite frankly it doesn't need that to show the differences, nor is anything more than 50% needed to replicate what a print looks like.  However the sheer amount of extra pixels captured by the A7 (6000 vs 4000 high) makes it hard to present them similarly. So I've chosen to present half size images for the A7 and full size GH1.
So lets dive in...


Straight away, just like with the noodle packet (in the earlier above mentioned post usig the Pentax lens on the A7) we see that the amount of detail available in the A7 just outright exceeds that of the GH1, and also contrast is better. What surprised me was how much more the noise was intrusive in the dark areas and even on the sides of Johns glasses. To me I'd expected that 2 Stops would see them on more equal footing than this. That the noise is larger in size (because its pixels are effectively bigger in the picture) makes it harder to apply Noise Reduction to without destroying detail.

To make the noise clearer (and highlight why colour channel noise is the big culprit here), lets look at just the Red channel.


The A7 noise looks more like "random noise" and the GH1 shows that its on the ragged edge of electrical read (sensor) noise.

This has a very detrimental effect on fine detail, which becomes clear when we look at the shot below.


So you can see the stripes clearly in the cushion, but the pattern in the covering of the chair is simply missing in the GH1 image (buried in noise). Click that image (indeed you should all of them) and look at the larger size ... the details vanished because the "feature size" of the image was about the same as the size of the noise ... so... gone.

Conclusions / discussion

Throughout this discussion I've used the names of the two cameras (the A7 and the GH1) however what really is significant is that one is FF (the A7) and one is m43 (the GH1). This is really the most significant point. All of the observations I've made are consistent with the expectations of the understanding of what is different between these two systems.

Big ticket items first, the FF camera gives better handling of noise and (via lower magnification) better contrast if not higher details. While its true that m43 can resolve as much as FF can its also true that things can erode that in practice, which we see here.


This experiment has shown to me that FF does indeed have benefits that extend beyond megapixel count, that even scaling the FF image down to a smaller than what the m43 camera still has greater details.

The smaller pixel size (relative to detail) allows greater use of post processing NR without destroying what RAW file posses.

So if you (like me) don't mind using manual focus (which requires a better skill set as a photographer) then you can take advantage of some great optics and get great images. Looking at the "native lens" options (meaning with AF and electronic control) available in a "normal" I see:
  • Sigma 50mm f1.4 for US$950
  • Sony E 50mm f1.8 with OSS (an advantage) US$300
  • Sony FE 50mm f1.4 US$1500
  • Panasonic 25mm f1.7 $150 
  • Olympus 25mm f1.8 $350
  • Panasonic 25mm f1.4 $600
Given that the FD50f1.4 costs about $100 it means that for a photographer, you have access to quality optics (but without the "bells and whistles") for a lot less. Yet still it on this camera the system outperforms the m43 option in terms of image quality and noise when in challenging light. Even if the P25f1.4 above could resolve more it would be eroded by the sensor noise. Perhaps its possible that some of the newer sensors (like the G80/85 I wrote about recently would come close to improving that situation, but  I doubt it would equal it let alone improve it.

So in conclusion (as I've expected) the m43 system will in good light give results in sharpness and detail that are well and truly "good enough". FF only pulls ahead when low light shooting is called for.

I'm in no way intending as a result of these findings to move away from m43 ... it offers so many things that FF does not (especially without having AF lenses). I will however be (now that I have a FF system {or at least not a film one}) rationalising my m43 lenses to be more in keeping to what I believe that basis of m43 has been right from the get go ... a compact and light weight flexible and system capable of excellent results as long as you don't have to contend with low light / high ISO.

Some random examples




No comments: