Sunday 14 October 2018

GH-1 vs G85 (big generational change)

I've occasionally done a look at the successive generations of m43 cameras, seeing as I started with a G1 (nearly 10 years ago) and then moved to the GH-1 (about 8 years ago,which I've sat with ever since); where I've found that pretty much all the developments have been "interesting" but not enough to "tip me over the scales" and buy a new one.

To my observations the changes have been about conveniences; which would of course be worth having, but for me (who works in RAW and seldom uses the Out Of Camera JPEG file (which I classify as one of the conveniences)) the actual end results in images just don't warrant spending a grand on a new body.

I've looked at different generations twice on my blog:


  1. GH1 vs EM5 (mk 1)
  2. GH1 vs GH4
and both times found that when processing a RAW file there was little actually in it. Meaning that Panasonic just got the GH1 so right back then that every subsequent camera could more or less be a re-tune off all that research and development.

Today I had the opportunity to have a decent look at the output from the G80 / G85 (depending on the market you're in) and compare it directly with my GH1. I' ve decided to put the spec comparison graphs at the end instead of the beginning this time.

Summary: its a game changer for me.


The GH-1 runs out of ISO at 3200, but soon after my first image I knew that it wasn't worth going over 1600ISO for me (which to be honest fits in well with over 90% of what I do or even want to do). Accordingly I'd resolved to not upgrade my GH1 until either it died, or something substantial came along. Well to me it has.

The G80 is well capable of being used at ISO right up to 12800, with RAW results just way nicer to work with than JPG. Throw in the 5 axis stabilisation and its an incredible camera.

1600 ISO

Given this is my comfortable upper limit lets have a look at what the GH1 and G80 produce at that point. Interestingly they both have different views of what "incandecent" colour balance should look like, but that's nothing really problematic.


These are OOC JPG, and as you can see they're both fairly acceptable at this size. You can see a little more flare from the lamp as I clearly sat down in a slightly different position when I returned to the table with the different camera (using the same 45f1.8 lens)

So lets look at the results from the RAW, stepping right into 100% (pixel peeping) we can see a couple of things, shutter speed was 1/60th and the eye reflection shows that while the subject didn't move, the photographer did (shake) and you can see the drop in detail in the pores of the skin and the eyelashes


not to mention (if you open it up and view it full screen) the amount of (not horribly invasive) noise already present in the GH1 (right image).

So marks awarded to the G80 right there ... and its even clearer when you look at the hairs on his head (both the clarity and the colour channel noise).



this becomes patently obvious when we examine just the Red Channel (if you happen to be one blind to what colour noise looks like)



chalk and cheese right?

Worth noting is that on a single grey hair there you can see the artifact of a stepping pattern which is I believe exactly what you'd expect with the Anti Alias Filter removed (which the designers of this this camera have chosen to do) (if you don't understand what an Anti Alias Filter is please read this).

So to me this is a nice step in improvement and something I'd not seen in any previous sensors I'd tested.

Very encouraging, so now lets go beyond my comfort zone and look at the higher ISO

3200

3200 ISO is the limit that the GH1 goes to, but for some reason I failed, and didn't take a 3200 ISO with the G80, so we have to compare the 6400 ISO (which the GH1 doesn't even go to)


even at this magnification we can see the reasons I've always stayed below 3200 on the GH1, sensor noise is becoming clear and quite marked ... there would essentially be no hope of clearing that up.  As well there is some strange colour casts starting (which correction will exacerbate noise further).

The image from the G80 though is looking a bit "gritty" or "sandy" but is clean enough even if colour vibrance is falling off (something has to give right). Looking in close (50% this time) we see that skin tones are pretty reasonable with the GH1, but its just lost it with noise ...


and the G80 is actually a full stop higher ISO at 6400 ISO ... like wow

Again, looking at the red channel we see clearly that sensor noise is the real source of problems here, but the G80 is just amazingly showing nothing more than just noise which is rather akin to Grain ... so sensor pattern types of noise are amazingly well controlled here.



Lets push this further

25600 ISO

Well I mean this results in (our dim room) me having a shutter speed of 1/500th of a second ... shit it wasn't long ago that such was daylight exposures ... yet this is what we got:



Simply amazing ... and having a closer look:



Grain is becoming more apparent, but we've gone up 4 stops ... so yes, in the same light we've gone from 1/100th of a sec exposure to 1/500th of a second ... that's amazing.

Further this grain is amenable to noise reduction with even simple post processing NR like Nix



Lets compare that to the GH1 at 3200 again


So certainly its losing colour saturation (which fits what we know from the measurements) but with a little more tweaking (just some saturation and local area contrast masking) amazingly this shows you can still get a usable image from the G80 in what would really be impossible portrait lighting just a few years ago.



So to me this a big generational change, not just a small fiddle.

Wow

Specs - and what they reveal to support or even predict this

Despite the contention that exists on forum about DxO measurements of lenses (which I happen to find meets my personal observations) their methodology for sensor assessment is pretty robust. So here is the comparison between actual measured ISO and the camera setting ISO for my GH1, the GH4 and the G80/85


So looking at this graph we see that while the GH4 on its setting appears to goe to a higher ISO it progressively actually only goes to 12800 with 25600 being just a pat on the head to make you feel better (and makes worse images along the way).

Interestingly when the GH1 is set to 1600 (and gives an actually higher ISO of 2154) and the G80 is set to 2 stops higher 6400 ISO (and actually gives a lower effective ISO of 4408) they are infact of shutter speed given only one stop apart (not two) and have similar dynamic ranges (which can be seen in my images above if you go back and look.


I encourage you load the DxO comparison (which I've snapshotted above) and go through the test results carefully.

As I've discussed previously the GH1 is actually performing equivalently at a setting of 1600ISO (when you look at obtained shutter speed) to other cameras (such as the Olympus OM-D E-M5) when set to 6500 (and by that to when its set to 5000 it will give similar shutter speeds to the GH1 when its set to 1600). Meaning that there has been no significant sensor improvement between cameras which are six years apart in release date (2009 and 2015).

However as we've seen the G80/85 camera (and from what I've seen but not presented here the G9 too (and probably the GH5)) indeed does.

I encourage you to have a read of those linked in articles.

bottom line

I feel that the G80/85 is a worthy upgrade for anyone who does a lot of low light shooting, you can actually get usable images from light where the GH1 (or indeed most other m43 cameras) would struggle. Combined with some of your own image processing of the RAW files you can get quite decent images, and in my view probably the equal (in terms of signal to noise and dynamic range) of most full frame cameras (if you are one of those poor bastards somehow in total ignorance and misapprehension lusting after one of them because it'll make you take better images).


1 comment:

Lens Bubble said...

That little sensor in the G85 is indeed amazing; it rivals or even surpasses some APS-C sensors.