Panasonic G85 Focus Bracketing and Stacking Test

[tom mann]

Well, you asked for it and here it is.  I picked up a Panasonic G85 and their f2.8 30mm macro lens within the last week and I wanted to test how it performs for focus stacking. 

First, this image shows how the model looks with the 30mm lens at f3.5.  This aperture should mean the lens is near it sharpest, and shows the problem that focus stacking solves.
 


So again, when adjusting aperture to something like f16 or f22 to get the whole model in focus, diffraction sets in and the image becomes blurry. This isn't too much of a problem when resizing for the web, but it actually is a problem for me for print (or so I think ). 

Panasonic has two modes to support focus stacking:

1.  4k Focus Stacking (all done in camera).  The camera takes a 4k movie of the subject and changes focus for the duration. The camera can focus stack frames of this movie, resulting in an 8MB image file.  This might sound small in the world of 24MB+ digital images, but 8MB is still good enough to print a 12" wide image. The image is cropped slightly, and you can see it doesn't quite have the same perspective.



I was pretty pleased with this, especially since the example on www.cameralabs.com wasn't great. There is a little wavering around some of the handrails.

2. Focus Bracketing.  You tell the camera how many images you want (up to 999!!) and the step. Step is the amount of focus change and appears to be a unitless number, and I initially went with 1.  I barely noticed any change in focus, and I actually thought it wasn't working.  I tried again with 5 but didn't experiment with any other combination.  I then loaded up the 25 images into Photoshop as one layered file and auto-aligned and focus blended them.



This is way more usable, with only a few focus oddities around the handrails.  I'll keep testing.

[Santa Fe Guy]

That is damned impressive Tom.
Rod.

[basementcalling]

Drool and slobber.

How's it compare to other camera set ups you've tried?

I'm about to get serious looking for a camera for my railroad needs.

[tom mann]

After a couple more attempts to get some nice focus stacking with Photoshop, I gave up.  PS makes it too hard to retouch areas that it didn't calculate correctly.  So I tried Helicon and got much better results.  Helicon worked in about 1/10 of the time, and it produced a much more usable result.  Plus, the retouch tools made fixing up the areas around the handrails easy.

This still isn't perfect, but click this and take a look at the 100% file.  Ignore the fuzz and dust.

[nscaleSPF2]

That is truly amazing, Tom.  Too bad that the cost of the camera is equivalent to the annual budget for the Waynesboro and Mt. Alto.  The vice-chairman of the W&MA has already put out a cease and desist order....

[tom mann]

How's it compare to other camera set ups you've tried?

@basementcalling : For what it is, it's nice.  I got it for two reasons:

1.  A small travel size camera that is good for 4k video. Weather sealing is a must, as is the ability to change lenses.  I had looked at the Sony Rx10 and Canon G3x, but I really wanted the flexibility of different lenses.  Plus, the 1" sensor cameras yield photos that look a little too much like the iPhone (which is actually pretty good for what it is).
2.  I wanted the 60mm FoV for macro shots. The tests I did with the Sony Rx1r showed that the look I was going for for modeling was about 60mm.

The camera is more of an electronic toy than a camera.  There are so many menus of settings, all jumbled up on the display that it's overwhelming. However, it has a nice touch screen so it's somewhat easy to navigate around. But there are annoying things: aperture value doesn't have a f in front of it, so it can be mistaken for shutter speed (which is shown as an 8 instead of 1/8).  So shooting at f8 for 1/8s shows up as 8 8 on the display.  The custom white balance doesn't show the temperature, so it's not clear if it registered correctly.  I know my lighting is exactly 5000K, but the ambient room lighting brings it down to around 4500K.  It would be nice to see if this is how the camera sees it.

You can set it up for what you need in a custom setting and just use that, so it's not really a problem.  It's just I'm used to more basics I guess. But I like the camera so far, but I don't know if I would want it as my only camera. The Rx1r's full frame sensor and lens combination is beautiful. It just doesn't do #1 or #2 above.

I like the idea of micro 4/3 now, as I said in another thread.  The small sensor size allows for high quality lenses at a fraction of the price of full frame. Plus, the lens' sweet spot is lower than larger lenses, which means the exposure time is significantly reduced - and this offsets the noise penalty of a smaller sensor.

That is truly amazing, Tom. 

Thanks! Look around for used/open box deals. 

[SSW7771]

Very nice! I have been looking to get back into photography after being out for many years. This smaller set-up might be the ticket especially since we travel a lot.

[tom mann]

Here's an n scale 50' car to show the perspective of the 30mm macro and how it makes the car look more massive.  Also, I shot this at f22 (note that the right side is still out of focus) to test a setting on the camera that corrects for diffraction.  It does make a difference as the inset photos show, but I'm not sure if this is just basic sharpening.



edit: It dawned on me that I might be able to get the whole car in focus if I focus on the center.

[peteski]

Yes, that technique works well. I take 3/4 view photos of model cars (1:24 scale) at my model clubs annual contest. They are sent into Scale Auto Magazine (Kalmbach publication) and are published in their contest annual issue.

I don't used focus stacking in those photos. When taking the photos I focus on the area right behind the front wheel well. That gives me acceptable focus of the entire model  (with slight preference for keeping the nose of the car in sharper focus than the rear end).

[Cajonpassfan]

Tom, thank you for venturing where I was a bit reluctant to go... I've been pondering sinking SPF2's annual railroad budget into one but couldn't bring myself to do it...yet
A question for you: Did you get the 12-60 (24-120) lens that comes with the G85 body for under a $100, and if so, what do you think of it? I'm not as interested in the macro (for admittedly gorgeous single model photos) as I am in doing layout photography and the in-camera focus stacking got my attention. I also love the outdoor-proof construction, for our outdoor adventures, even if it is a bit bigger and heavier than I'd like (price to pay for interchangeable lenses I guess).
A grand for your thoughts
Otto K.

[nscaleSPF2]

Otto, this may help you make up your mind:

https://www.dpreview.com/products/panasonic/slrs/panasonic_dmcg85

Still trying to lobby the board of the W&MA, to expand its budget...

[tom mann]

Tom, thank you for venturing where I was a bit reluctant to go... I've been pondering sinking SPF2's annual railroad budget into one but couldn't bring myself to do it...yet
A question for you: Did you get the 12-60 (24-120) lens that comes with the G85 body for under a $100, and if so, what do you think of it? I'm not as interested in the macro (for admittedly gorgeous single model photos) as I am in doing layout photography and the in-camera focus stacking got my attention. I also love the outdoor-proof construction, for our outdoor adventures, even if it is a bit bigger and heavier than I'd like (price to pay for interchangeable lenses I guess).
A grand for your thoughts
Otto K.

I didn't get that lens.  I was able to snag an open box deal on the body only, so I would have spent 200 more for the lens.  But for $100, it should be a great deal -  I just would rather have the 14-140 or there other, faster 12-60.

The in-camera focus stacking would only be suitable for 1024px wide web images.  There are too many artifacts for anything bigger - look at some of the handrail stanchions on the switcher.  Some are wavy. I'll post the full res photo tonight for you to see.

As you can see from the DPReview article, the camera is getting a lot of positive press. If you don't care about video, get the Olympus OMD EM5II.

[basementcalling]

@tom mann , I'll keep my eyes open. Definitely too rich for my train budget new.

By wavy stanchions, do you mean the left rear step handrail on the body having a wiggle part way up? I assumed that was just a bend in the wire.  I also noticed the first stanchion in front of the cab is perfectly straight, but the others apear to lean in towards the body. The one in front of the radiator screen seems to twist. Are these the stacking affects you are referring to?

It's still an impressive set of photos.

[Cajonpassfan]

Thanks guys! I did see the review, but there's little reference to the in-camera focus stacking capabilities. As I mentioned before, I'm mostly interested in that feature for "long shot" layout photography (like the fuzzy examples below) where decent depth of field is important, and I'd like to see how this camera stacks up, pun intended.

Tom, the 1024 limit is probably okay for my purposes, although it would be nice to have resolution good enough for magazine publication. I'll be following this to see what else I can learn here. Your info sharing is much appreciated!
Otto K.

[tom mann]

By wavy stanchions, do you mean the left rear step handrail on the body having a wiggle part way up? I assumed that was just a bend in the wire.  I also noticed the first stanchion in front of the cab is perfectly straight, but the others apear to lean in towards the body. The one in front of the radiator screen seems to twist. Are these the stacking affects you are referring to?

Some of these are real life bent stanchions due to my son taking this "train" outside to "show it to me" a year or two ago.

But this is what I mean:



At first glance it looks good, but the 100% inset crops show the issues. It could be something that can be fixed in a firmware release - who knows? Again, this is totally done in camera and takes a few seconds to generate.  So it's pretty amazing, it just needs some refinement.  It's possible that in other scenarios that lack discreet hard edges it would work just fine.