Sunday, December 6, 2009

Panasonic ZS3 versus Fuji F70EXR -- round 4 -- ISO Ladders -- high ISO remains Panasonic's Achilles Heel

Finally ... I have ISO ladders for the two cameras of an essentially identical scene shot within minutes of each other in identical light. Very low light ... shutter speed of 1" at 100 ISO ...

The test was pretty simple ... put a few elements together to test several aspects of low light performance for the two cameras. I was planning on three series ... one for ZS3 and 2 for F70EXR ... on at half resolution and one at full resolution. But then I realized that it was more work and would be pointless, since the full resolution images always look awful above base ISO ...

So I shot the two series only. Note that I shot this series last week and tossed the data. I shot it earlier today (twice) and tossed the data. I was having all sorts of issues with controlling the variables. With two cameras that have completely different control capabilities, it is a bear to even get them to properly expose in low light. Turns out the Fuji must be shot in manual mode if you need shutter speeds slower than 1/4 second. The Panny can be set to a minimum shutter speed of 1 second and that was enough for the test.

I had to tweak the focus and macro settings differently for each cam to get the same framing and magnification while still getting focus lock. They take a very different approach to minimum focus distance, but I was finally able to find common ground.

I used custom white balance on both cams, which is easy to do and works really well. I shot the images under about 300 watts of light bounced from the ceiling. These are very old and very warm halogens, which means that the custom white balance is pushing the blue channel *very* hard. This really shows up as an issue in the ZS3 images.

***Edit: To verify that the color issues in this post are caused by pushing the blue channel, I shot a pair of images this morning in very low but filtered daylight ... (cloud) ... on both auto white balance and custom white balance, the red came out perfectly. So this is an issue in tungsten light only. I also flooded the room with the same tungsten light and the red was again a bit more sane, but moving towards the purple again. I've found the ZS3 very sensitive to the red and blue balance, even in daylight ... and this confirms that it is really finicky.

On with the test.

Here are a pair of low ISO images ... full frame at 80 ISO for the ZS3 and 100 ISO for the F70EXR. I think the Panny does a nice job with the fur, although the Fuji might be a bit more details. That could come down to sharpening, as it does in the crops.

The colors here are very accurate, with strong reds. But the Panasonic is an entirely different story.

Reds on this camera are horrid. The blue channel takes over completely, causing the pure red of the hat and the top of the cam to render as purple instead of red. It's really hard to imagine how this could have got by the manufacturer's testing ... it's simply terrible.

Now comes the ISO ladder itself. These 100% crops at 5mp ... the ZS3 having been downsized to perfectly match the F70EXR's dimensions before taking these crops. Thus, the Panasonic has the advantage in noise characteristics.

Yet the Fuji pretty wipes the floor with the Panny in this series of crops.

Click on any of the images as usual and you will see the full version. WIth these crops, the Fuji has the better sharpening out of camera. That helps with crops, but is not really relevant when looking at finished prints as those would have been sharpened anyway.

Chroma noise starts to be visible in the white fur on the ZS3 at 200 ISO, while it is visible on the F70EXR at 400 ISO. The Fuji retains a one stop lead right up to ISO 3200 where chroma noise is concerned. However, 3200 ISO retains detail very poorly ... so the ZS3's 1600 ISO does tend to look a little better than the F70EXR's 3200.

The yellow can with lettering shows really poor saturation starting at 800 ISO opn the ZS3. The F70EXR never shows that kind of desaturation, however the highest ISOs do show massive chroma noise.

The lens cap portion of the crop shows how the cameras react to dark surfaces at higher ISO. The surprise here is that the F70EXR starfts showing chroma noise obviously as early as 200 ISO. The ZS3 of course starts there too, but that is expected based on previous results.

The black plastic cap is a goow way to judge 3 dimensionality for an image ... with the Fuji, all semblance of clear tone is lost at 1600 ISO. Before that, it makes a game of it. The Panasonic, on the other hand, does not get overwhelmed by grain the way the Fuji does ... but it changes color completely by 800 ISO, essentially being completely iverwhelmed by the blue channel. This will pose massive challenges for anyone shooting concerts ... by 800 ISO, the images will start showing a tendency to too much blue. I've seen good 800 ISO concert images from the ZS3 though, so it is not always fatal ...

Here are a pair of images at 1600 ISO. The F70EXR comes first ... this was tweaked a bit in ACR and then Topaz Denoise took a swing ... but there were no brush level tweaks, so this is an easy fix for anyone to achieve.

I have massively intervened in the Panasonic image with the hue/sat brush to change the magenta back to red, the desaturation brush to repair the fur, the color brush to paint and the yellow back into the tin (color blend mode.)

The detail retained by the ZS3 is remarkable ... easily the equal of the F70EXR. But the work required to get a usable color result is far too high ... and would not work on detailed images with a lot of colors ... and would be even worse with people.

Here's what the original 800 ISO ZS3 shot looked like straight from ACR ... not bad. No extra noise reduction here, but the color is pretty ghastly.

And here is the 800 ISO black and white image that shows how clean it can get. Excellent for 1600 ISO in my opinion. Note the smoothness of the color ... this is possible because I use the adjustment layer black and white converter, which is a very sophisticated channel mixer. I find it far better than using the equivalent in ACR. Strange that the photoshop one would be that much better. I've noticed this every time I've tried both.

I added a bit of glow and threshold work to the B&W to give it a softness that the others don't have. This is why the black and white images are pretty decent in round 3 ... this cam has some chops at high ISO.

And here are the 1600 ISO images from ACR and then after B&W conversion ...

Pretty challenging to get this far ... byt they are still passable as memories at web sizes. What I would recommend is, in really dark lighting at a pub or party, one can shoot at 1600ISO in black and white mode. That way, none of the brutal color blotches ever intrude.

And even after all the work I did ... the Fuji looks better in color. So for those who wonder about the two cameras as low light shooters, there is really no contest. The Fuji's reputation preceded it, and it easily lived up to it. But the Panny holds some surprises for those who like to process their images ...


Mark said...

Thanks for the comparison.

I have the F30 and the TZ7. I can't recreate the Red colour problem that you found but I don't have a Santa Hat, the can or the lighting to match.

The colour of Red on my TZ7 is slightly destaurated with a hint of orange.

Totally agree with your ISO findings. Just back from a weekend of gigs and the TZ7 is OK up to 400 but after that you have to have some luck with the lighting! The old F30 is still awesome at 800 but the thing is just so damn slow that you to be prepared (final 3 burst mode is handy!)
Mark (the TZ7 Nickleback Video guy)

Bob Crawford said...

I recently purchased the F70 and have been doing a number of tests on the camera comparing it to the F30 and the Panasonic TZ1. My testing definitely shows that the F70 is a great low light camera at high ISO's. There is one disturbing characteristic of the camera however that may be a deal breaker for me and this relates to the lack of resolution showing in the corners and around the outer perimeter of the frame compared to the center portion at f3.3. There is some reference to this in the forums and web reviews. I do not get the same resolution decay in the TZ1 at f2.8.
Have you noticed this in your F70 and how does it compare to the ZS3 in this respect? There is some vignetting in the corners as well. Using a f4.9 allows for a much better edge to edge resolution but this seems to defeat the purpose of having a relatively wide lens.