PowerShot S100 reviewed and compared to the S95

[neuroanatomist]

Thanks for the comments, feedback, and your testing, Paul!

I did notice your statement at the outset that said "converted in dpp using whatever the defaults were". So dpp applies different default conversion settings specific to each camera. Presumably, those defaults are chosen to optimize the final image, but to what standard?

Right, there is absolutely no doubt that dpp applies different settings based on the camera model.

The standard is...whatever Canon thinks looks the best, lol.

That's certainly the basis behind Picture Styles, ALO, etc. But, are you suggesting that if I use the same NR settings in DPP with different cameras, I will get different amounts of NR? Certainly, the defaults vary by camera. So, for example, the defaults for ISO 3200 are:

• 5DII - luminance 6, chrominance 6
• 7D - luminance 5, chrominance 12
• S95 - luminance 15, chrominance 16
• S100 - luminance 17, chrominance 17

Say I set DPP to luminance 8, chrominance 8 for all for cameras, are you suggesting those same numerical settings would be applied differently to the different images, based on the sensor/body? I'm not sure that this is the case (but knowing Canon, I wouldn't be surpirsed in the least).

Incidentally, I used the same NR settings in DPP for the tests, using the slightly more conservative S95 default values (i.e. I reduced the S100 NR settings a little bit from their defaults).

Wouldn't it be a more apples-to-apples comparison to use the same conversion settings and possibly turn off any software NR when trying to compare the noise performance of one camera to another? The final image is what matters so it might be appropriate to apply the optimal NR algorithms and compare images particularly if for some reason the NR algorithms would work better on one file than the other but I'm not sure that's the case.

No, not in my opinion.

First, what most people are really interested in (including myself) is the actual results they will get out of the camera in my own day to day use. If Camera A is theoretically better than Camera B, but in reality Camera B produces better pics because of more advanced processing, and I usually use the default processing, who cares if Camera A is theoretically better?

In this case the only way to do a "fair" comparison is to use the exact same workflow and tools you would use...that is also a little impossible as different people use different tools, and new versions of those tools come out later, but I try to stick with what I will actually use.

I agree, which is why I did not turn off NR for the testing. Although I normally use DxO for RAW conversions, in this case, using DPP for the 'tool' seemed best, since everyone will have access to it (and more importantly from a practical standpoint, at the present time no other software can process S100 .CR2 files!)

 

[Meh]

@PaulRivers
@Neuroanatomist

As long as the NR is being applied equally then it's a fair comparison and would be more representative of what users can expect but if there's any chance the NR is not being applied equally or if different NR techniques are applied then I don't think it's an absolutely objective camera-to-camera comparison.

Now if the noise characteristics of two image files are vastly different (which could be the case here) then an argument can be made that each image is best served by different NR techniques and settings but it still has to be noted. The argument to use the default settings because that's what most users would do has merit but a footnote would be in order. What if the default settings were extremely different... without knowing that, I don't think any objective conclusion can be reached. For example one might look at the final image and conclude that images from camera A are less noisy but also less sharp than images from camera B.

That being said, I do appreciate and agree that the quality of final images that can be attained is relevant.

 

[neuroanatomist]

@PaulRivers
@Neuroanatomist

As long as the NR is being applied equally then it's a fair comparison and would be more representative of what users can expect but if there's any chance the NR is not being applied equally or if different NR techniques are applied then I don't think it's not an absolutely objective camera-to-camera comparison.

Agreed. I assumed that by setting identical NR values in DPP, NR would be applied equally. As Paul points out, that may be an incorrect assumption. Certainly, not all NR is created equal - as I've stated, I find that DxO does a better job than DPP at NR, yielding less noise while preserving more detail. I hope DPP is applying the same NR setting equally. The fact that the default NR settings are generally higher for the 7D than for the 5DII, and even higher for the S95/S100, suggests that the scale is constant from camera to camera.

I could redo the comparison with NR set to off...but that would be assuming that 'off' is really OFF.

 

[anthony11]

With respect to the stated loss of max aperture at the long end: does that factor in the increase in focal length? Ie., the S95 went to 105mm equivalent but the S100 goes to 120mm -- is the max aperture still f/5.9 at 105mm on the S100, or does it maybe match what the S95 offers *at 105mm*? If it does, there really isn't anything lost.

 

[neuroanatomist]

With respect to the stated loss of max aperture at the long end: does that factor in the increase in focal length? Ie., the S95 went to 105mm equivalent but the S100 goes to 120mm -- is the max aperture still f/5.9 at 105mm on the S100, or does it maybe match what the S95 offers *at 105mm*? If it does, there really isn't anything lost.

The max apertures for FF-equivalent focal lengths are:

S100

• 24mm - f/2.0
• 28mm - f/2.2
• 35mm - f/2.8
• 50mm - f/4.0
• 85mm - f/5.0
• 100mm - f/5.6
• 120mm - f/5.9

S95

• 28mm - f/2.0
• 35mm - f/2.5
• 50mm - f/3.2
• 85mm - f/4.5
• 105mm - f/4.9

So, at equivalent focal lengths, the S100 is 'slower' by the following (approximately):

• 28mm - 1/3 stop
• 35mm - 1/3 stop
• 50mm - 2/3 stop
• 85mm - 1/3 stop
• 105mm - 2/3 stop

 

[PaulRivers]

That's certainly the basis behind Picture Styles, ALO, etc. But, are you suggesting that if I use the same NR settings in DPP with different cameras, I will get different amounts of NR? Certainly, the defaults vary by camera. So, for example, the defaults for ISO 3200 are:

• 5DII - luminance 6, chrominance 6
• 7D - luminance 5, chrominance 12
• S95 - luminance 15, chrominance 16
• S100 - luminance 17, chrominance 17

Say I set DPP to luminance 8, chrominance 8 for all for cameras, are you suggesting those same numerical settings would be applied differently to the different images, based on the sensor/body? I'm not sure that this is the case (but knowing Canon, I wouldn't be surpirsed in the least).

Hey, thanks for writing back. I'm always happy to see someone else do a review as well.

fyi, looks like in the link I posted earlier for low light -
http://forums.dpreview.com/forums/readflat.asp?forum=1010&thread=39700614

And for a listing of all my s95 vs s100 comparison threads -
http://photography-on-the.net/forum/showthread.php?t=1110007

I had thought the "View Original" link was turned on, it's turned on now.

Back to noise reduction, there are 2 separate things that I'm saying -

1. As you noted, different cameras have different defaults. For example, from the s90 to the s95, at iso80, detail noise reduction was by default set to "0" on the s90, but on the s95 it's set to "2" (in my opinion sensor noise was the same but with 0 blue skies were often slightly noisy, with 2 they were clear).

2. However, on an entirely different note, the actual noise reduction algorithm used by dpp is entirely and completely different based on whether the raw file was shot with an s95 or an s100. If you are used to how the s95 shots looked, and take high iso pics of people indoors, the differences are very noticeable.

I imagine Canon came up with new noise reduction with it's new processor, and wanted files converted with dpp to look about the same as in-camera jpg's, so they apply the new noise reduction to s100 files and the old noise reduction to s95 files. The s100 noise reduction nearly completely eliminates noise, but can sometimes give peoples faces a plasticy, or "to much makeup" look. The s95 pics never have this look. (On the other hand on other photos the s100's noise reduction manages to eliminate noise without giving people or objects this look, so sometimes it's an improvement and sometimes it's worse). It doesn't matter whether you shoot raw or jpg (for my comparisons I shot RAW+JPG so I would have both) - dpp converts them differently. And you know it's the noise reduction because if you turn off the noise reduction in dpp the s100 files no longer have that different look, they look fairly similar to the s95 files.

 

[neuroanatomist]

Interesting, thanks!

Actually, I'll be most interested to see how DxO does with the S100 files (I use DxO for the S95 and other RAW file conversions).

 

[Meh]

the actual noise reduction algorithm used by dpp is entirely and completely different based on whether the raw file was shot with an s95 or an s100.

Interesting, I wouldn't have thought so. Do you know this for sure or an educated guess based on comparing images. The S100 is an entirely new CMOS sensor so it stands to reason the images may look different primarily for that reason rather than how they are processed by DPP.

I imagine Canon came up with new noise reduction with it's new processor

Quite possible they used different NR techniques but not likely that they came up with new noise reduction. DIGIC5 is said (by Canon) to be far more powerful than DIGIC4 so certainly Canon could have implemented better NR techniques that were too computationally intensive and time consuming for DIGIC4 to process quickly enough.

On the other hand on other photos the s100's noise reduction manages to eliminate noise without giving people or objects this look, so sometimes it's an improvement and sometimes it's worse.

I think your observations are valid but it's very difficult to make comparisons about technical issues such as NR techniques without really knowing what's being used. But certainly what you see is what matters. Nice work on doing the comparisons. Much appreciated as I've been on the fence about whether to pick up an S95 on sale now or get an S100 in a few weeks.

 

[Woody]

The DPReview RAW comparison widget is up for the Canon S100. Does not appear to have any improvement over the S95. Better than Oly XZ1/Panny LX5 but much worse than Nikon J1/V1.

Canon's backside illuminated CMOS sensor is nothing to get excited about. I get the feeling Canon has reached the end of their sensor development capabilities.

A crying shame.

 

[neuroanatomist]

DPR's 'RAW comparisons' are entirely dependent on Adobe Camera RAW's JPG conversion. Take them with a grain of salt.

 

[PaulRivers]

the actual noise reduction algorithm used by dpp is entirely and completely different based on whether the raw file was shot with an s95 or an s100.

Interesting, I wouldn't have thought so. Do you know this for sure or an educated guess based on comparing images. The S100 is an entirely new CMOS sensor so it stands to reason the images may look different primarily for that reason rather than how they are processed by DPP.

If you take the time to read through my thread you'll see that's not the case, when doing raw processing and turning off detail noise reduction the s95 and s100 files look very similar. Put detail noise reduction at the defaults and the results are very different. They are more different in some pics than others, but in some pictures the difference is very apparent even at screen size.

Unless I worked for Canon specifically on s100 development (which I don't in any way), there's no way to say that one "knows for sure". But the results with people's faces in them make it obvious that the noise reduction method is very different on the s100.

I imagine Canon came up with new noise reduction with it's new processor

Quite possible they used different NR techniques but not likely that they came up with new noise reduction. DIGIC5 is said (by Canon) to be far more powerful than DIGIC4 so certainly Canon could have implemented better NR techniques that were too computationally intensive and time consuming for DIGIC4 to process quickly enough.

You say it's not likely that they came up with new noise reduction, then say they implemented better NR techniques with the new processor - sounds like the same thing to me. Not sure what you're trying to say there.

On the other hand on other photos the s100's noise reduction manages to eliminate noise without giving people or objects this look, so sometimes it's an improvement and sometimes it's worse.

I think your observations are valid but it's very difficult to make comparisons about technical issues such as NR techniques without really knowing what's being used. But certainly what you see is what matters. Nice work on doing the comparisons. Much appreciated as I've been on the fence about whether to pick up an S95 on sale now or get an S100 in a few weeks.
[/quote]

Thanks!

 

[Aputure]

Am I the only one who finds the operational speed of the S100 to be slower than the S90, as well as it being a touch softer? Can't comment on the S95. Here's my comparison/review: http://www.aputure.com/blog/?p=3271

 

[Seanlucky]

The vast majority of compacts have similar amounts of shutter lag.

There is one that I know of that does not - the Panasonic lx5. It's not "jeans pocketable", but it's still "could possibly cram it in a jeans pocket would just be really uncomfortable to walk around like that" sized.

Imaging resource's timing gave it's shutter lag timing almost the same as an entry level dslr - and more importantly I tried it out myself at my local camera store next to a Canon t2i and could not tell the difference in focus speed (unlike the s95, g12, and Nikon p7000 where I could tell the difference in focus speed). The lx5 has similar low light performance to the s100 (though not the s100's new noise reduction, which sounds like it would be a plus for you).

Good to know! I've been debating between the S100 and the LX5 for quite some time, going back and forth many times. Shutter lag (not just auto focus) is the big thing for me but I haven't been able to do a side by side comparison of the cameras (or even find either of them seperately in a local store...). The ONE thing I wish the LX5 had was the front control ring that I could set to aperture. That's one design the S95/S100 really did right. It feels like going back to my Contax RTS and controlling aperture on the lens, and shutter speed on the dial... Oh well, I'm sure I can get used to the LX5's control, it seems pretty easy to switch between the two controls on that rear dial.