Reading mode: Light Dark

G1 and GH2 pushed three stops.

Go to last post
Louis_Dobson said:
One cheerfully says one will test stuff, and then one thinks about and realises it isn't that easy.
Click to expand...
Hello.

Thanks for this. My skip filling too longer than I anticipated and then the light went, so I'll have to do my G1 v G3 tests in the next day or so.

Interesting.

Looking at these particular comparison doesn't make me want to rush out and swap my current cameras for a GH2. Which is good news for my bank account.
However.

The GH2 is indeed a smidge smoother (although how much of that is the downsizing I'm not sure).

Also, 2000th of a sec .... hmmmm... and a white wall .... double hmmmm. Nastiness in the shadows manifests itself at its worst when you have darker shades - and lower light to start with. I am concerned that much slower shutter speeds and a different subject matter might extend the difference and the GH2's lead could extend.

Extrapolating from these, I am tempted to say that the GH1 might indeed be the cleanest of the lot as DXO suggests - I think it could be with that white wall for example. However as other people have mentioned, the GH1 Achilles heel is banding. You might not see it in the white wall, but if there were large dark patches, it would probably show up.

This is the tricky thing. A couple of people have mentioned (possibly in my thread) that I should download the DPR samples and peer into the bobbin cabinet (ooh err). However I suspect that if I did that with the GH1 sample, I wouldn't see banding. Even though I am 100% positive that if you gave that same camera to me and I went out shooting with it ... I could get it to band!!!

Ta for the samples anyway.

I think you might be interested in the G3 ones I'll produce. I suspect you won't be happy, as they will probably show the G3 has visibly worse noise (at 100%) than the G1 at base ISO. And unlike others, I don't have faith that Oly will be able to 'tweak' the RAW files in any meaningful way (in fact, I'd be very, VERY worried that they'll do what they usually do, and sprinkle in a generous helping of nasty banding which will be in wait for those of us wanting to push and pull on the files)

Cheers
G.
 
Big Ga said:
Extrapolating from these, I am tempted to say that the GH1 might indeed be the cleanest of the lot as DXO suggests - I think it could be with that white wall for example. However as other people have mentioned, the GH1 Achilles heel is banding. You might not see it in the white wall, but if there were large dark patches, it would probably show up.
Click to expand...
Banding in the GH1 can show up at ISO 1600, in underexposed areas. However I'm pretty sure "pushing" the exposure from a lower ISO, like was done here, would not show any banding, the pattern in the noise would not be there to start with.
 
PatrickC said:
Big Ga said:
Extrapolating from these, I am tempted to say that the GH1 might indeed be the cleanest of the lot as DXO suggests - I think it could be with that white wall for example. However as other people have mentioned, the GH1 Achilles heel is banding. You might not see it in the white wall, but if there were large dark patches, it would probably show up.
Click to expand...
Banding in the GH1 can show up at ISO 1600, in underexposed areas. However I'm pretty sure "pushing" the exposure from a lower ISO, like was done here, would not show any banding, the pattern in the noise would not be there to start with.
Click to expand...
And I'm even more sure the banding is there from the word go.
 
There clearly is a difference there, and it is nice to see how a GH2 might improve my post processing. Will wait another sensor generation before I decide, but the prices (the G3 double lens kit is down to 47,000yen!) on Panasonic are always dropping.

--
-CW
 
Big Ga said:
PatrickC said:
Big Ga said:
Extrapolating from these, I am tempted to say that the GH1 might indeed be the cleanest of the lot as DXO suggests - I think it could be with that white wall for example. However as other people have mentioned, the GH1 Achilles heel is banding. You might not see it in the white wall, but if there were large dark patches, it would probably show up.
Click to expand...
Banding in the GH1 can show up at ISO 1600, in underexposed areas. However I'm pretty sure "pushing" the exposure from a lower ISO, like was done here, would not show any banding, the pattern in the noise would not be there to start with.
Click to expand...
And I'm even more sure the banding is there from the word go.
Click to expand...
Do you even have a GH1? I do, and it's pretty clear that pushing a low ISO image does not reveal any banding that was previously invisible. Since I can't prove a negative, perhaps you could post a counter-example?
 
I think the downsizing is essential, a pixel to pixel comparison is meaningless....

Quite often a 2000th is my problem, differential development is my issue.

However, I will simulate a wedding for you later (I'll brush my hair and put on a tie).

what would you like, 1/15 wide open, pushed three stops?
Big Ga said:
Louis_Dobson said:
One cheerfully says one will test stuff, and then one thinks about and realises it isn't that easy.
Click to expand...
Hello.

Thanks for this. My skip filling too longer than I anticipated and then the light went, so I'll have to do my G1 v G3 tests in the next day or so.

Interesting.

Looking at these particular comparison doesn't make me want to rush out and swap my current cameras for a GH2. Which is good news for my bank account.
However.

The GH2 is indeed a smidge smoother (although how much of that is the downsizing I'm not sure).

Also, 2000th of a sec .... hmmmm... and a white wall .... double hmmmm. Nastiness in the shadows manifests itself at its worst when you have darker shades - and lower light to start with. I am concerned that much slower shutter speeds and a different subject matter might extend the difference and the GH2's lead could extend.

Extrapolating from these, I am tempted to say that the GH1 might indeed be the cleanest of the lot as DXO suggests - I think it could be with that white wall for example. However as other people have mentioned, the GH1 Achilles heel is banding. You might not see it in the white wall, but if there were large dark patches, it would probably show up.

This is the tricky thing. A couple of people have mentioned (possibly in my thread) that I should download the DPR samples and peer into the bobbin cabinet (ooh err). However I suspect that if I did that with the GH1 sample, I wouldn't see banding. Even though I am 100% positive that if you gave that same camera to me and I went out shooting with it ... I could get it to band!!!

Ta for the samples anyway.

I think you might be interested in the G3 ones I'll produce. I suspect you won't be happy, as they will probably show the G3 has visibly worse noise (at 100%) than the G1 at base ISO. And unlike others, I don't have faith that Oly will be able to 'tweak' the RAW files in any meaningful way (in fact, I'd be very, VERY worried that they'll do what they usually do, and sprinkle in a generous helping of nasty banding which will be in wait for those of us wanting to push and pull on the files)

Cheers
G.
Click to expand...
--
http://www.flickr.com/photos/acam
http://thegentlemansnapper.blogspot.com
 
Big Ga said:
I think you might be interested in the G3 ones I'll produce. I suspect you won't be happy, as they will probably show the G3 has visibly worse noise (at 100%) than the G1 at base ISO. And unlike others, I don't have faith that Oly will be able to 'tweak' the RAW files in any meaningful way (in fact, I'd be very, VERY worried that they'll do what they usually do, and sprinkle in a generous helping of nasty banding which will be in wait for those of us wanting to push and pull on the files)
Click to expand...
According to DxO the G3 beats the G1 in both (engineering) DR and SNR (18%), so it should have more 'usable DR' than the G1.
 
Good test Louis. The GH2 has substantially less noise than the G1. As for dynamic range, I think spacenegroes might have it right when he gives the G1 the advantage. There is more detail in the highlights, in the bottle labels and bamboo torch in front of the telescope. The shadows are a very close call and look pretty much identical in both (look at black railings against black background on the left.

So I think I'd give the G1 the advantage here for dynamic range, but does it lose that advantage when it's cleaned up to what the GH2 gives?

What it does show is that for dynamic range the sensors are very close, but advancement in sensor design has dramatically reduced noise. Again, useful test - thanks.

Michael
 
And open to interpretation!

DR is what's left when the noise is stripped. So to me the GH2 is clearly ahead on DR. As for detail, hmm. I suspect that is either:
  • Shot variation. What's there comes from the right hand side of the frame, and the two shots were not angled identically.
  • Poor downsiizing. I'm not sure what I should have done, but the G1 shot is untouched and the GH2 shot is downsized by bicubic in LR. Perhaps I should have downsized both to 4MP? Not sure.
  • AA filter. Possibly the GH2 has a more vicious AA filter.
mfj197 said:
Good test Louis. The GH2 has substantially less noise than the G1. As for dynamic range, I think spacenegroes might have it right when he gives the G1 the advantage. There is more detail in the highlights, in the bottle labels and bamboo torch in front of the telescope. The shadows are a very close call and look pretty much identical in both (look at black railings against black background on the left.

So I think I'd give the G1 the advantage here for dynamic range, but does it lose that advantage when it's cleaned up to what the GH2 gives?

What it does show is that for dynamic range the sensors are very close, but advancement in sensor design has dramatically reduced noise. Again, useful test - thanks.

Michael
Click to expand...
--
http://www.flickr.com/photos/acam
http://thegentlemansnapper.blogspot.com
 
Louis_Dobson said:
Downsized the GH2 picture in LR to the same dimensions as the G1.
Click to expand...
Isn't it a problem to compare one image in native resolution to another that has been scaled?

Have you tried working with a resolution not native (or a simple multiple) to either sensor so both have been through scaling? It would be very interesting to see if that has an impact.
 
I agree it's a problem.

One answer is to downsize both pictures to 4MB, but then do we get the detail? The other answer is to upscale both to 48MP, and then back down to 12MP....

Advise me!
iano said:
Louis_Dobson said:
Downsized the GH2 picture in LR to the same dimensions as the G1.
Click to expand...
Isn't it a problem to compare one image in native resolution to another that has been scaled?

Have you tried working with a resolution not native (or a simple multiple) to either sensor so both have been through scaling? It would be very interesting to see if that has an impact.
Click to expand...
--
http://www.flickr.com/photos/acam
http://thegentlemansnapper.blogspot.com
 
Amin, have you tried GH2 files in LR4b, prepare to be amazed they are infinitely better again and recoverable highlights is a new ball game? I dont know what this tells us about dr of gh2 but the recoverable amount is off the scale.
Amin Sabet said:
This sort of thing, when done well as you've done here, is much more useful than the various DR charts we're shown on the review sites. It's one thing to know which sensor records more DR and another to know how much that difference matters in a practical sense.

I did a similar test with the GH2 and Pentax K5 here: http://www.mu-43.com/f102/dynamic-range-whats-all-about-panasonic-gh2-vs-pentax-k5-comparison-11793/
Click to expand...
 
I think I'd probably show both at native resolution. Yes, they will be different magnification, but for dynamic range assessment purposes that wouldn't really matter. It would take resampling out of the equation.
Louis_Dobson said:
DR is what's left when the noise is stripped.
Click to expand...
DR is best assessed by looking for detail in black on black (or shadows), and white on white (or highlights). It would be interesting to see how the G1 fares when it has noise reduction applied to bring its output in line with the GH2 image. The slight advantage the G1 has in my perception may well be lost!

Michael
 
Amin Sabet said:
This sort of thing, when done well as you've done here, is much more useful than the various DR charts we're shown on the review sites. It's one thing to know which sensor records more DR and another to know how much that difference matters in a practical sense.
Click to expand...
Yes, a 3 stops push of the shadows seems like a good test of how much 'usable DR' a camera has. If it can handle that with a reasonably good result, then it should be capable of handling 99% of real world scenarios. Maybe such a test would be most realistic if the pushed shadows ended up somewhere in the midtones, since it's only the shadows that sometimes needs such a huge lift. We can't normally push the whole image like Louis did here, since that would blow the highlights and upper midtones in a 'properly exposed' image.
Amin Sabet said:
I did a similar test with the GH2 and Pentax K5 here: http://www.mu-43.com/f102/dynamic-range-whats-all-about-panasonic-gh2-vs-pentax-k5-comparison-11793/
Click to expand...
Fine test/comparison. The final image (with "auto tone" in LR) isn't much different from the original image though. Maybe the biggest problem here isn't so much lack of DR, but more how we can take advantage of the DR our cameras have, without making the resulting image look 'flat' and unnatural.
 
I think it DOES matter for DR assessment. Getting rid of noise is easy, just turn up NR. But that kills detail.

First off, I think we are disagreeing on what we mean by DR :-)

What I mean by DR is perhaps best described as latitude - "how many stops can I pulll and push (an area of) a photo and still have a usable shot?"

Which means asking what stops it being unusable, and that is normally lack of detail after NR has been applied.

So a higher res shot automatically has more "DR" because I can apply more NR.
mfj197 said:
I think I'd probably show both at native resolution. Yes, they will be different magnification, but for dynamic range assessment purposes that wouldn't really matter. It would take resampling out of the equation.
mfj197 said:
DR is what's left when the noise is stripped.
Click to expand...
DR is best assessed by looking for detail in black on black (or shadows), and white on white (or highlights). It would be interesting to see how the G1 fares when it has noise reduction applied to bring its output in line with the GH2 image. The slight advantage the G1 has in my perception may well be lost!

Michael
Click to expand...
--
http://www.flickr.com/photos/acam
http://thegentlemansnapper.blogspot.com
 
Louis_Dobson said:
I think it DOES matter for DR assessment. Getting rid of noise is easy, just turn up NR. But that kills detail.

First off, I think we are disagreeing on what we mean by DR :-)

What I mean by DR is perhaps best described as latitude - "how many stops can I pulll and push (an area of) a photo and still have a usable shot?"

Which means asking what stops it being unusable, and that is normally lack of detail after NR has been applied.
Click to expand...
Well, that's fine - so then to complete the comparison you could increase the NR on the G1 shot until it matches the GH2 shot, and see what is left over? :-)
 
Louis_Dobson said:
And open to interpretation!
Click to expand...
Louis_Dobson said:
  • Poor downsiizing. I'm not sure what I should have done, but the G1 shot is untouched and the GH2 shot is downsized by bicubic in LR. Perhaps I should have downsized both to 4MP? Not sure.
Click to expand...
.
Louis_Dobson said:
I agree it's a problem.

One answer is to downsize both pictures to 4MB, but then do we get the detail? The other answer is to upscale both to 48MP, and then back down to 12MP....

Advise me!
Click to expand...
Recently looked into what I could learn regarding (upwards and downwards) re-sampling.
.

To compare relative image-noise , it seems (to me) that downsizing the larger pixel-size image to the smaller image pixel-size (as you did) makes the most sense. The (visible) noise-spectrum will be reduced in some amount approaching that of the square-root of ratio of the pixel-areas. Lanczos-3 (as opposed to the PS Bicubic) might possibly be a better tool to use to perform such a task ? ...
.

To compare relative image-detail , it seems (to me) the upsizing the smaller image to the larger image pixel-size might make more sense. I am thinking that the interpolation processes of upwards re-sampling do not increase the (visible) noise-spectrum. The image-noise comparison would not be representative by this method, but the image-detail comparison seems to me like it would be valid.

The interpolation processes implementing upwards re-sampling do not appear to need or to use a (digital) anti-aliasing filter preceding them - whereas downward re-sampling processes do need one.

Downwards re-sampling processes use decimation only (if the down-sampling ratio is an integer number), but (in the more common cases of down-sampling ratios in between integer values) precedes the that decimation (which divides by integer numbers only) with interpolation processes implementing upwards re-sampling. However, to handle all possible down-sampling ratio cases, the downwards re-sampling process needs to be preceded by a digital (and non-ideal) anti-aliasing filter. Like all filter-alignments, no anti-aliasing filter is ideal in it's response. Best to avoid them if possible ?

So, my thought is that upwardly re-sampling the lower pixel-resolution image to the larger pixel-size of the higher pixel-resolution image may well be a better way to compare images at like pixel-sizes .

From what I have read, (certain) Bicubic Spline (B-Spline) algorithms are considered the best, with Lanczos following closely. There are many forms of Bicubic algorithms (as in PS) using different tap-weightings in the 4x4 cells. However, it looks like the B-Spline algorithms are a distinct (and more computationally intensive) affair (as compared to Bicubic algorithms implemented in PS/PSP).
.

One (8-bit TIFs only) implementation (of a B-Spline as well as a Lanczos-3 option) exists in the high quality, free, and non-invasive XnView image viewer's Resizing Tool options (along with Lanczos-3):

XnView 1.986 (just updated): http://www.xnview.com/en/downloadwin32.html

If nothing else, a great general purpose image-viewer (in my book) !

NDD Image Scaler portable freeware includes B-spline as well as Lanczos upwards/downwards re-sampling algorithm options, although it is older in it's origin, and XnView (may) be more current.
 
I believe dynamic range is the range between the brightest signal with detail and the darkest signal with detail. One component of this is noise, as noise kills detail. I believe tonal range is the number of tones between those values (lots of tones with smooth gradation). These are related and both of these are important for image quality.

How many stops you can pull and push seems a function of processing software as well as sensor qualities. We may be able to push and pull an image more with LR4 than PS or LR3, without any change in the sensor.
Louis_Dobson said:
I think it DOES matter for DR assessment. Getting rid of noise is easy, just turn up NR. But that kills detail.

First off, I think we are disagreeing on what we mean by DR :-)

What I mean by DR is perhaps best described as latitude - "how many stops can I pulll and push (an area of) a photo and still have a usable shot?"

Which means asking what stops it being unusable, and that is normally lack of detail after NR has been applied.

So a higher res shot automatically has more "DR" because I can apply more NR.
Click to expand...
--
http://fruminousbandersnatch.blogspot.com/
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top