sony a7r vs sigma sd1 18-35mm resolution test

Started 2 months ago | Discussions
docmaas
Veteran MemberPosts: 4,730
Like?
sony a7r vs sigma sd1 18-35mm resolution test
2 months ago

These are shots of a koren 2003 resolution chart using the a7r and sd1 with the 18-35 f18 lens.

camera to target distance same for all images.

all images are unmodified except that sharpening has been set to -7 for the sigma

white balance on both was set to auto

aperture priority mode was used compensation was set to 0

as usual sigma needs more care than I gave it. sony was much better in this respect

a7r shot full frame mode at 27mm.

SD1 shot at 18mm

SD1 enlarged via lanczos in Irfan view (b-spline and simple resize lower quality but same cutoff)

The amplitude of the max mins indicate the shades of gray so the low point is white and the high darkest. The lower the amplitude the lower the mtf or resolution (the less amplitude the closer the max is to the min and the less the difference between the two in degree of gray.

The x axis is length of the measured target in pixels nothing to do with resolution

These are relative measurements only and cannot be compared to any other measurements.

the flattening of the response indicates extinction.

There is more information on koren and ImageJ here: http://www.normankoren.com/Tutorials/MTF5.html

Under these conditions sigma does not match A7r full frame images when upsized appropriately by a factor of 1.5.

The images of the Koren charts are full size crops of the charts from the images as shot for the sony and uprezzed x 1.5 for the sigma.

full size crop of chart photographed image A7r

On these charts the top band and the fourth band are sin wave resolution charts running from 2 to 200 lpmm. But those numbers are only relevant if the charts are used as directed on the Koren link above THEY ARE NOT RELEVANT HERE. The second band is the same lpmm in linear format. I used the 4th band for the analysis.

ImageJ resolution plot of chart from A7r

full size crop of chart from SD1 uprezzed x1.5 on Irfan view using lanczos algorithm

ImageJ resolution plot of uprezzed chart from sigma SD1

Remember:

The images are unmodified with the exception of reducing the sharpness setting on the sigma.

They are only indicative of what the camera/sensor/lens SYSTEM is capable of without modification by a user.

The resolution difference is what I was looking for the the fact that the sigma drops off earlier than the sony indicated that it can't keep up in pure resolution.

The graphs are logrithmic and should not be interpreted linearly. The differences are greater than they appear when looked at linearly.

You can use the lpmm scale at the bottom of the actual chart to see the relative differences.

I will now don my fireproof suit and await the arrows.

Mike

Leica X1 Sigma SD1 Sony Alpha 7R
If you believe there are incorrect tags, please send us this post using our feedback form.
Kendall Helmstetter Gelner
Forum ProPosts: 18,152Gear list
Like?
Perfect demonstration of representative detail.
In reply to docmaas, 2 months ago

I think the comparison is really interesting, and that's the perfect lens to use with the Sigma for it...

The main thing I get out of it is what a great example it is of representative detail looks like - that is, detail captured beyond "real" detail.  The reason I call it representative is that while in theory it's not really what is there, you can tell clearly that what is there is straight up and down... whereas in the Sony as it goes beyond the edge of detail it breaks down and you start to see slanted lines, never mind the colors.  So in the end the usable detail captured is greater than the Sony, even if the pure numbers claim otherwise.  That's true regardless of shooting for color or monochrome conversion, where the meandering artifacts would be even more evident.

And of course, it's a demonstration why color moire can matter even just on a web image level never mind 100%...  small color shifts can easily get coagulated into way more visible ones in downsampling.

 Kendall Helmstetter Gelner's gear list:Kendall Helmstetter Gelner's gear list
Sigma 70-200mm F2.8 EX DG OS HSM Sigma 8-16mm F4.5-5.6 DC HSM Sigma 24-70mm F2.8 EX DG HSM Sigma 50-500mm F4.5-6.3 DG OS HSM Sigma 85mm F1.4 EX DG HSM +4 more
Reply   Reply with quote   Complain
MarkWW
Regular MemberPosts: 437
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

Thanks for this interesting comparison, it's nice to get some hard facts amid all the speculation.

A few thoughts.

1. The two images seem to have been exposed differently, which could give one camera an advantage over the other (specifically the one that has whites closer to white, especially with the Foveon sensor which is very sensitive to underexpsoure). The gray value passes 200 in the top image, it doesn't even reach 150 in the bottom image.

2. Does the actual measurement require the image to be uprezed? If not then why do it, surely the upsizing algorithm will introduce artifacts.

Otherwise, excellent job.

Reply   Reply with quote   Complain
tammons
Veteran MemberPosts: 4,928
Like?
Re: We need......
In reply to docmaas, 2 months ago

A full frame 24mp x 3  mirrorless Sigma with an EVF and a versatile lens mount system that you can mount most any lens on - period.

That would do it for me.

-- hide signature --

www.troyammons.com
www.pbase.com/tammons

Reply   Reply with quote   Complain
docmaas
Veteran MemberPosts: 4,730
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to MarkWW, 2 months ago

MarkWW wrote:

Thanks for this interesting comparison, it's nice to get some hard facts amid all the speculation.

A few thoughts.

1. The two images seem to have been exposed differently, which could give one camera an advantage over the other (specifically the one that has whites closer to white, especially with the Foveon sensor which is very sensitive to underexpsoure). The gray value passes 200 in the top image, it doesn't even reach 150 in the bottom image.

Yes the sigma is underexposed. I did correct it and measure the resolution but if there was a difference it was so little I couldn't see it so.  The software has better eyes than us.;<)

2. Does the actual measurement require the image to be uprezed? If not then why do it, surely the upsizing algorithm will introduce artifacts.

Only if you want to compare whether or not a smaller sigma sensor can match a larger bayer sensor which is a frequently discussed topic here and has been as long as sigma has been making digital cameras.

Otherwise, excellent job.

thanks and thanks for the good questions.

Mike

Reply   Reply with quote   Complain
khunpapa
Regular MemberPosts: 135Gear list
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

docmaas wrote:

I will now don my fireproof suit and await the arrows.

Mike

Beware your knees!! The arrow will be there <grin>
Oh .. I contribute nothing to this post except laughing ..

-- hide signature --

Flashes of my Memory.

 khunpapa's gear list:khunpapa's gear list
Sigma DP2 Sigma DP1 Merrill Sigma DP3 Merrill Olympus E-1 Olympus E-M1 +10 more
Reply   Reply with quote   Complain
mike earussi
Veteran MemberPosts: 4,681
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to khunpapa, 2 months ago

Good clear test, and about what is expected given that the max resolution of the Merrill sensor is 102 lpm. But do those test charts come in colors as well? It would be interesting to see how the resolution comparison holds for red or blue.

Reply   Reply with quote   Complain
docmaas
Veteran MemberPosts: 4,730
Like?
another chart
In reply to Kendall Helmstetter Gelner, 2 months ago

This chart better exhibits what I think you are speaking of. The charts I posted earlier don't show this because the lens is too close to the camera to capture the aliasing that comes after extinction.

But this chart does. Ignore the -20 and other stuff; it was originally a sharpness comparison of settings in spp but it has to be done again. Nevertheless it does show what happens with extinction and then subsequent aliasing. At least that is my interpretation.

Have a look here for some more relevant information: http://www.dpreview.com/forums/thread/3668227?page=1

If I break out the green red and blue channels on a sony image the green is pretty smooth in the rough area just before extinction the roughness comes from the red and blue channels.  I'm not sure this is color aliasing; I tend to think more of it just being marginal intensity as resolution breaks down.  I think the color aliasing comes after extinction just as it does in sigma images but with the color that sigma doesn't have.

I could be completely wrong here about what these areas are but it makes sense to me.

The charts for this test don't show this because as I said the camera is too close and while we do see extinction of real resolution at the end of the chart we don't see the aliasing that happens after that as we do here.

Kendall Helmstetter Gelner wrote:

I think the comparison is really interesting, and that's the perfect lens to use with the Sigma for it...

The main thing I get out of it is what a great example it is of representative detail looks like - that is, detail captured beyond "real" detail. The reason I call it representative is that while in theory it's not really what is there, you can tell clearly that what is there is straight up and down... whereas in the Sony as it goes beyond the edge of detail it breaks down and you start to see slanted lines, never mind the colors. So in the end the usable detail captured is greater than the Sony, even if the pure numbers claim otherwise. That's true regardless of shooting for color or monochrome conversion, where the meandering artifacts would be even more evident.

And of course, it's a demonstration why color moire can matter even just on a web image level never mind 100%... small color shifts can easily get coagulated into way more visible ones in downsampling.

Reply   Reply with quote   Complain
docmaas
Veteran MemberPosts: 4,730
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to mike earussi, 2 months ago

I thought of this too but was never able to figure out how to convert the Koren chart from black lines to colored lines.  It would seem there must be a way to do it but I never found it.  Maybe you can do it.  The charts are on the koren link.  If you do, I need the canon chart 5906 pixels length with .25 inkspread gamma 2 5mm size.

Mike

mike earussi wrote:

Good clear test, and about what is expected given that the max resolution of the Merrill sensor is 102 lpm. But do those test charts come in colors as well? It would be interesting to see how the resolution comparison holds for red or blue.

Reply   Reply with quote   Complain
D Cox
Senior MemberPosts: 6,955
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

docmaas wrote:

I thought of this too but was never able to figure out how to convert the Koren chart from black lines to colored lines. It would seem there must be a way to do it but I never found it. Maybe you can do it. The charts are on the koren link. If you do, I need the canon chart 5906 pixels length with .25 inkspread gamma 2 5mm size.

Mike

Illuminate the chart with coloured light. Stage lighting gels come in strong red, green or blue colours.

This gives you black-and-red lines rather than white-and-red (which is what you had in mind).

How to get red-and-green etc ? You might have to make your own charts on an inkjet printer.

mike earussi wrote:

Good clear test, and about what is expected given that the max resolution of the Merrill sensor is 102 lpm. But do those test charts come in colors as well? It would be interesting to see how the resolution comparison holds for red or blue.

Reply   Reply with quote   Complain
docmaas
Veteran MemberPosts: 4,730
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to D Cox, 2 months ago

Guess I figured it out last time after all.  I'll try something tomorrow.

Mike

Reply   Reply with quote   Complain
mike earussi
Veteran MemberPosts: 4,681
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

This is more like what I had in mind:

blue

red

Reply   Reply with quote   Complain
docmaas
Veteran MemberPosts: 4,730
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to mike earussi, 2 months ago

Even better Mike.

Mike

Reply   Reply with quote   Complain
Erik Magnuson
Forum ProPosts: 11,681Gear list
Like?
Use the channel mixer
In reply to mike earussi, 2 months ago

Use the channel mixer:

Blue/Black: (R=(0,0,0), G=(0,0,0), B=(100,100,100))

Blue/White: R=(100,0,0), G=(0,100,0),B=(0,0,0) then invert

-- hide signature --

Erik

 Erik Magnuson's gear list:Erik Magnuson's gear list
Canon EOS 5D Mark II Canon EOS 450D Sigma SD10 Sony Alpha NEX-5 Nikon D3200 +24 more
Reply   Reply with quote   Complain
bobsphotos
Forum MemberPosts: 72Gear list
Like?
Are these Charts printed on paper?
In reply to Erik Magnuson, 2 months ago

...or a sheet of plastic?

Reply   Reply with quote   Complain
bobsphotos
Forum MemberPosts: 72Gear list
Like?
nOT
In reply to bobsphotos, 2 months ago
No text.
Reply   Reply with quote   Complain
rjjr
Veteran MemberPosts: 5,720
Like?
Re: We need......
In reply to tammons, 2 months ago

tammons wrote:

A full frame 24mp x 3 mirrorless Sigma with an EVF and a versatile lens mount system that you can mount most any lens on - period.

That would do it for me.

I'd actually prefer a DSLR/OVF but I could live with that for a while.

Reply   Reply with quote   Complain
jennyrae
Senior MemberPosts: 1,373
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

you need metal armor for neck. guillotine is punishment for heresy.

Reply   Reply with quote   Complain
Scottelly
Senior MemberPosts: 1,138Gear list
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

docmaas wrote:

These are shots of a koren 2003 resolution chart using the a7r and sd1 with the 18-35 f18 lens.

camera to target distance same for all images.

all images are unmodified except that sharpening has been set to -7 for the sigma

white balance on both was set to auto

aperture priority mode was used compensation was set to 0

as usual sigma needs more care than I gave it. sony was much better in this respect

a7r shot full frame mode at 27mm.

SD1 shot at 18mm

SD1 enlarged via lanczos in Irfan view (b-spline and simple resize lower quality but same cutoff)

The amplitude of the max mins indicate the shades of gray so the low point is white and the high darkest. The lower the amplitude the lower the mtf or resolution (the less amplitude the closer the max is to the min and the less the difference between the two in degree of gray.

The x axis is length of the measured target in pixels nothing to do with resolution

These are relative measurements only and cannot be compared to any other measurements.

the flattening of the response indicates extinction.

There is more information on koren and ImageJ here: http://www.normankoren.com/Tutorials/MTF5.html

Under these conditions sigma does not match A7r full frame images when upsized appropriately by a factor of 1.5.

The images of the Koren charts are full size crops of the charts from the images as shot for the sony and uprezzed x 1.5 for the sigma.

full size crop of chart photographed image A7r

On these charts the top band and the fourth band are sin wave resolution charts running from 2 to 200 lpmm. But those numbers are only relevant if the charts are used as directed on the Koren link above THEY ARE NOT RELEVANT HERE. The second band is the same lpmm in linear format. I used the 4th band for the analysis.

ImageJ resolution plot of chart from A7r

full size crop of chart from SD1 uprezzed x1.5 on Irfan view using lanczos algorithm

ImageJ resolution plot of uprezzed chart from sigma SD1

Remember:

The images are unmodified with the exception of reducing the sharpness setting on the sigma.

They are only indicative of what the camera/sensor/lens SYSTEM is capable of without modification by a user.

The resolution difference is what I was looking for the the fact that the sigma drops off earlier than the sony indicated that it can't keep up in pure resolution.

If what you are saying is the lens can not keep up with the sensor in the SD1 Merrill, when compared to what is achievable from the same lens on a fullframe camera, then I would have to say I agree. If you are saying the sensor of the SD1 Merrill (and its associated systems that store and "interpret" the image information captured by the sensor), then I have to disagree with you. What you are seeing here may just be simply related to the fact that you are capturing the resolution of the lens, and the fact that you are using a larger sensor in the Sony, you are seeing a greater amount of lines resolved across the frame. Of course!

The graphs are logrithmic and should not be interpreted linearly. The differences are greater than they appear when looked at linearly.

You can use the lpmm scale at the bottom of the actual chart to see the relative differences.

I will now don my fireproof suit and await the arrows.

Mike

Why do you claim you are not modifying either of the images, after you state that you have "uprezzed" the Sigma images by a factor of 1.5?

"Under these conditions sigma does not match A7r full frame images when upsized appropriately by a factor of 1.5.

The images of the Koren charts are full size crops of the charts from the images as shot for the sony and uprezzed x 1.5 for the sigma."

I wonder why you decided to change the resolution of the Sigma images, rather than downsizing the Sony images.

Presumably you are trying to prove a point here. I think the point is simple enough to prove though. Shoot two identical photos of some sign far away, and let people look at them in full-size versions, so they can see for themselves that they can read the words on the sign in one photo but not the other. This is basically what DPreview does, with their test images. Ultimately though, I would suggest you use a Zeiss 55mm f1.4 OTUS lens, mounted with adapters, instead of the Sigma 18-35mm. That Zeiss would make a great reference lens. Still, you might see similar results, because you could possibly just be seeing limitations of the lens resolution in this test that you have done here. Think about it . . . if the lens is capable of resolving 1,800 line pairs across a 35mm sensor in the Sony, then it will resolve fewer line pairs across the 24mm width of the Foveon sensor in the Sigma. Could it be possible that we are still seeing sensors out-resolve lenses, even with high-quality lenses like the Sigma 18-35mm f1.8 A?

 Scottelly's gear list:Scottelly's gear list
Sigma SD14 Sony DT 18-55mm F3.5-5.6 SAM Tamron SP AF 10-24mm F/3.5-4.5 Di II LD Aspherical (IF) +5 more
Reply   Reply with quote   Complain
Scottelly
Senior MemberPosts: 1,138Gear list
Like?
Re: sony a7r vs sigma sd1 18-35mm resolution test
In reply to docmaas, 2 months ago

I really can't believe you have made these statements:

"The images are unmodified with the exception of reducing the sharpness setting on the sigma.

They are only indicative of what the camera/sensor/lens SYSTEM is capable of without modification by a user."

You changed to total number of pixels in the image from the Sigma. How is that NOT modifying the image? Everyone with a Sigma camera here knows what happens when they do things like this - the image resolves less detail. This is the main reason why people do not use the "Double Size" export feature or the Super-HI mode, when shooting JPEG photos.

This test is flawed.

 Scottelly's gear list:Scottelly's gear list
Sigma SD14 Sony DT 18-55mm F3.5-5.6 SAM Tamron SP AF 10-24mm F/3.5-4.5 Di II LD Aspherical (IF) +5 more
Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads