Sony AR7IV vs GFX 50 for DSLR scanning.

I doubt that anything over 50 megapixels will make any perceptible difference, even from 6x7cm on 120 film. And certainly not from the puny 35mm format.

Apart from the huge file size, mainly containing 'empty' pixels - i.e. pixels showing no real detail - increasing the pixel number may cause a deterioration in perceived quality, due to interference effects between grain/dye-cloud clusters and the pixel spacing. Film doesn't play nicely with any digitising medium whose pixel spacing closely coincides with the average 'grain' clustering diameter.

I know from experience that there's no improvement in 'scan' quality between using a 24 megapixel camera to copy 35mm frames and using a 36 megapixel camera. Just slightly more visible patterning in the dye clouds.

You can see the effect here:

The film was 200 ISO colour negative material.

Great! That's good info. Any view on the benefits of pixel shift though? IQ test show improvements but not sure how it would apply to my needs though and how noticeable it would be when printing

Jim Klasson did an interesting test: Sony a7RIV with pixel-shift vs Fujifilm GFX 100

Any aperture used on the copying lens that takes care of film bowing and field curvature, is going to reduce the absolute resolution through diffraction. Plus your marked (scale) aperture is effectively made smaller through lens extension.

The diffraction vs depth-of-field trade-off varies across copying formats - e.g. FX, DX, etc. - but it's always a no win situation, with an almost equal outcome whichever copying format is chosen.The overall effect being to almost totally negate the use of a high-res sensor.

For example: I find I have to use a marked aperture of f/8 to get corner-to-corner sharpness in a copy from 35mm film onto a DX format camera, despite the aperture for best central resolution being f/5.6.

Now lens extension turns that marked f/8 into f/13 for the purposes of diffraction calculation, which means the best resolution I can hope for is around 60 lppmm, which is quite a bit short of the 125 lppmm that 24 megapixels on the DX format can theoretically resolve. Digital sharpening claws some of that diffraction loss back, but not 200% worth!

So no amount of pixel shift or extra sensor resolution is going to help. You're fighting a losing battle against the laws of physics.

Sure, you can open up your copying lens to a super-flatness-demanding f/4, but that still only gets you to a theoretical 120 lppmm, which no colour film is practically capable of resolving anyway. And remember, a colour film emulsion has measurable depth, which also has to be covered by depth-of-field, no matter how flat you squash the film. If you doubt it, just try to get the entire depth of a colour film in simultaneous focus under a microscope - an impossible task.

And that's my answer. Thanks. This is very informative!

Any aperture used on the copying lens that takes care of film bowing and field curvature, is going to reduce the absolute resolution through diffraction. Plus your marked (scale) aperture is effectively made smaller through lens extension.

 

The diffraction vs depth-of-field trade-off varies across copying formats - e.g. FX, DX, etc. - but it's always a no win situation, with an almost equal outcome whichever copying format is chosen.The overall effect being to almost totally negate the use of a high-res sensor.

 

For example: I find I have to use a marked aperture of f/8 to get corner-to-corner sharpness in a copy from 35mm film onto a DX format camera, despite the aperture for best central resolution being f/5.6.

 

Now lens extension turns that marked f/8 into f/13 for the purposes of diffraction calculation, which means the best resolution I can hope for is around 60 lppmm, which is quite a bit short of the 125 lppmm that 24 megapixels on the DX format can theoretically resolve. Digital sharpening claws some of that diffraction loss back, but not 200% worth!

 

So no amount of pixel shift or extra sensor resolution is going to help. You're fighting a losing battle against the laws of physics.

 

Sure, you can open up your copying lens to a super-flatness-demanding f/4, but that still only gets you to a theoretical 120 lppmm, which no colour film is practically capable of resolving anyway. And remember, a colour film emulsion has measurable depth, which also has to be covered by depth-of-field, no matter how flat you squash the film. If you doubt it, just try to get the entire depth of a colour film in simultaneous focus under a microscope - an impossible task.

Pixel Shift is far more than about "more resolution" (MPs) as outlined pretty well below:

Testing Sony's New Pixel Shift Feature in the a7R III

Advantages of this function are the removal of aliasing and moire, increased color accuracy, and most of all a great increase in sharpness.

Pixel Shift – What Is It and Why You Should Try It - Backcountry Journeys

Just to review, Pixel Shift is a feature that allows every single pixel on the sensor to have accurate values for R, G, and B color channels.

So yeah, I agree, you should try it and see if it does or doesn't help.

Anyhow, it seems that the majority feels that it will be minimal improvements in the end. Still unsure.

And you should be since the 'majority' here hasn't done any testing using Pixel Shifting vs. without. If you can, you should.

And you should be since the 'majority' here hasn't done any testing using Pixel Shifting vs. without. If you can, you should.

Yep! Will rent one next week. It is an unconvincing majority ;-)

Yep! Will rent one next week. It is an unconvincing majority ;-)

Love to know what you actually find. As for unconvincing, all I can think of is: If they have only imagined it, they haven't experienced it.;)

Love to know what you actually find. As for unconvincing, all I can think of is: If they have only imagined it, they haven't experienced it.;)

Will do!

But with pixel shift, it IS more information TRUE data albeit not more 'resolution' because the sensors are not sharing the data; each pixel is getting true RGB data, just like a scanner. It's not a trick. You are recording more actual information.

But it's still just a trick to take a higher res photo of a lower res subject. If your film frame has 20MP of image in it, a 100MP copy of the film frame doesn't do anything for you that a 50MP copy won't accomplish.

But it's still just a trick to take a higher res photo of a lower res subject. If your film frame has 20MP of image in it, a 100MP copy of the film frame doesn't do anything for you that a 50MP copy won't accomplish.

AGAIN; got NOTHING to do with resolution and EVERYTHING about capturing true, trilinear RGB data which isn't the case without Pixel Shift. Understand? ;)

I've taken many pixel-shifted images, and the color is indistinguishable from single frames without shifting, other than twice the effective resolution.

This is an Ektachrome slide, copied with a Sony A7Riii and Nikon 55/2.8 Macro without pixel-shiftin. The original slide (reduced) is shown, with a detail crop at 300%. The original slide was taken with a Leica M2 and 35/2.8 Summaron lens, of a valley near Jayuya, PR.

Of note, the grain is just as sharp at the edges as in the center, suggesting the slight bulging of the film can be ignored (at f/5.6). Secondly, more resolution would not produce more usable detail, unless you are fixated on dye clouds.

Thanks Ed. So your experience is that there is no benefit to pixel shifting at the resolution that I'm working out (ie 50mp)? I really appreciate your input done you have actually tried this!

I've taken many pixel-shifted images, and the color is indistinguishable from single frames without shifting, other than twice the effective resolution.

 

This is an Ektachrome slide, copied with a Sony A7Riii and Nikon 55/2.8 Macro without pixel-shiftin. The original slide (reduced) is shown, with a detail crop at 300%. The original slide was taken with a Leica M2 and 35/2.8 Summaron lens, of a valley near Jayuya, PR.

 

Of note, the grain is just as sharp at the edges as in the center, suggesting the slight bulging of the film can be ignored (at f/5.6). Secondly, more resolution would not produce more usable detail, unless you are fixated on dye clouds.

 

[ATTACH=full]1319589[/ATTACH] [ATTACH=full]1319590[/ATTACH]

AGAIN; got NOTHING to do with resolution and EVERYTHING about capturing true, trilinear RGB data which isn't the case without Pixel Shift. Understand? ;)

I understand what you are saying. I disagree with you. Pro tip: assuming that people who aren’t agreeing with you don’t understand you is a very annoying habit and breaking it would help people to like you better.

A 50MP Fuji captures enough information, including color information, to digitize 35mm and medium format film. Buying a new camera for this purpose is crazy, and it’s the kind of crazy that people on forums tend to encourage by spouting technical mumbo jumbo like you need pixel shift tricks to capture trilinear RGB data.

It is my feeling as well and I don't get swayed by forums unless there is a clear cut answer which is not the case here. Will keep on scanning with the Fuji in the interim!

I understand what you are saying. I disagree with you. Pro tip: assuming that people who aren’t agreeing with you don’t understand you is a very annoying habit and breaking it would help people to like you better.

 

A 50MP Fuji captures enough information, including color information, to digitize 35mm and medium format film. Buying a new camera for this purpose is crazy, and it’s the kind of crazy that people on forums tend to encourage by spouting technical mumbo jumbo like you need pixel shift tricks to capture trilinear RGB data.

I understand what you are saying. I disagree with you.

Wait, the facts are the facts about how the capture is produced with respect to true RGB data. I NEVER said it was better for use in capturing slides. I said the OP should do his own testing to see. But clearly, the examples provided in TWO URLs I provided DO show there IS a benefit to some captures with Pixel Shifting; you (well I) can clearly see this. Does this mean for capturing slides it's a benefit? Well I've never done any such testing. Nor it appears have you. And I've yet to see two side by side examples of a slide capture with and without Pixel Shifting. But none the less, if you take just a few seconds to examine actual side by side samples with and without Pixel Shifting of those subjects, there's no question there's a difference and benefit. Do you disagree with the actual examples in the URLs, the two captures are absolutely NOT identical and one is superior?

Understand now? :eek:

Wait, the facts are the facts about how the capture is produced with respect to true RGB data. I NEVER said it was better for use in capturing slides. I said the OP should do his own testing to see. But clearly, the examples provided in TWO URLs I provided DO show there IS a benefit to some captures with Pixel Shifting; you (well I) can clearly see this. Does this mean for capturing slides it's a benefit? Well I've never done any such testing. Nor it appears have you. And I've yet to see two side by side examples of a slide capture with and without Pixel Shifting. But none the less, if you take just a few seconds to examine actual side by side samples with and without Pixel Shifting of those subjects, there's no question there's a difference and benefit. Do you disagree with the actual examples in the URLs, the two captures are absolutely NOT identical and one is superior?

Understand now? :eek:

JFC. This is a discussion thread about film and slide copying.

JFC. This is a discussion thread about film and slide copying.

No, it's also about a capture technology called "pixel shifting" as asked, a subject you haven't tried and call mumbo jumbo and a "trick".