Meryl Arbing
Veteran Member
Wait for other reviews before jumoping to conclusions.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
V3 optimal aperture setting at Program mode?
- Thread starter hi161514
- Start date
stan ames
Active member
This is one of the most interesting threads I have seen for a long time and I’m grateful to all who have added to it, but I’m disappointed by those who wish to stifle discussion by saying wait for more reviews - people should be encouraged exchange ideas and think for themselves.
From the excellent data presented by Mike Davis it’s possible to plot the effect of pixel density on the aperture at which image softness due to diffraction will begin to appear. The latest 7Mp cameras don’t feature in the database but from the data on dpreview I calculated the pixel density as 427 pixels per mm. When added to the plot it is possible to extrapolate the limiting aperture to a little less than f5. It’s easy to see why this discussion has evolved.
The data provided below is for the purpose of forum discussion only and not guaranteed to be accurate! I trust that Mike will let me know if I have made any errors.
Stan
Live, learn & pass it on
From the excellent data presented by Mike Davis it’s possible to plot the effect of pixel density on the aperture at which image softness due to diffraction will begin to appear. The latest 7Mp cameras don’t feature in the database but from the data on dpreview I calculated the pixel density as 427 pixels per mm. When added to the plot it is possible to extrapolate the limiting aperture to a little less than f5. It’s easy to see why this discussion has evolved.
The data provided below is for the purpose of forum discussion only and not guaranteed to be accurate! I trust that Mike will let me know if I have made any errors.
Stan
Live, learn & pass it on
Unless Dcresource had a defective camera its pretty evident that this is a serious setback for the V3. However, I will wait for Phils thorough review to confirm this.
Mike Davis
Leading Member
Hi Stan!
Your chart is not only sound (mathematically), it beautifully illustrates the problem. Thanks!
Manufacturers will continue to increase pixel counts without simultaneously increasing sensor size until they've found that point at which the average digicam consumer is actually noticing his images are degraded at all but the widest apertures. More discriminating consumers will stay clear of high density sensors despite the tolerance exhibited by the masses.
Softness caused by visible diffraction is somewhat insidious in that it affects the entire image uniformly (ignoring for the moment any discussion of how Airy disk diameters vary with subject color - which is actually neglible, anyway). Unlike insufficient depth of field, which readily reveals itself to even an untrained eye (as foregrounds or backgrounds that are visibly softer than the plane of sharpest focus), diffraction's degradation, again, uniform across the entire image, is not easily detected by the inexperienced until you make side-by-side comparisons of the same image, one taken at a "safe" aperture and the other at an aperture that forces diffraction's Airy disks to become visible.
Ironically, legions of photographers, in their diffraction-ignorant quest for depth of field, routinely make the mistake of using apertures that actually cause diffraction's Airy disk diameters throughout the entire image to exceed the Circle of Confusioin diameters they are striving to shrink at the Near and Far sharps of the subject space! The images of pinhole cameras are an extreme example of what happens when you have lots of DoF (tiny Cirlces of Confusion) at the expense of lots of diffraction (huge Airy disks).
If Joe Consumer doesn't start rejecting these cameras with high density sensors soon, the manufacturers might as well just get rid of the glass lenses altogether. Pinhole digicams might be all we need.
Actually, there is an alternate solution - but equally expensive for the manufacturers: Give us faster lenses along with the higher densities. I would have no objection to using a sensor that's diffraction limited to stopping down no further than f/4 if the lenses available for that camera simultaneously offered maximum apertures as wide as f/1.0. That would at least give us a five-stop range to work with (f/1.0, f/1.4, f/2.0, f/2.8, and f/4).
Mike Davis
--
http://www.accessz.com
Your chart is not only sound (mathematically), it beautifully illustrates the problem. Thanks!
Manufacturers will continue to increase pixel counts without simultaneously increasing sensor size until they've found that point at which the average digicam consumer is actually noticing his images are degraded at all but the widest apertures. More discriminating consumers will stay clear of high density sensors despite the tolerance exhibited by the masses.
Softness caused by visible diffraction is somewhat insidious in that it affects the entire image uniformly (ignoring for the moment any discussion of how Airy disk diameters vary with subject color - which is actually neglible, anyway). Unlike insufficient depth of field, which readily reveals itself to even an untrained eye (as foregrounds or backgrounds that are visibly softer than the plane of sharpest focus), diffraction's degradation, again, uniform across the entire image, is not easily detected by the inexperienced until you make side-by-side comparisons of the same image, one taken at a "safe" aperture and the other at an aperture that forces diffraction's Airy disks to become visible.
Ironically, legions of photographers, in their diffraction-ignorant quest for depth of field, routinely make the mistake of using apertures that actually cause diffraction's Airy disk diameters throughout the entire image to exceed the Circle of Confusioin diameters they are striving to shrink at the Near and Far sharps of the subject space! The images of pinhole cameras are an extreme example of what happens when you have lots of DoF (tiny Cirlces of Confusion) at the expense of lots of diffraction (huge Airy disks).
If Joe Consumer doesn't start rejecting these cameras with high density sensors soon, the manufacturers might as well just get rid of the glass lenses altogether. Pinhole digicams might be all we need.
Actually, there is an alternate solution - but equally expensive for the manufacturers: Give us faster lenses along with the higher densities. I would have no objection to using a sensor that's diffraction limited to stopping down no further than f/4 if the lenses available for that camera simultaneously offered maximum apertures as wide as f/1.0. That would at least give us a five-stop range to work with (f/1.0, f/1.4, f/2.0, f/2.8, and f/4).
Mike Davis
--
http://www.accessz.com
Meryl Arbing
Veteran Member
I have yet to see the effect on my camera and I have taken several series of shots at all available apertures without noticible differences.
Of course, you need to make sure you have the camera on a tripod just to account for camera shake. As aperture gets smaller, shutter speed gets slower. Numbers aside, people may be seeing nothing more than this simple photographic reality.
Of course, you need to make sure you have the camera on a tripod just to account for camera shake. As aperture gets smaller, shutter speed gets slower. Numbers aside, people may be seeing nothing more than this simple photographic reality.
jsonic
Veteran Member
OK, guys, I think I kind of follow this. I always had the feeling that smaller sensors with more pixels was a bad idea, but can I assume in the chart that the dots towards the left are all SLRs and the ones towards the lower-right are all the consumer digicams? Am I misinterpreting the chart to say that if I buy a non-SLR (with a good number of pixels), the max f-stop I should use is probably anywhere from 4.8 to 6, depending on brand/model? It would be interesting to plot my camera (3 mp), since it probably has fewer pixels/mm.; there might not be much loss just due to the limited resolution! 
From Mike's message, this is what I gathered: manufacturers allow an aperture of 8, knowing that it will create an inferior photo as far as detail is concerned, in that they believe users will prefer a great depth-of-field where the entire photo is equally blurred. Um, I'm not sure I get it, but maybe there is a tradeoff there. If I'm taking snapshots, a slight blur really isn't going to be noticable at 4x6 (or even 5x7). Still, the idea of guaranteeing blur just seems wrong!
Why not just stop it (in the internal firmware) at, say, f7 or f6 and not even bother with f8? (I assume that the effect would be much less noticable at, say, f6.) Why allow the camera to occasionally decide to use f8? I guess I can see where you might allow a user to do whatever he wants if he insists (in Aperture-priority mode).
My only concern would be that the camera NOT choose f8 often while in one of the automatic modes! It sounds like this would be the exception, so I'm not too worried about it; I look forward to seeing the V3 in person. (One of these days...)
--
Gary W.
From Mike's message, this is what I gathered: manufacturers allow an aperture of 8, knowing that it will create an inferior photo as far as detail is concerned, in that they believe users will prefer a great depth-of-field where the entire photo is equally blurred. Um, I'm not sure I get it, but maybe there is a tradeoff there. If I'm taking snapshots, a slight blur really isn't going to be noticable at 4x6 (or even 5x7). Still, the idea of guaranteeing blur just seems wrong!
Why not just stop it (in the internal firmware) at, say, f7 or f6 and not even bother with f8? (I assume that the effect would be much less noticable at, say, f6.) Why allow the camera to occasionally decide to use f8? I guess I can see where you might allow a user to do whatever he wants if he insists (in Aperture-priority mode).
My only concern would be that the camera NOT choose f8 often while in one of the automatic modes! It sounds like this would be the exception, so I'm not too worried about it; I look forward to seeing the V3 in person. (One of these days...)
--
Gary W.
Mike Davis
Leading Member
Hi Gary,
It's important to note that the spreadsheet compares each of the sensors listed there in terms of the largest 300 dpi prints they are capable of producing. The dimensions obtainable at 300 dpi with each camera are provided in the sheet.
Many digicams that would produce visible diffraction at f/8 in their largest possible 300 dpi prints would not suffer visible diffraction at lesser enlargement factors. Unfortunately, these cameras doesn't ask you what size print you intend to make before the aperture is selected for each exposure.
If all one intends to produce is 4x5-inch snapshots, f/8 will not induce visible diffraction with most digicams. But a 300 dpi 4x5-inch snapshot can be had with a sensor resolution of only 1200 x 1500. What are all those extra pixels for in the latest rash of 7.1 MP digicams? They are for prints larger than 4x5, of course - but you'd better stick to shooting wide open (at f/4.1 or wider) if you really want to prevent diffraction from corrupting detail in a 300 dpi print made with all of those pixels (speaking of cameras like the Sony V3 - with its outrageous pixel density of 430 pixels/mm.)
Pixel Density is the enemy here only because the manufacturers haven't given us faster lenses to go with their cheap, high density sensors. Having lots of pixels encourages us to make large prints and many people don't stop at the 300 dpi dimensions given in the spreadsheet - they go to even greater enlargement factors with image densities as low as 150 dpi or worse.
Think of diffraction's Airy disks as being projected on to the sensor. Anything in the subject space that's smaller than an Airy disk will simply not be recorded. For example, a model standing might be holding a nickel such that it is facing the camera. It should be easy to imagine that at some point, as diffraction's Airy disks are allowed to get larger, the abillity to read the date on that coin will be lost. No amount of sharpening or rezziing up can recover that lost detail. The final print may appear sharp, but the coin could be devoid of detail just the same. The more you enlarge Airy disks to get to your final print dimensions, the more likely they will be visible in the final print. A large, low density 8 MP sensor suffers less enlargement than a smaller, high density 8 MP sensor, and thus, the larger sensor's prints will have smaller (hopefully invisible) Airy disks for a given f-stop.
Mike Davis
--
http://www.accessz.com
It's important to note that the spreadsheet compares each of the sensors listed there in terms of the largest 300 dpi prints they are capable of producing. The dimensions obtainable at 300 dpi with each camera are provided in the sheet.
Many digicams that would produce visible diffraction at f/8 in their largest possible 300 dpi prints would not suffer visible diffraction at lesser enlargement factors. Unfortunately, these cameras doesn't ask you what size print you intend to make before the aperture is selected for each exposure.
If all one intends to produce is 4x5-inch snapshots, f/8 will not induce visible diffraction with most digicams. But a 300 dpi 4x5-inch snapshot can be had with a sensor resolution of only 1200 x 1500. What are all those extra pixels for in the latest rash of 7.1 MP digicams? They are for prints larger than 4x5, of course - but you'd better stick to shooting wide open (at f/4.1 or wider) if you really want to prevent diffraction from corrupting detail in a 300 dpi print made with all of those pixels (speaking of cameras like the Sony V3 - with its outrageous pixel density of 430 pixels/mm.)
Pixel Density is the enemy here only because the manufacturers haven't given us faster lenses to go with their cheap, high density sensors. Having lots of pixels encourages us to make large prints and many people don't stop at the 300 dpi dimensions given in the spreadsheet - they go to even greater enlargement factors with image densities as low as 150 dpi or worse.
Think of diffraction's Airy disks as being projected on to the sensor. Anything in the subject space that's smaller than an Airy disk will simply not be recorded. For example, a model standing might be holding a nickel such that it is facing the camera. It should be easy to imagine that at some point, as diffraction's Airy disks are allowed to get larger, the abillity to read the date on that coin will be lost. No amount of sharpening or rezziing up can recover that lost detail. The final print may appear sharp, but the coin could be devoid of detail just the same. The more you enlarge Airy disks to get to your final print dimensions, the more likely they will be visible in the final print. A large, low density 8 MP sensor suffers less enlargement than a smaller, high density 8 MP sensor, and thus, the larger sensor's prints will have smaller (hopefully invisible) Airy disks for a given f-stop.
Mike Davis
--
http://www.accessz.com
Meryl Arbing
Veteran Member
I think you have done a lot of work on your mathematical model but I don't see what size CoC you are assuming for the different sized digital sensors.
We know that, in the film world, Medium format Coc is defined as 1/1000th of the 'normal' focal length but in 35mm it is 1/1500 of a 'normal' 50mm lens.
A CCD, such as the one on the Sony V3 has a diagonal measurement of only 9mm and, as such, the CoC for a sensor of this size would need to be significantly smaller than a 35mm film camera. I would expect it to be closer to 1/2000th of its 10mm 'normal' focal length (not converted to 35mm equivalents of course).
Since this has a direct effect on perceived sharpness and DOF concepts, it should be a variable that is accounted for in the model.
We know that, in the film world, Medium format Coc is defined as 1/1000th of the 'normal' focal length but in 35mm it is 1/1500 of a 'normal' 50mm lens.
A CCD, such as the one on the Sony V3 has a diagonal measurement of only 9mm and, as such, the CoC for a sensor of this size would need to be significantly smaller than a 35mm film camera. I would expect it to be closer to 1/2000th of its 10mm 'normal' focal length (not converted to 35mm equivalents of course).
Since this has a direct effect on perceived sharpness and DOF concepts, it should be a variable that is accounted for in the model.
Oh yeah, I like the night frame and night shot. The new thing they added too. Where you can zoom in on your picture when you view it. Wow. So people stop beating around the bush. Just get one. You'll be amazed. Also, that guy who took them pictures at that other site. He sure can't take sharp photos. Because my photos are sharp. And I'm using the same camera. Well that's about it. For me Sony V3 is great. Way better then dullsville Canon G6. Pictures that come out with no contrast. Ilk. Don't knock it till you try it. Thanks
imbsysop
Veteran Member
you may want to check the data at
http://www.uni-mainz.de/%7Esprec000/Digicam7.html
although in German the data for the Sony P150 are claimed to be useable for both V3 & Canon G6 (same CCD?)
Assuming 3072 pixels along the long side of the CCD this gives a pixel density of 426.666 pix/mm aka a pixel size of 2.3 micron
the COC seems to be defined as 2x the physical pixel size (dunno why)
this brings the COC for the V3 (& others) to .0046
dunno if all this makes sense .. but that seems to be the value to be used in Hyperfocal focussing calculations ..
Have you seen the V3 vs G6 brick wall comparison shot on Dcresource? The difference really hits you in the face. It doesnt leave much room for doubt.
Meryl Arbing
Veteran Member
That seems about right.
My rough calculations estimated the CoC for the V3 (and others) to be about .005mm (5 microns) so .0046 is probably the more accurate figure.
My rough calculations estimated the CoC for the V3 (and others) to be about .005mm (5 microns) so .0046 is probably the more accurate figure.
Mike Davis
Leading Member
Hi Meryl,
It's refreshing to see someone is actually delving into this with me.
The CoC used to calculate DoF for each sensor varies from one sensor to the next. It is the same as the diameter of the Airy disk (twice the Rayleigh radius) at the f-stop where diffraction would prevent a resolution of 5 lp/mm in a 300 dpi print made with that sensor.
An entirely different spreadsheet was used to calculate the Near Sharp distances you see in Column P, but to answer your question, here's how I came up with the CoC for each sensor:
Maximum permissible CoC = 1 / desired resolution / enlargement factor
For every sensor, it is assumed we will be printing a 300 dpi print, which is equivalent to resolution of 5 lp/mm. So our the "desired resolution" in the equation above is a value of 5 for every sensor in the spreadsheet. The enlargment factor varies greatly from one sensor to the next (as you noted). That is available in column N of the spreadsheet.
So, for the Mamiya ZD, for example, the CoC used to calculate a Near Sharp at the f-stop given in column 0 is:
CoC = 1 / 5 / 9.53 = 0.02099
For the Panasonic DMC-FZ20, the enlargement factor necesary to produce its 300 dpi print is a ridiculous 37.63. For this sensor, the CoC used to calcualte a Near Sharp at the f-stop given in column 0 is:
CoC = 1 / 5 / 37.63 = 0.005315
Thanks,
Mike Davis
--
http://www.accessz.com
It's refreshing to see someone is actually delving into this with me.
[snipped several excellent comments I agree with.]
The CoC used to calculate DoF for each sensor varies from one sensor to the next. It is the same as the diameter of the Airy disk (twice the Rayleigh radius) at the f-stop where diffraction would prevent a resolution of 5 lp/mm in a 300 dpi print made with that sensor.
An entirely different spreadsheet was used to calculate the Near Sharp distances you see in Column P, but to answer your question, here's how I came up with the CoC for each sensor:
Maximum permissible CoC = 1 / desired resolution / enlargement factor
For every sensor, it is assumed we will be printing a 300 dpi print, which is equivalent to resolution of 5 lp/mm. So our the "desired resolution" in the equation above is a value of 5 for every sensor in the spreadsheet. The enlargment factor varies greatly from one sensor to the next (as you noted). That is available in column N of the spreadsheet.
So, for the Mamiya ZD, for example, the CoC used to calculate a Near Sharp at the f-stop given in column 0 is:
CoC = 1 / 5 / 9.53 = 0.02099
For the Panasonic DMC-FZ20, the enlargement factor necesary to produce its 300 dpi print is a ridiculous 37.63. For this sensor, the CoC used to calcualte a Near Sharp at the f-stop given in column 0 is:
CoC = 1 / 5 / 37.63 = 0.005315
Thanks,
Mike Davis
--
http://www.accessz.com
stan ames
Active member
Meryl
I have done exactly what you suggest and my pictures show exactly the same effect as on DC resource. There is a perceived critical lack of sharpness at f8. My shots were all from a tripod so we can't simply dismiss it as camera shake.
There's a lot more to this issue than "is the V3 good or bad". Diffraction is at the very heart of current digital camera development. "Numbers aside"? We can't put the numbers aside, they are essential to understanding the problem and they enforce our objectivity.
Stan
ps I hope you are having as much enjoyment with your V3 as I am with mine.
Live,learn & pass it on
I have done exactly what you suggest and my pictures show exactly the same effect as on DC resource. There is a perceived critical lack of sharpness at f8. My shots were all from a tripod so we can't simply dismiss it as camera shake.
There's a lot more to this issue than "is the V3 good or bad". Diffraction is at the very heart of current digital camera development. "Numbers aside"? We can't put the numbers aside, they are essential to understanding the problem and they enforce our objectivity.
Stan
ps I hope you are having as much enjoyment with your V3 as I am with mine.
Live,learn & pass it on
Meryl Arbing
Veteran Member
I'll post my series later tonight. No problems seen.
Meryl Arbing
Veteran Member
As many are aware, I dispute the conclusions made in the dcresource review and the subsequent negative assertions based on it.
The model developed by Mike is quite impressive but the value of any model is its ability to predict performance. The test for the model must be empirical observation and then, if the predictions are confirmed, the model is accurate. But if the observations contradict the model then it is not the observations that need to be changed.
I don't deny that diffraction exists, but the degree to which it is an issue for the V3 user is the question.
To this end, I setup a simple scenario.
I mounted the V3 on a tripod, pointed it at a clock on the wall, set to Aperture Priority, used the self-timer and then to a shot at EVERY aperture setting available at both ends of the zoom range (wide angle and tele)
I have reduced the size of the images here (for the sake of those with dial-up connection but the Full size images with full EXIF data intact.
Here are the reduced images:
Wide Angle f2.8
Wide Angle f8
Tele f4
Tele f8
The full size shots, direct from the camera with no alterations are available at:
http://www3.sympatico.ca/marbing/Pictures/
The model developed by Mike is quite impressive but the value of any model is its ability to predict performance. The test for the model must be empirical observation and then, if the predictions are confirmed, the model is accurate. But if the observations contradict the model then it is not the observations that need to be changed.
I don't deny that diffraction exists, but the degree to which it is an issue for the V3 user is the question.
To this end, I setup a simple scenario.
I mounted the V3 on a tripod, pointed it at a clock on the wall, set to Aperture Priority, used the self-timer and then to a shot at EVERY aperture setting available at both ends of the zoom range (wide angle and tele)
I have reduced the size of the images here (for the sake of those with dial-up connection but the Full size images with full EXIF data intact.
Here are the reduced images:
Wide Angle f2.8
Wide Angle f8
Tele f4
Tele f8
The full size shots, direct from the camera with no alterations are available at:
http://www3.sympatico.ca/marbing/Pictures/
Meryl Arbing
Veteran Member
This is a diagonal slice and recombination of two shots. One side is f4 and the other is f8.
Which side is which? If the stated 'problem' exists it should require no effort to distinguish them based on the sharpness.
I don't think it is that apparent and, as such, the so-called aperture 'problem' is NOT significant to the use of the V3 in any circumstance.
Which side is which? If the stated 'problem' exists it should require no effort to distinguish them based on the sharpness.
I don't think it is that apparent and, as such, the so-called aperture 'problem' is NOT significant to the use of the V3 in any circumstance.
Keyboard shortcuts
- f
- Forum