bwcolor
Veteran
I placed this in this forum because the topic may be relevant to your purchasing decision with regards to the NEX-7.
Mike Davis @ Dpreview Forums posted data regarding diffraction and Bayer Sensor Pixel Density. This data was presented as a graph and as an Excel spreadsheet.
I downloaded the spreadsheet that he kindly linked and used the Nex 3 sensors pixel density and upscaled it to the NEX-7 density and came up with 339 Bayer Pixel Density/mm. The Contax TVS Digital comes in at 359, so using that row, 5lines/mm for an 8x10 print is diffraction limited at around f/4.2 So, in order to minimize diffraction and maximize image quality, we need optics that are optimized for f/4.2 and wider optics. Compare this to the same column for the Fujifilm X100, which is limited to f/13.7. Of course, we are talking a lower resolution sensor and we assume that both have equivalent AA filters. Also, the number calculated in the spreadsheet is not an absolute number beyond which image quality is poor. It is simply a standard set. It was suggested that this standard could have been set as low as 4 lines/mm to as much as 8 lines/mm.
You can find the spreadsheet here
The Graph Here
and the thread here
I thought that there might be a good bit of expertise on this board that might be able to address this concern. Of course, reviews and image samples will ultimately tell the story.
Mike Davis @ Dpreview Forums posted data regarding diffraction and Bayer Sensor Pixel Density. This data was presented as a graph and as an Excel spreadsheet.
I downloaded the spreadsheet that he kindly linked and used the Nex 3 sensors pixel density and upscaled it to the NEX-7 density and came up with 339 Bayer Pixel Density/mm. The Contax TVS Digital comes in at 359, so using that row, 5lines/mm for an 8x10 print is diffraction limited at around f/4.2 So, in order to minimize diffraction and maximize image quality, we need optics that are optimized for f/4.2 and wider optics. Compare this to the same column for the Fujifilm X100, which is limited to f/13.7. Of course, we are talking a lower resolution sensor and we assume that both have equivalent AA filters. Also, the number calculated in the spreadsheet is not an absolute number beyond which image quality is poor. It is simply a standard set. It was suggested that this standard could have been set as low as 4 lines/mm to as much as 8 lines/mm.
You can find the spreadsheet here
The Graph Here
and the thread here
I thought that there might be a good bit of expertise on this board that might be able to address this concern. Of course, reviews and image samples will ultimately tell the story.
douglasf13
Well-known
It doesn't really matter. It'll still have similar resolution to a lower MP sensor at f5.6 and above, and it'll have more resolution at f4 and below. There is still a net gain in resolution.
I feel dumb.
bwcolor
Veteran
Good point, but careful choice of lenses is required to make this work. We are talking about a small lens that is optimized for apertures greater than f/4-5.6. Telephoto lenses, such as the Zeiss 100-300mm used yesterday for my kids soccer now require high ISO, given the same sensor technology, yielding decreasing returns with greater pixel density. Don't know what this means outside of this discussion, but it probably does mean that we will not be seeing 40Mpix APS-C sensors.
BTW... "Good point" was not in reference to "I feel dumb."
BTW... "Good point" was not in reference to "I feel dumb."
Last edited:
ampguy
Veteran
another point is that these calculations assume your aperture opening is round. If it is not, e.g. hexagonal, etc., you are diffraction limited more, and need to open up more.
Also, depends on color wavelength, reds will diffract sooner than other colors.
Also, depends on color wavelength, reds will diffract sooner than other colors.
squirrel$$$bandit
Veteran
I think the time to start worrying about such things is when the camera comes out and the images suck. Until that time, I intend to assume that it is a magic box of super sexy lovin'.
Field
Well-known
Do you think this is a design flaw, limitation, or intentional thing to encourage people to buy their DSLRs?
bugmenot
Well-known
Your math is wrong.
1 mm/(23.5 mm/6000 pixels) = 255.319 pixels
http://www.centplay.com/affiliate/id_366/
1 mm/(23.5 mm/6000 pixels) = 255.319 pixels
http://www.centplay.com/affiliate/id_366/
John Robertson
Well-known
Nokton48
Veteran
When the camera is finally available, I'd like to see some actual photographs from it.
aad
Not so new now.
I am amazed-are these cameras even available? And people would make a decision of some sort on this type of thing?
bwcolor
Veteran
I used the spreadsheet to look up the Bayer pixel density of the NEX-5/NEX-3 and it was reported to be 196. These are 14.03 MPix and I believe that the sensor size is APS-C. So, I came up with 339 by this formula: 196 x 24.3/14.03
Field
Well-known
OK mate!!
I haven't a clue what they are talking about either!!
If there are too few pixels then when the image gets to big the pixel size will have to increase, eventually creating a visible grid pattern.
Diffraction is an issue because... ok so if the light coming is a lot of information say at f16, but there is not enough surface space on the censor to capture all of it so you end up losing some "sharpness". Does that make sense? That is why Medium format and Large format can clamp down f-stops farther, because they can hold more information. They suffer from diffraction but it is less noticeable.
If diffraction is too strong the image over you will end up with so much repeat information that the resolution of the image will dramatically drop, and limit how big it can get before it starts to create a grid look (think Nintendo mario bros). It is a bit more complicated than that, but I think that answer will work to help you guys out.
Granted maybe I do not understand digital as well as I think I do.
bwcolor
Veteran
It is an attempt at self-education in order to make a purchasing decision regarding a compact interchangeable lens camera for my 'M' and perhaps Contax glass. I'm trying to decide if I just want to wait for a new 'full size' sensor camera in a smallish body.
Of course, the reviews will be a major part of my decision. DxOMark reported that my favorite lens, the Zeiss 35mm f/2.8 compact didn't shine when mounted on one of the earlier NEX cameras, so I've been on the lookout for threads that are relevant to pixel density, AA filters, CMOS vs. CCD and lens quality issues. I've also learned that a reduced AA filter can be used with high pixel sensors without worrying about moire. So, it is a complicated affair.
Last edited:
bugmenot
Well-known
Megapixels is a square value. Pixel density is not.
"For instance, a 100 x 100 pixel image that is printed in a 1-inch square could be said to have 100 dots per inch (DPI)"
in your spreadsheet you can calculate Pixel density this way
"Bayer Sensor Width in Pixels" / "Sensor Width (mm)"
for example for P65+(number two in spreadsheet) : 8984/54.9 = 163.643
number differs because they averaging it for height in case pixels aren't square enough.
Btw all this "Stop (f/N) at Which Diffraction Prevents 5 lp/mm After Enlargement" is completely pointless.
douglasf13
Well-known
Let me try to make it a little simpler and explain why diffraction should have no real bearing in deciding on a camera. Say you're looking at a camera that is 12mp vs. a camera that is 24mp, and, for the sake of argument, we'll just say that diffraction sets in on the 12mp camera at f8, and it sets in at f5.6 on a 24mp camera. All the 24mp camera is doing is resolving the diffraction better, and both cameras may end up having about the same resolution at f8 or more, so it's a draw. However, below f4, you'll still have more resolution with the 24mp camera, so it has the resolution advantage overall. All the diffraction is doing is keeping the 24mp camera from beating the resolution of the 12mp at every aperture, but it still beats it in many of the apertures. This is simplifying things a bit, but does that make sense?
The ZM 35/2.8 is my most used lens on the NEX-5, and it is still a good performer. The NEX-C3 has been shown by many to have improved sensor corners, so I expect the ZM 35/2.8 to be even better with the newer NEX cameras.
Really, as far as rendering detail is concerned, I'd be more concerned about a good tripod and raw converter choice, rather than worrying about diffraction. You'd be amazed how much more detail that a floating bit raw processor, like RPP, resolves compared to the run of the mill converters like LR3 (or worse yet, Sony's own raw program.)
The ZM 35/2.8 is my most used lens on the NEX-5, and it is still a good performer. The NEX-C3 has been shown by many to have improved sensor corners, so I expect the ZM 35/2.8 to be even better with the newer NEX cameras.
Really, as far as rendering detail is concerned, I'd be more concerned about a good tripod and raw converter choice, rather than worrying about diffraction. You'd be amazed how much more detail that a floating bit raw processor, like RPP, resolves compared to the run of the mill converters like LR3 (or worse yet, Sony's own raw program.)
Last edited:
umcelinho
Marcelo
i hate math.
bugmenot
Well-known
bwcolor
Veteran
OK, great input and explanation.... thanks. I thought that I could get a better set of answers on this board. So, it seems that sensor noise/ISO performance and peripheral vignetting are the only serious negatives with these higher density sensors. Of course, Sony's NEX sensor design is built around providing a lower noise floor than previous technologies for a sensor of this size.
gavinlg
Veteran
DXO is rubbish. They report the zeiss distaon ZE 21mm f2.8 as one of the lowest resolving lenses they've tested for canon EOS (when in fact it's probably one of - if not THE highest resolving lens available for EOS). Same goes with other zeiss's tested...
They're just plain wrong.