Size matters ??

Zack Arias's take on crop vs full-frame:

Crop or Crap :: Math or Moment

Just thought it could be of interest to those of you who could be down-sizing you camera load.

It's kind of an absurd video, watched it the other day.  APS-C is only 44% of a full frame in area terms...

Overall I get what he is trying to do and remove some of the angst of different sensor sizes but he does it by basically insulting our intelligence.

If you extend his logic, size of the sensor doesn't matter at all so my iPhone should be just as good as a Phase One IQ series

I don't think he's saying there're no differences. I think he's just saying if the quality is good enough for your photographs.

Wow it was like I just watched a Fuji advertisement. lol, Wonder why he chose the big Nikon full frame? and not a Sony A7r full frame, Ide like to see those two cameras side by side, for a size comparison.

This just adds to the confusion about sensor sizes. A major part of the confusion is thinking constant f/ratios are equivalent. They are not. The only equivalence is light density (photons per square micron per second). The difference in depth of field is due to aperture diameter, not f/ratio. One tends to use a longer focal length lens on the larger sensor to get the same field of view. Then people think the f/2.8 lens on the 2x crop is the same as the f/2.8 lens on the full frame; they are not. The f/2.8 lens on the full frame, for the same field of view and the 2x crop, has twice the focal length, twice the lens diameter and collects 4x the light. The cause of the difference people observe in the images is mainly due to the lenses used, not the sensor size.

Roger

Here's another video:

Crop Factor with ISO & Aperture: How Sony, Olympus, Panasonic, Canon, Nikon & Fuji Cheat You

More technical. In case some of you want to be confused more

Tony Northrup's responses to critics:

Crop Factor Part 3: Responding to Critics, Corrections


A bit about pixel density and if f2.8 is f2.8 is f2.8.

Roger is correct.

I also agree that the right camera and sensor size is an individual thing.  When making the decision based on knowledge, there really isn't a wrong decision since you are aware of the compromises for each sensor size.  Or perhaps the decision is a financial one for somebody.

While the original video posted in this thread attempt to make that point, I think it does it badly and confuses the issue more than helps by misleading the viewer.  That's basically how I interpreted it but others certainly may get the original intent.

I watched the Tony videos, and sent him this response:

Hello Tony,

First I want to compliment you on your videos and efforts to educate people about photography. The series on ISO, Aperture and Focal Length, https://www.youtube.com/watch?v=DtDotqLx6nA
is well done, though perhaps too complex for most photographers to remember in the field (for example, to square the crop factor or not to square it when applying a correction).

I agree with your math as presented, but it uses a basic assumption that is not always valid (assumption: ISO is tied to exposure time) and you barely talk about exposure time, the real underlying key factor.

Let me give an example of where your presentation and theory falls apart. At about 11 minutes in, you talk about
ISO 1600 on a full frame getting the same total light as ISO 400 on a 2x crop camera. ISO does not change the sensitivity nor the amount of light collected; exposure time does. So your model works if you are using your camera's light meter to set the exposure. By decreasing ISO in a metered situation, the camera exposure time lengthens. It is the longer exposure time, not ISO that controls how much light the camera records.

Digital camera sensors have only one sensitivity; ISO is simply a change in post sensor (and post light collection) gain to boost signals for the A/D conversion.

Here is where your model fails. Say two guys go out to do some night photography. One with a full frame camera and a 20 mm f/2.8 lens, and the other with a 2x crop camera and a 10 mm f/2.8 lens. They decide on doing 20 second exposures to keep stars from trailing too much. The guy with the full frame camera sets his ISO to 1600,
and so the guy with the crop camera, having watched your video sets his camera to ISO 400 thinking he will get the same total light. No he doesn't. With 20 second exposures, he gets the same amount of light regardless of where he sets the ISO, and only 1/4 the light that the guy with the full frame camera gets, because his lens aperture area is only 1/4 that of the lens on the full frame camera.

ISO does not change sensitivity nor the amount of light collected; only the exposure time and lens aperture area (not f/ratio) control the amount of light collected. ISO just changes the range of light that gets digitized.

So to simplify all your math, all one needs to do to equalize the real total amount of light collected is to equalize the lens area times exposure, or slightly simpler: lens aperture diameter squared times exposure time.
ISO is irrelevant (except for handling downstream electronics noise and making sure the highest signal one wants to digitize falls withing range of the electronics).

Similarly, using f/ratio and scaling factors for different crop cameras is confusing. It is more simple: forget f/ratios and just look at the lens aperture diameter. When you have the same field of view on two cameras, the depth of field will be the same when the lens aperture diameters are the same and you produce the same size
final image.

For more information on what ISO really is, and true exposure, please see my series on ISO and exposure:
http://www.clarkvision.com/articles/index.html#exposure

Sincerely,
Roger Clark

Roger, your reply is probably too complicated for my understanding but iso in digital cameras is linear from the base to a certain sensitivity (Sony sensor typically around a 1000 iso) and any iso over that is just boosted signal and corresponding SW to manage noise. People use to film iso #s have a tough time relating to digital iso which by its very nature is just brightening from the sensor's base iso.

Mike in O wrote:Roger, your reply is probably too complicated for my understanding but iso in digital cameras is linear from the base to a certain sensitivity (Sony sensor typically around a 1000 iso) and any iso over that is just boosted signal and corresponding SW to manage noise.  People use to film iso #s have a tough time relating to digital iso which by its very nature is just brightening from the sensor's base iso.

Hello Mike,

Sensitivity is the wrong term to use.  Digital sensors have one one sensitivity, given by the Quantum Efficiency of the device.  ISO changes the gain of the signal read off the sensor.  Changing ISO does not change the amount of light collected, thus not the sensitivity.  I don't know about Sony sensors, but Canon and Nikon's latest cameras do not do a software multiplication of the data to increase ISO until at least several thousands of ISO.  Canon does not appear to do any software filtering to manage noise in the raw files.  The noise is dominated by photon statistics for all image data were a few/few tens of photons are collected in a pixel.

Roger

I would imagine Sony does the same thing since they use essentially the same sensors as Nikon cameras. But then again Sony does weird $#!+ that is inexcusable to the RAW files

Here are some graphs showing the switch analogue to digital (sw dependent); as you can see with the 1dx sensor, the switch happens about iso 800.
http://www.dpreview.com/forums/post/52755911

Mike in O wrote:Here are some graphs showing the switch analogue to digital (sw dependent); as you can see with the 1dx sensor, the switch happens about iso 800.
http://www.dpreview.com/forums/post/52755911

There are some misconceptions on that page.  Side note, bobn2 has the sensorgen web site where he references my site for methodology, but bobn2 is not using my methodology and I disagree that he can derive quantum efficiency and some of the other parameters from DXOMark data (perhaps relative numbers, but not the absolute values he gives).

The flattening of apparent read noise has nothing to do with digital scaling of the data.  "Apparent read noise" is a combination is sensor read noise and noise from downstream electronics.  As gain (ISO) increases, the signal from the sensor, including sensor read noise is boosted above downstream electronics noise.  At high ISOs, (typically around ISO 1600 and higher on recent Canon cameras) that gain is high enough that downstream electronics noice is relatively insignificant.  That does not mean above that value, digital scaling is happening.

To determine if there is digital scaling, one needs to look at the histogram of raw linear data from the sensor (before any demosaicking) and see if there are regular gaps.  I have not seen such gaps in Canon raw data.  Nikon raw data do have gaps, though strange and not regular like one might expect from digital scaling up the data.

Roger

The easy way to test isoless sensor is to take a picture at base iso, under expose and try to raise the shadows until they show degradation, you will find Canon sensors do not compete with the sony ones.  Most camera are using Sony sensors now because of this analog difference.  This is the basis of DR which everyone knows is superior in the Sonys.  The point of analog to digital can easily be measured.

E.J. Peiker wrote:I would imagine Sony does the same thing since they use essentially the same sensors as Nikon cameras.  But then again Sony does weird $#!+ that is inexcusable to the RAW files

What weird things? Just curious.

Markus

They use a lossy compression algorithm. Even though the sensor is 14 bits, after their compression you only get about 11 bits of data per channel per pixel. Here's an article on it. Why they do this is beyond comprehension and you can't turn it off:
http://diglloyd.com/blog/2014/20140212_ ... ation.html

E.J. Peiker wrote:They use a lossy compression algorithm.  Even though the sensor is 14 bits, after their compression you only get about 11 bits of data per channel per pixel.  Here's an article on it.  Why they do this is beyond comprehension and you can't turn it off:
http://diglloyd.com/blog/2014/20140212_ ... ation.html

The newer cameras use the compression, but my A900 has the choice and I can't tell the difference.

E.J. Peiker wrote:They use a lossy compression algorithm.  Even though the sensor is 14 bits, after their compression you only get about 11 bits of data per channel per pixel.  Here's an article on it.  Why they do this is beyond comprehension and you can't turn it off:
http://diglloyd.com/blog/2014/20140212_ ... ation.html

I agree, that makes zero sense. At least one should be able to turn it off.

Markus

Mike in O wrote:

E.J. Peiker wrote:They use a lossy compression algorithm.  Even though the sensor is 14 bits, after their compression you only get about 11 bits of data per channel per pixel.  Here's an article on it.  Why they do this is beyond comprehension and you can't turn it off:
http://diglloyd.com/blog/2014/20140212_ ... ation.html

The newer cameras use the compression, but my A900 has the choice and I can't tell the difference.

In many cases you wouldn't but it is very easy to construct photos that will have a big problem, like star trails or high contrast fine edges at night.