Graeme Nattress
RED Problem Solver
The slanted edge approach leads to silly results though. We all know you can't have more resolution than a sensor has pixels, but the numbers calculated by that approach result in such numbers as it doesn't distinguish between real resolution and aliases. Jack posted a similar response to Adam Wilt on the PVC article: "The method allows spatial frequency response above Nyquist to be measured, although such frequencies will alias".
The two issues there are the issue of counting aliasing as resolution, which I think we can all agree is a wrongheaded approach, and the 2nd, which is in some ways more serious, which is that figures are presented to an audience without adequate explanation of what they mean, and it's only later when questioned by a careful observer like Adam that any kind of answer is given at all.
And if it's MTF that's being measured - why is the chart labeled "maximum resolution", and why is not an MTF plot provided instead with the area that represents aliasing shown? Even with with the bar graph used, aliasing could have been noted in another colour on the bar to show this, but it is not.
And remember here - I'm complaining about a test that rates the resolution of a RED One much higher than I've measured and quoted to customers. We neither want nor need to show resolution figures greater than we deem possible with our camera systems as that helps nobody, and gives false expectations. So what happens when someone goes out and shoots something and they can't get the resolution quoted - who takes the flak - is it me, or you? If you can sense my frustration here, you're right - I'm frustrated by this.
"so I can't tell you in simple terms why this was the preferred method. " - but I can tell you simply why it's NOT a preferred method :
it produces overly large numbers not representative of what the camera system can practically achieve,
is prone to produce "silly" numbers where the resolution is greater than the sensor pixel count,
counts aliasing as resolution,
and very hard to understand! It's a computational approach rather than a direct measuring approach. With a direct measure approach we can all "see" the answer quite clearly from a good chart, whereas with a computational approach you rely on a lot of math which is not always easy to follow.
Graeme
The two issues there are the issue of counting aliasing as resolution, which I think we can all agree is a wrongheaded approach, and the 2nd, which is in some ways more serious, which is that figures are presented to an audience without adequate explanation of what they mean, and it's only later when questioned by a careful observer like Adam that any kind of answer is given at all.
And if it's MTF that's being measured - why is the chart labeled "maximum resolution", and why is not an MTF plot provided instead with the area that represents aliasing shown? Even with with the bar graph used, aliasing could have been noted in another colour on the bar to show this, but it is not.
And remember here - I'm complaining about a test that rates the resolution of a RED One much higher than I've measured and quoted to customers. We neither want nor need to show resolution figures greater than we deem possible with our camera systems as that helps nobody, and gives false expectations. So what happens when someone goes out and shoots something and they can't get the resolution quoted - who takes the flak - is it me, or you? If you can sense my frustration here, you're right - I'm frustrated by this.
"so I can't tell you in simple terms why this was the preferred method. " - but I can tell you simply why it's NOT a preferred method :
it produces overly large numbers not representative of what the camera system can practically achieve,
is prone to produce "silly" numbers where the resolution is greater than the sensor pixel count,
counts aliasing as resolution,
and very hard to understand! It's a computational approach rather than a direct measuring approach. With a direct measure approach we can all "see" the answer quite clearly from a good chart, whereas with a computational approach you rely on a lot of math which is not always easy to follow.
Graeme
