A few posts ago, I started presenting MTF data in terms of cycles per picture height. instead of my previous units of cycles per pixel, which required the introduction of an explicit correction factor when comparing cameras of different resolution.
This morning, I received a message from a reader containing an interesting question and a perspicacious observation, both of which I think merit some discussion.
First, the question:
In your last post you used the same scale (cycles/picture height) for both horizontal and vertical resolution. And you did not mention any H:V scaling factor – 1.5 I presume.
The way I look at it, the picture height, measured in pixels, doesn’t change when you consider horizontal and vertical edges. The camera was in landscape orientation, so the picture height for both measurements was the number of pixels along the short side of the image.
In mentioning the aspect ratio, the reader has found the soft underbelly of the supposedly-impartial cycles/picture height measurement. In landscape orientation, it favors cameras with squarer proportions, like the MF standard of 4:3, over cameras with more panoramic images, like the 35mm-derived 3:2, or HDTV’s 16:9. Fortunately, so far I’ve been dealing with cameras all of which have an aspect ratio of 3:2, so I haven’t had to deal with that, but I’m sure I will at some time.
The obvious way to do so is to come up with a new measurement unit, cycles per picture diagonal. I’m tempted, but such a measurement loses the advantage that drew me to cycles per picture height in the first place, the fact that it is a standard unit.
Which brings me to another point that my reader got me thinking about. Should the standard measure change just because we hold the camera up to our eye in a different orientation? I think not. I believe the unit comes from the television world, who, in the analog era, measured resolution in lines per picture height. They didn’t have to deal with the definition of height: it was always the short direction. Therefore, I’m gong to sitck with cycles per picture height, but what I really mean is cycles per pixel times the number of pixels in the short direction. Doesn’t flow trippingly off the tongue, does it?
Now to my reader’s observation:
By the way, when switching to cycles/picture height scale couple of days ago you did not remove the correction factor, you just hid it: multiplying the a7 MTF50 in cycles/pixel by 4000, and the a7R MTF50 in cycles/pixel by 4912 keeps proportion at 1,228.
You are entirely correct. I never said my idea for correction was wrong, just that it was confusing. And this way, when people come after me with their knives and pitchforks for manipulating the data, I can take cover in the fact that I am manipulating the data in an entirely standard and long-accepted manner.