17
« on: April 06, 2014, 02:17:10 PM »
@rainless: This is my understanding of what is known in ML and this forum as "crop mode".
1. Raw video is not created by ML. It is simply captured directly from the already-existing data stream which the camera sends to the LV display in Video mode.
2. Therefore, when you are in, let's call it "normal" mode (you haven't hit the Zoom button yet), you can get, as a maximum, the native video resolution of the LV display. This resolution has increased since the older cameras, so that the 50D, for example, can do "only" 1584x1058 whereas the 5D Mark III can do 1920x1080.
3. This represents the entire image sensor dimensions mapped onto the size of the LV display. For example, the entire 5760x3240 of the 5D Mark III sensor is mapped [via a black-box process we don't completely understand] to the dimensions of the LV display in Video mode (1920x1080 for this camera, less for other cameras.) This mapping is done in the camera, and ML has no control over how it is done. It simply uses the result.
4. BUT, once you press the "Zoom" button, it's another story. Now you go into "crop mode". By zooming into the image we see on the LV display in "normal" Video mode, we see objects larger than we did before. Exactly how much larger varies by camera. But what happens when you're running ML is that now the data stream sent to the LV display can in effect be larger (or smaller) than the native video resolution of the LV display. That's why, in "crop mode", you can record, for example 2000x1080 on the 50D and 3584x1320 on the 5D Mark III. The difference is that now this rectangle you are recording is a 1-to-1 pixel mapping of only a portion of the image sensor, that is, a smaller rectangle which represents only part of the entire image sensor rectangle. The highest resolution you can record in "crop mode" depends on the camera model you have and is probably a hardware limitation in most cases.
5. The display, on the other hand, which you see when you are in "crop mode" is selectable. It is also ONLY a display, it is not what you are recording. Basically you have the choice between ML grayscale, which shows an accurate representation of the boundaries of the rectangle you are recording, but a low-frame-rate and low-pixel-count picture, and Canon internal, which displays in the correct frame rate and in color, but as far as I understand, with no guarantee that you are seeing the correct boundaries of what you are recording. You just have to pick which is more helpful to you while shooting, remembering that it's only the display and doesn't change what you are recording.
I hope others will chime in with corrections and/or additions, and that this helps a bit. Cheers.