Originally Posted by Don Landis
With top bottom you do lose half your vertical twice, once with the FPR and again with the sharing or squeezing into the screen height of two images. The effective horizontal resolution on FPR remains x1 so the top bottom is not recommended unless you like 230 x 1920 3D.
The only reason why T/B is even present is because there were some early productions that used it for a distribution format before frame packing technology became standard. There also may be some broadcasts that still use the uncommon format. T/B full, however is a good choice to use for file size compression in intermediate rendering for video editing 3D. Here the format doesn't matter much because it is not viewed or distributed.
For best passive viewing, use SBS half which will give you 540 x 960 or with frame packing from a BD, it will be 540x1920. Never use SBS half for blu ray. When we get the 4K TVs we will have 3D Blu Ray at the full 1080 x 1920. And with 4k 3D content, the resolution will be 1080 x 3840.
Wow this forum is acting up... just typed out a long repy and it duplicated, then deleted dupe and both copies are gone... >
So basically I disagree and think you are not representing how FPR and passive 3D works accurately.
With O/U you are getting each eye 1920x540 lines of data (half a 1080p frame split along the horizontal axis) and with SBS you are getting 960x1080 to each eye (half a 1080 split along the vertical axis).
When you feed each to an FPR display o/u is now either line doubled or just just has blank lines inserted between each horizontal line of data (it doesn't matter becuase you will never see those newly made lines whatever they are) and then displayed effectively makig each eye see 1920x1080 where 540 horizontal lines are original data and the other are garbage that will be blocked. The end result: You see you original 1920x540 lines of data to each eye. Which is half the pixels of 1080p.
With SBS each eye has 960x1080 lines of original data, however this has to be stretched horizontally to fill a 1920x1080p pixel space so each vertical line must be doubled/interpolated to create the missing lines. However of the 1080 lines of horizontal data, half will be lost to the FPR. So the result is that you are down to 960x540 with half SBS of original data or 1/4 the pixes of a 1080p frame.
My empiracle testing shows that when I encode to O/U with all other settings the sam ethe image is noteably sharper and cripser and with SBS the image is much softer backing up the math.
So the basic idea is that lets say the screen resolution is actually 6x4 pixels for easy math. A full frame packed 3D would show 24 pixels per eye (a full 6x4 frame) with half ou you would have a 6x4 rame of video where the top 2 horzitonal lines of data are all left eye and the bottom 2 horizontal are all right eye (or vice versa whatever) giving your left eye 6x2 (12 pixels) of image data and the right eye 6x2 (12 pixles of image data).
When displayed on an FPR TV the left eye will be stretched back out to 6x4 (lets go with line doublling as the effect is the same as inserting blank lines) so if you number each horizontal line of data 1 2 you would essentially end up with
1122 all of which are 6 pixels accorss.
When the FPR blocks every other row to that eye (represented by an x) you get 1x2x in other words you get all 2 rows of original resolution displayed. So the only info you lost was garbage info created to stretch the image out anyway and each eye gets 6x2 real image data for 12 pixels per eye... that's the original 12 pixels that were encoded in the o/u video stream.
Now with half sbs the image is split the other way so of the 6 vertical lines 3 are for the left eye and 3 for the right and each eye maintains all 4 horizontal lines as it's not split that way. So your left eye is encoded at 3x4 (12 pixels) and the right eye is encoded at 3x4 (12 pixels).
What happens when you feed that 3x4 left eye to an FPR dispaly? It has to stretch it out to a 6x4 image which means doubling the 3 vertical lines of resoultion to get 112233 each 4 pxiels tall.
The problem is when you feed that to the FPR display 2 of those 4 horizontal lines will be blocked... and those were original image data... so now you what each eye sees is 112233 vertical lines (really 3 original lines of data) and 2 horizontal lines of data for 6 pixels per eye of original data.
Now granted you do see 12 pixels per eye of something on screen, but half of those pixels are either line doubled or interpolated lines because remember the original sournce only had 3 lines of vertical data per eye so something got created somewhere and created datra is never as good as original data.
So what we get is that o/u on a n FPR shows you all original pixels of image data which is one 1920x540 frame per eye, but SBS on an FPR shows you half the original pixels (becuase half of them are blocked for each eye) with an end result of 960x540 per eye.
This was all discussed over here as well... http://www.avsforum.com/t/1364877/top-bottom-3d-vs-side-by-side-on-passive-display/90
So I think the mistake is when you say "once with the FPR and again with the sharing or squeezing into the screen height of two images."
You have the order backwards (you lose half to the squeezing first then you lose have to the display) but more importantly you are forgetting there is a doubling going on somewhere in the middle in order to stretch the image to fit a 16:9 disiplay (as each eye is encoded to 16x4.5). So half, double, then half with a very important detail being that the second half that is thrown away is the result of the doublling so no original data is thrown away, just the data taht was created during doubling for a result of 1/2 original full frame image data.
With SBS you get half a different direction, doubled, then halved BUT the second halving does not throw away only the data taht was created during doubling.... half of what it throws away is original image data for a total result of 1/4 original full fram data.
Unless I am missing something here...Edited by Devedander - 2/9/13 at 10:13am