Hi guys,
I have been projecting from the right side, testing out the 4 corner adjustment manually and it turns out the horizontal distortion is pretty bad as you can see from picture 1 that Netflix icon is much longer than the YouTube icon.
Did some research and clearly shows this is a “H Linearity” issue (pic 2).
Wonder if Philip will fix this?Thanks.
WOW, that’s unfortunate. Looks like the 3D transform isn’t perspective correct. As the 4-point stuff is all GPU there’s nothing stopping them from fixing this in firmware as long as the GPU is having enough horsepower to do so (which it should have).
We’ve seen that the optics path seems to be fairly linear (when no corrections are being applied and projected perpendicular to the screen) so this has to be 100% a digital linearity problem. You can tell that either very little QA time was spent validating image correctness OR QA was bypassed with the “WE NEED TO SHIP” card. Happens all the time in product development.
I agree this can and has to be fixed, or geometry correction would be useless even with internal content.
I also assume that V linearity is kept in check by the TI DLPC firmware when using vertical keysone?
That’s what I’ve seen on previous chipsets, V linearity is being compensated for (but can’t be used in 4-corner mode of course)
Just adding my two cents of agreement, that this definitely needs to be worked on
This issue would be a big refund request for me if I receive it and it is not doing what it’s supposed to
One thing that realized just a short while ago, when distorting the geometry the relative brightness of the pixels also have to be adjusted (the pixels will appear larger where the geometry has been compressed to compensate for a longer distance), which means those larger pixels will appear darker than the non-compressed ones. There should be a setting that enables/disables this though as the total brightness will be lowered in order to match the dimmest parts of the picture (largest physical pixels).
@Philips_Support_P please elaborate on this. If linearity distortion happens on PPM can it be fixed with a future software update? If not, it will render 4 corner adjustment useless even with the internal content.
That is technically true, and I suppose that should be quite the same also for projectors with optical correction, however, i never noticed any decay in luminance in typical installation (although optical correction rarely allows more than a 30% offset from center, which is usually more than enough for most cases). I don’t see a decay in the photo above, although the angle must be quite extreme, to produce that kind of linearity.
Hmm, you’re right, it isn’t obvious on camera that the luminance differs left to right. I’ve seen the obvious errors when stitching multiple projectors together, but then we’re talking a lot more brightness. Luminance is not the main issue right now.
I have seen this as well, Philips please look into this. This is not something users can correct and it makes keystone a very rough implementation.
I´m satisfied with everything, but there is an isse wht the image on the lieft side of the screen when looking at HDMI-input (Apple TV) The Keystone seems wrong and the border is not vertical. The right side is fine and if I switch input source to internal the screen is perfect. The ATV is updated and not defective, works fine with other screens
@Andreas_Bjurling it’s a known issue and on our task list to solve! But maybe you’d like to attach some pictures to explain more clearly, since this issue only happens for internal android, not HDMI sources.
I´ll see if I can get some pictures for you. But it is the opposite for me. Internal UI works perfectly, when I switch to HDMI, the left side is not vertical anymore. It did not change after software update, but remained the same
External sources (HDMI/type-c) only supports vertical keystone correction. Have you tried turning that off and see?
My initial impression is that maybe the internal is actually wrong so it’s not right when in HDMI so you might need to reposition your projector physically.
Just for reference that this Topic is listed as Task #29 according to:
