AVS Forum banner

Dark Frame Insertion at VP - Compatible with PJs?

2K views 25 replies 9 participants last post by  sarangiman 
#1 ·
I'm not particularly interested in frame interpolation, but I am quite interested in dark frame insertion (DFI) - loss in brightness notwithstanding - as a means of reducing/eliminating the 'sample and hold' (SAH) induced motion blur that affects many current projectors (especially LCD/LCOS). While the latter do suffer from some inherent motion blur as a limitation of the technology, I would think the SAH component is probably more significant, and remedying that would get one closer to the look of actual projected film (nicely replicated by the naturally decaying phosphors in CRTs).


I know the Sony VW200 incorporates this at the projector level, but are there any video processors that feature DFI, and if so, would they work with the current crop of projectors?


I would think DFI at the VP level would be a relatively simple feature to incorporate - simply adding in a black frame/image once every 24 seconds (1/48th of a second separated from the native frame/image) should almost certainly be easier to implement than interpolating a detailed image at the same frequency.


DFI as articulated above would result in 1080p23.98 material encoded on disc being fed to the projector as 1080p47.96 (23.98 image frames/second, and 23.98 black frames/second). The key question is whether current projectors would be capable of accepting such a signal. I know that many (all?) will accept 1080p24, as well as 1080p60 and 1080p50 (PAL), but will they accept 1080p48 (shorthand for 47.96)? The JVC RS-2 for eg. does not spec 1080p48 among accepted digital video timings/resolutions, but it does list a number of other timings/resolutions for PC signals, and the fact (I think) that it accepts 1080p24 input and doubles (or quadruples) it to 1080p48 or 96 suggests that it should (perhaps with a minor firmware tweak?) be able to accept a 1080p48 digital video input. Certainly the latter would not present any bandwidth issues, since the projector happily accepts 1080p60. I no longer have a Lumagen HDP, but someone could verify this by using a Lumagen (or HTPC, etc.) to feed an RS-2 a custom 1080p timing.


To recap:


1. Are there any VPs that perform DFI?


2. Can most currently available projectors (I'm particularly curious about the RS-2) accept a 1080p48 'custom' timing?
 
See less See more
#2 ·

Quote:
Originally Posted by Zenjabil /forum/post/16033992


DFI as articulated above would result in 1080p23.98 material encoded on disc being fed to the projector as 1080p47.96 (23.98 image frames/second, and 23.98 black frames/second).

Flashing each frame once like that would flicker too much. Film projectors flash each frame twice (and some even flash thrice). So you would probably want at least a 1080p96 signal for DFI.
 
#4 ·

Quote:
Originally Posted by Erik Garci /forum/post/16035797


Flashing each frame once like that would flicker too much. Film projectors flash each frame twice (and some even flash thrice). So you would probably want at least a 1080p96 signal for DFI.

Might not be an issue. If I'm not mistaken, the RS-2 displays 1080p24 at 1080p96 (it's doubled twice, i.e. effectively quadrupled). What this means is that both the image and dark frames within the 1080p48 input would be displayed (flashed) twice...which should go a long way towards ameliorating flickering.


Let me know if you disagree.
 
#5 ·

Quote:
Originally Posted by LilGator /forum/post/16036312


I know my InFocus SP-7210 accepted 48p (47.952Hz), perhaps the IN-83 does?

Good to know. If anyone's using a Lumagen or HTPC with an RS-2, can you confirm if the latter will accept 1080p47.952? If it does, the only question that remains is whether any commercially available VPs are capable of dark frame insertion. If not, I would think that an HTPC should be flexible enough to do so. Can anyone confirm that?
 
#6 ·

Quote:
Originally Posted by Erik Garci /forum/post/16035797


Flashing each frame once like that would flicker too much. Film projectors flash each frame twice (and some even flash thrice). So you would probably want at least a 1080p96 signal for DFI.

I'm pretty sure that for 24p material DFI on the Sony's is implemented at 96hz for this very reason. Also inserting a pure black frame results in a significant brightness drop which is why Sony offer 2 additional modes using gamma corrected "image" frames.
 
#7 ·

Quote:
Originally Posted by Zenjabil /forum/post/16037182


Might not be an issue. If I'm not mistaken, the RS-2 displays 1080p24 at 1080p96 (it's doubled twice, i.e. effectively quadrupled). What this means is that both the image and dark frames within the 1080p48 input would be displayed (flashed) twice...which should go a long way towards ameliorating flickering.


Let me know if you disagree.

I disagree.


1080p24: ABC (each frame shown continuously, not flickery)

1080p48: AdBdCd (each frame flashed once, very flickery)

1080p96: AdAdBdBdCdCd (each frame flashed twice, slightly flickery)


where "d" is dark frame.
 
#8 ·
The problem though is that 1080p96 isn't a standard frame rate. A VP manufacturer could add it, but I don't think that most digital displays will accept it. Overall, I think it's best to keep it in the display. The benefit of keeping it in the display are 1) It doesn't have to be tied to an exact multiple of the frame rate. 2) The bandwidth requirements over HDMI are reduced which is especially important for long projector runs and 3) The display manufacturer can apply a DFI algorithm that works best for their particular flavor of technology and specific refresh requirements.


The reason why an exact multiple of the frame rate isn't needed is because the panel can be driven to black for a subframe interval for any given video frame. As an example say the source is 24hz and we're talking about a LCD or LCOS display. The panel could be refreshed to say 72hz, 96hz or 120hz and a different black subframe interval could be used for each of those refresh rates. With DLP I would assume it's pretty easy to drive it for subframe timing, but for LCOS it really requires a digital backplane, but both JVC and apparently Sony have gone that route.
 
#9 ·

Quote:
Originally Posted by Erik Garci /forum/post/16038005


I disagree.


1080p24: ABC (each frame shown continuously, not flickery)

1080p48: AdBdCd (each frame flashed once, very flickery)

1080p96: AdAdBdBdCdCd (each frame flashed twice, slightly flickery)


where "d" is dark frame.

I don't think you understood what I'm driving at. Under the scenario I'm envisioning - and with native 1080p24 information encoded on disc - 1080p48 would be the output from the VP (dark frames inserted in between the native images) and thus the input into the projector. If it were then displayed 'as is' i.e. at 1080p48, then yes, one would detect significant flicker in the manner you've described. However (if I'm not mistaken), the RS-2 quadruples a 1080p24 input, and should thus double a 1080p48 input to display it as 1080p96 - the last (slightly flickery) scenario you've described, which may be quite acceptable if the reduction in 'sample & hold' derived motion blur is material.

EDIT...you are right, my response above is not correct...simple doubling of a flicker ridden 1080p48 input would do nothing to reduce flicker...see posts below
 
#10 ·

Quote:
Originally Posted by Zenjabil /forum/post/16038877


I don't think you understood what I'm driving at. Under the scenario I'm envisioning - and with native 1080p24 information encoded on disc - 1080p48 would be the output from the VP (dark frames inserted in between the native images) and thus the input into the projector. If it were then displayed 'as is' i.e. at 1080p48, then yes, one would detect significant flicker in the manner you've described. However (if I'm not mistaken), the RS-2 quadruples a 1080p24 input, and should thus double a 1080p48 input to display it as 1080p96 - the last (slightly flickery) scenario you've described, which may be quite acceptable if the reduction in 'sample & hold' derived motion blur is material.

Sending it 48 hz would give you AADDBBDDCCDD displayed when doubled to 96 hz. This is different from what you described and would result in longer exposure to dark frames and would still be very flickery (is that a word?) and wouldn't help the sample & hold effect.
 
#11 ·

Quote:
Originally Posted by Mark Petersen /forum/post/16038837


The problem though is that 1080p96 isn't a standard frame rate. A VP manufacturer could add it, but I don't think that most digital displays will accept it. Overall, I think it's best to keep it in the display. The benefit of keeping it in the display are 1) It doesn't have to be tied to an exact multiple of the frame rate. 2) The bandwidth requirements over HDMI are reduced which is especially important for long projector runs and 3) The display manufacturer can apply a DFI algorithm that works best for their particular flavor of technology and specific refresh requirements.


The reason why an exact multiple of the frame rate isn't needed is because the panel can be driven to black for a subframe interval for any given video frame. As an example say the source is 24hz and we're talking about a LCD or LCOS display. The panel could be refreshed to say 72hz, 96hz or 120hz and a different black subframe interval could be used for each of those refresh rates. With DLP I would assume it's pretty easy to drive it for subframe timing, but for LCOS it really requires a digital backplane, but both JVC and apparently Sony have gone that route.


Good point re: desirability of on-board (the projector) implementation. However, given that I'm interested in enabling DFI in a a non-DFI enabled projector (eg. the RS-2), a video processor or HTPC seems to be the only viable route.


Excellent point about black insertion for subframe intervals, allowing one to dial in the ideal compromise between motion blur and flickering. I don't see why this couldn't also be done at the VP/HTPC level, though given the likely limits on max acceptable input resolutions/timings, as well as how they're handled (automatically doubled for eg.?) by the projector, such an approach would be far less flexible than if done on board the projector.


Once again, all of the above is academic unless we can affirmatively answer the following:


1. Does the RS-2 accept 1080p48 inputs


2. Are existing VPs or HTPCs capable of performing dark frame insertion on 1080p24 material?


N.B: Sub frame insertion would be even better - for eg. 2 image frames (same frame doubled) to every dark frames which should yield a perfectly acceptable 1080p72 signal, though I don't think most projectors with accept inputs north of 1080p60
 
#12 ·

Quote:
Originally Posted by chadly25 /forum/post/16039024


Sending it 48 hz would give you AADDBBDDCCDD displayed when doubled to 96 hz. This is different from what you described and would result in longer exposure to dark frames and would still be very flickery (is that a word?) and wouldn't help the sample & hold effect.

Hmmm...you do have a point.


1080p24: ABCD, etc., etc.


1080p48 (w/ DFI performed by VP): AdBdCdDd, etc., etc.


In order to avoid the doubling of the 1080p48 signal w/in the projector to yield 1080p96 as you (and Erik Garci) describe (AAddBBddCCddDDdd), one would need to feed the projector a signal that would NOT automatically be doubled, but would still be high enough in frequency to avoid flicker. How about the sub-interval frame insertion postulated by Mark Peterson?
 
#14 ·

Quote:
Originally Posted by stanger89 /forum/post/16041614


You might want to take a look at this:
http://forum.doom9.org/archive/index.php/t-144276.html

Yup, looks like they're trying to achieve the same thing.


It seems pretty clear that simply inserting a dark frame very 24 secs into native 1080p24 (ABCD) material results in 1080p48 (AdBdCdDd) with far too much flicker. Furthermore, having the projector double this 1080p48 does nothing to diminish the flicker.


I think Mark Peterson's really on to something with sub-interval dark frame insertion, and doubling 1080p24 into 1080p48 and then adding in a dark frame would give you 1080p72 (AAdBBdCCdDDd) material that should be far more flicker friendly. That said, I don't think the RS-2 will accept a 1080p72 input (1080p60 max?).


Perhaps its possible to go the interlaced route (with a higher custom refresh frequency) and then have the projector deinterlace?
 
#15 ·

Quote:
Originally Posted by Zenjabil /forum/post/16042975


I think Mark Peterson's really on to something with sub-interval dark frame insertion, and doubling 1080p24 into 1080p48 and then adding in a dark frame would give you 1080p72 (AAdBBdCCdDDd) material that should be far more flicker friendly.

(AAdBBdCCdDDd) is still very flickery, but the image is probably brighter, because the dark frame is shown 33% of the time, instead of 50% of the time for (AdBdCdDd).
 
#16 ·

Quote:
Originally Posted by Erik Garci /forum/post/16043942


(AAdBBdCCdDDd) is still very flickery, but the image is probably brighter, because the dark frame is shown 33% of the time, instead of 50% of the time for (AdBdCdDd).

Really? I would have thought a dark frame every 1/72nd of a second would not be an issue on the flicker front. Also, on the brightness front, it probably doesn't need to be completely black...perhaps a dark gray would better mimic the average interpeak brightness of CRT phosphors (given that they take a small amount of time to decay - and recover - from on to off).
 
#17 ·

Quote:
Originally Posted by Zenjabil /forum/post/16052963


Really? I would have thought a dark frame every 1/72nd of a second would not be an issue on the flicker front.

Even though the duration of each dark frame is 1/72 sec, the frequency of the dark frames is still 24 Hz. In other words, the dark frames are still shown 24 times per second, so it will still appear very flickery.
 
#18 ·

Quote:
Originally Posted by Zenjabil /forum/post/16052963


Also, on the brightness front, it probably doesn't need to be completely black...perhaps a dark gray would better mimic the average interpeak brightness of CRT phosphors (given that they take a small amount of time to decay - and recover - from on to off).

Using dark gray frames would wash out the original frames. A better alternative would be to use darker copies of the original frames, such as AAaBBbCCc (where lower case is a darker copy), but it would still be very flickery.
 
#19 ·

Quote:
Originally Posted by Erik Garci /forum/post/16053484


Even though the duration of each dark frame is 1/72 sec, the frequency of the dark frames is still 24 Hz. In other words, the dark frames are still shown 24 times per second, so it will still appear very flickery.

Hmmm...that would mean you'd have this problem regardless of length of interval. i.e. even if the duration was 1/120th of a second, the frequency of the dark frame would still be 24Hz, and you'd still suffer from flicker. Is that correct?


If so, sounds like you'd need a 96hz signal (AaAaBbBbCcCcDdDd, etc.) to truly get at a low enough flicker...problem is that I don't think the RS-2 accepts a 96Hz signal.


I guess getting a VP or HTPC to do this is a bit of a pipe dream then?
 
#20 ·

Quote:
Originally Posted by Zenjabil /forum/post/16054403


Hmmm...that would mean you'd have this problem regardless of length of interval. i.e. even if the duration was 1/120th of a second, the frequency of the dark frame would still be 24Hz, and you'd still suffer from flicker. Is that correct?

Correct.
Quote:
Originally Posted by Zenjabil /forum/post/16054403


If so, sounds like you'd need a 96hz signal (AaAaBbBbCcCcDdDd, etc.) to truly get at a low enough flicker...

Yes, and you might prefer 1080p120, which would work not just for 24p but also for 30p, 60i, and 60p sources.
Quote:
Originally Posted by Zenjabil /forum/post/16054403


problem is that I don't think the RS-2 accepts a 96Hz signal.


I guess getting a VP or HTPC to do this is a bit of a pipe dream then?

I think some HDMI chips support 1080p96 and 1080p120 signals, so I think it would be technically possible, but I'm not aware of any existing devices.
 
#21 ·

Quote:
Originally Posted by Erik Garci /forum/post/16055364


Correct.


Yes, and you might prefer 1080p120, which would work not just for 24p but also for 30p, 60i, and 60p sources.


I think some HDMI chips support 1080p96 and 1080p120 signals, so I think it would be technically possible, but I'm not aware of any existing devices.

Thanks for the clarification. Pity about the lack of devices capable of outboard DFI. If (and that's a big if) the RS-2 does accept a 1080p96 INPUT, I would have thought one could get an HTPC to convert ABCD (1080p24) to AdAdBdBdCdCdDdDd (1080p96) without too much trouble, but then again, I've never written any software.
 
#22 ·

Quote:
1080p96: AdAdBdBdCdCd (each frame flashed twice, slightly flickery)

Not sure that the 'd' between A & A would really help w/ motion judder, since there's no reason to clear the retained 'A' image if you're going to show the 'A' image again after the dark frame.


For 96Hz, maybe AAAdBBBdCCCd would work to not dim the image too much. But, again, the dark frame is 24Hz, so would likely flicker.


What about:


AdAAdBdBBdCdCCd at 120Hz?


That's effectively 2 dark frames every 1/24s, though they're not equally spaced, which might introduce its own sort of artifact.


Fun to think about
 
#23 ·
For the sake of posterity, I've drawn out some possible scenarios for Dark Frame Insertion here:
http://cl.ly/MQB9


(3) is likely most desirable. Though I label it as "Frame-tripling', you can think of it as 'Frame-sextupling', where you replace every other frame with a black frame.


This'll lead to 72Hz flicker, & is what Sony should do.


Let's hope they implement it in the future.
 
#25 ·
Huh. Thanks for reporting your experience. However, if you don't see flicker in DFI using the aforementioned method, that indicates to me that 'Film Mode' isn't working. Actually, I really don't think 'Film Mode' works properly at all with most content. It usually introduces weird stuttering artifacts with 60p content.


If it's really re-creating the original 24 frames, you'd have to notice flicker with DFI, since it'll introduce a dark frame in between each frame (48Hz).


With 60p content, DFI doesn't introduce noticeable flicker, since the flicker frequency is 120Hz in that case.


One option is to sort of do what you say: disable 24p output on your Blu-Ray player, so it does a 3:2 pulldown to 60p. Then, DFI introduces 120Hz flicker. But, IMHO, that's a poor solution, b/c now you re-introduce 3:2 pulldown motion judder. Which bothers me more than lack of DFI bothers me!
 
#26 ·

Quote:
Originally Posted by whitetrash66  /t/1129204/dark-frame-insertion-at-vp-compatible-with-pjs#post_23021742


on my VW95, i have found that 24p material shows flicker with the DFI. I have now set my BD player to output 60 hertz, then, on the projector, i set film mode to film 1 to recreate the 24p look, and with DFI, i notice no flicker in this mode. works for me

So I tried your method on my HW50.


Unfortunately, disabling 24Hz output on your Blu-Ray player (i.e. feeding your projector a 60p signal), then using Film Mode (set to 1) to recreate the original 24p content, & then enabling DFI undoes the effects of the 60p --> 24p algorithm. You can easily see this with a good moving scene from Planet Earth (the helicopter shots from 'Mountains', e.g.). In other words, 3:2 pulldown motion judder re-emerges as soon as you enable DFI. So, essentially, you may get the benefits of DFI, but you have to live with 3:2 pulldown judder. The latter is a huge step backward, so I would never suggest this method to getting DFI working without flicker.


However, interestingly, if you then enable 'Low' Motion Enhancer, w/ DFI on & Film Mode set to 'Auto 1' (again, feeding the projector 60Hz content), the motion judder largely disappears, and you've still got DFI without too bad a flicker (60Hz flicker). Which is great for stuff like documentaries & Planet Earth sort of stuff. However, to me, even in 'Low' mode, the Motion Enhancer creates too much of a soap opera effect for movies.


So I can't really recommend this workaround for movies.


However, if you really do want to use DFI, I feel it can only be used with 60Hz content, not 24Hz Blu-Ray output. You're then left w/ the choice of: 3:2 pulldown motion judder, or soap-opera effect with less judder (by enabling 'Film Mode' to Auto 1, then turning on Motion Enhancer)


Sorry if that's a little complicated. I wish Sony (and every other manufacturer) just spelled out what all these modes do.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top