Return to the CurtPalme.com main site CurtPalme.com Home Theater Forum
A forum with a sense of fun and community for Home Theater enthusiasts!
Products for Sale ] [ FAQ: Hooking it all up ] [ CRT Primer/FAQ ] [ Best/Worst CRT Projectors List ] [ Setup Tips & Manuals ] [ Advanced Procedures ] [ Newsletters ]

 
Forum FAQForum FAQ   SearchSearch   MemberlistMemberlist  Photo AlbumsPhoto Albums  RegisterRegister 
 MembershipClub Membership   ProfileProfile   Private MessagesPrivate Messages   Log inLog in 
Blu-ray disc release list and must-have titles. Buy the latest and best Blu-ray titles to show off in your home theater!

E-shift in software?
Goto page 1, 2, 3  Next
 
Post new topic   Reply to topic   Printer-friendly view    CurtPalme.com Forum Index -> Lumagen Products
View previous topic :: View next topic  
Author Message
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Mon Dec 26, 2016 6:26 am    Post subject: E-shift in software? Reply with quote


        Register to remove this ad. It's free!
Just curious about this:

I have a basic understanding of how E-shift works, with the displayed image shifted diagonally half a pixel
and the shifted and non-shifted frames alternate. You might even consider it to be a form of interlacing, in
very broad and simplistic terms.

Would it be practical for Lumagen to be able to add a software E-shift feature to any of its processors?

I'm asking because I'm of the opinion that the best CRT projectors still in service in home theaters just might be able
to benefit from the e-shift concept as implemented in software by a suitably capable projector.

The idea is simple enough in theory: Add very tiny delays to the vertical and horizontal sync information on an every other frame basis. First frame (no shift) comes in with normal timings, second frame (shifted) comes in with both H and V sync signals delayed by half the pixel clock time for H shift and half the vertical time interval between the start of scan lines, for V.
Back to top
garyfritz




Joined: 08 Apr 2006
Posts: 12024
Location: Fort Collins, CO


PostLink    Posted: Mon Dec 26, 2016 3:15 pm    Post subject: Reply with quote

I don't think there would be any benefit from horizontal shift. There are no pixels and I don't think anybody's going to see a "pixel" difference from one pixel clock tick to the next. Vertical shift, maybe, if your spot size is so tight that you can see scan lines on 1080p. But I would think just tweaking your spot to be oval instead of round, so the scan lines just barely overlap, would have the same effect.
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Mon Dec 26, 2016 3:24 pm    Post subject: Reply with quote

I'm not sure either. But if the option was available, I'd try it. It doesn't seem like the idea is very complex to implement in software,
but I'm no software developer.


Actually my spot size is indeed so tight that I can see scan lines at 1080p but I have to vertically expand the image by
a fair amount to really see this. So that's not really quite making it at 1080p. But I do believe that I can sharpen this up even more. That's an ongoing, you might even say permanent project.
Back to top
garyfritz




Joined: 08 Apr 2006
Posts: 12024
Location: Fort Collins, CO


PostLink    Posted: Mon Dec 26, 2016 4:54 pm    Post subject: Reply with quote

Tightening up the horizontal focus might be worthwhile. You'll get more horizontal resolution in your picture. With a true analog signal there is a huge amount of horizontal bandwidth and the CRT can show basically all of that. But nobody displays analog any more, and with your 1080p image you only have 1920 horizontal "pixels." Not sure there's much point in tightening the focus beyond that. I don't know if the scanning spot moves continuously along the scan line, or if it steps for each "pixel." I suspect the former but I'm not sure how those details work.

But tightening up the vertical focus is pointless. It just makes your scan lines more visible, and that's an undesirable artifact. The simple solution for that is to make your spot oval. An e-shift hack *might* work just as well, but why? The spot shape is easy and I think it would do at least as well, with less complexity. And no "vibration" of the scan lines.
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Mon Dec 26, 2016 8:44 pm    Post subject: Reply with quote

Well, consider this: If the spatial resolution of a single horizontal scan line is sufficient to resolve all 1920 pixels' worth of data on that scan line, Then it stands to reason that another scan line stepped half a pixel to the side and half a scan line down, in the "interlaced" second field would also be able to spatially resolve all THOSE 1920 pixels as well. So for CRT, with the sharpest
projectors, I hypothesize that an E shift in the raster scan system MIGHT be able to deliver more picture detail information than 1080p, but of course I would not be able to promise 4K-like performance.

I mean, we know that 1080i has more actual picture detail in it than 720p. For the same reason I think that an eshift done at the video processor has some potential to improve apparent picture resolution.

The idea of beam shaping the scanned electron beam has occured to me in the past. The difficult part is that the shape of the beam changes as the beam to phospor layer angle changes across the tube face. We already have beam shape correction systems, and
some projectors take this to extreme. (Such as active astig and triangularity adjustments on both the Barco 909 and the Sony G90)
Changing the beam shape to have a narrow vertical profile (|) instead of (O) would act to maximize frequency responsiveness and reduce cross-the-spot blurring effects. But it's a challenge to achieve it and it's still going to be distorted as you get farther off center.
Back to top
gjaky




Joined: 05 Jun 2010
Posts: 2789
Location: Budapest, Hungary


PostLink    Posted: Mon Dec 26, 2016 9:59 pm    Post subject: Reply with quote

I have a friend who had one of his dreams come true, when he could afford a used 3chip DLP DPI TITAN HD250 projector, which is a great projector except for its 720P resolution... This guy happens to be very clever actualy, he also designs astronomical telescopes, so he knows his way around optical devices. He simply took the idea of JVC's e-shift technology and built his own DIY variant do overcome the shortcomings of his 720P projector.
It needs a HTPC to do the signal processing (and video playback), and needs a projector which can display the source frame sequental (ie. not droping any frames) and is not 24Hz -the DPI Titan does this at 60Hz. And needs of course the proprietary vibrating mirror device -that he made, and a small electronic circuit that drives the mirror -which... ahem... I made... Very Happy

With this it is possible to virtually double the resolution of ANY digital projector being able to display pictures without dropping frames. I also was at his place to check out this system in person (which is working almost a year now without problems...) and I must say it was great! Really! No screendoor was seen even though it is only a 720P projector, yet fine details still looked crispy, fine text edges looked antialiased etc. This stuff made the 720P projector to be a virtual 1440P projector.
Again, this method can be used on any systems, so with a 1080P projector (which have good enough lens) can be upgraded to capable of display virtual 4K at a very low cost.
So far the system was tested on 3 projectors:
Qumi LED DLP -it works, although the lenses are cheap plastics and low light output make it useless
Samsung A600 -it did not work because it has no synchronized output, ie. the projector is constantly throwing frames away
DPI TITAN 250 -it works

PIC1
PIC2

_________________
projectors in the past : NEC 6-9PG xtra, Electrohome Marquee 6-7500, NEC XG 1351 LC ( with super modified Electrohome VNB neckboard !!!)
current: VDC Marquee 9500LC
The MOD: VNB-DB, VIM-DB
Back to top
View user's photo album (1 photos)
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Mon Dec 26, 2016 10:30 pm    Post subject: Reply with quote

Exactly! The only limitations that would keep a software based E-shift system from working properly with a CRT projector would
be having a phosphor spot size that's too big which covers up the theoretical added detail and a phosphor grain structure that is also too big to show the increase in detail resolution.

I think that there's probably someone on the forums who's bright enough to implement this in software on an HTPC, for testing purposes. Any takers?
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Wed Dec 28, 2016 7:25 pm    Post subject: Reply with quote

Continuing to think about this, I see absolutely no reason why a software E-shift can't benefit a CRT projector.

I say this because a JVC D-ILA projector with e-shift has what is functionally the same visual characteristic of a CRT that can resolve
1080P with some level of quality. That being, there is no room on the raster for the extra pixels found in the shifted frame, but
it doesn't matter because when the shifted frame is being painted, that's the only frame currently being painted. While the pixels may overlap between the shifted and non-shifted frames, they're separated temporally. Separated in time. Again, this would work
very much like 1080i as compared to, properly considered, 540P rather than 720P.
Back to top
xmob135lc




Joined: 15 Sep 2012
Posts: 80



PostLink    Posted: Thu Dec 29, 2016 1:06 pm    Post subject: Reply with quote

Commercial arrangements have double "ADDRESSED" resolution , effective is about ~1.5x , and that only comes after an MTF hit , & motion is bad to begin with .

The Hewlett Packard paper about wobulation is >10yr old and they are clear about addressed 2x vs. effective 1.5x . This mostly helps with digitals 90% fixed pixel fill , bc. that can be quite a distraction (edgy), and motion is traded off already, so that's "not hurt".

CRT can't take the MTF hit , it has overlap to begin with, and if you are not keen on using "single flash" on CRT , IMO you don't have a reason to stick with analog really.

Double flash 48p or triple flash 72p or quadruple 96i is the same bad "quasi motion picture" as digital.

Sports, "HFR" ,web, 60fps/ 60p / fast vertical retrace is CRT's stuff , and >720p @ 120hz would just mess that up (bc. double flash and double vertical retrace).
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Thu Dec 29, 2016 2:17 pm    Post subject: Reply with quote

OK, nobody would expect a software e shift to actually double apparent resolution, but actually the argument regarding MTF may not apply as the e-shifted image doesn't appear at the same time as the non-shifted image. This is never asking the projector to do more than 1080p in reality, just asking it to display slightly different content half the time and shifted half a pixel and half a scan line.

Presumably the unshifted and shifted images would be processed to reveal slightly different details in the image.

I'd rather have this feature available and decide for myself whether or not it adds anything than to not have the option at all
because somebody else doesn't think it'll achieve anything useful.

There was a time when people thought no CRT projector would be able to resolve better than 720p or 1080i but we know that didn't turn out to be true.
Back to top
xmob135lc




Joined: 15 Sep 2012
Posts: 80



PostLink    Posted: Thu Dec 29, 2016 2:31 pm    Post subject: Reply with quote

Quote:
may not apply as the e-shifted image doesn't appear at the same time as the non-shifted image.


electronically not the same , perceptually it is the same time , hence overlap / fuse.

Try to do it somehow within a "single" flash then you have something new.
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Thu Dec 29, 2016 3:33 pm    Post subject: Reply with quote

Then e-shift on a JVC D-ILA unti would also have the same overlap/fuse issue. Why would it not? In both cases you have 1080p native devices (We'll say that the CRT is doing that since we're not going to deviate from displaying 1080p) alternating two full 1080p fields with a slight physical iimage offset. In the case of the D-ILA machine this is done mechanically, with software e-shift it's done by adjusting signal timing.

Net result should be same-same. If not, do go into some detail as to why it would not work.

And it's the EASIEST way to even give the CRT crowd any reasonable hope of getting significantly more out of our machines than 1080p. Why would there be any resistance to trying it?

I can even see, generally, how the software engineer would do it. There's code to handle interlacing, so use that, change the resolution data, add a routine to offset H and V timing sync signals by half a pixel's worth on half the frames. Conceptually simple.
Back to top
xmob135lc




Joined: 15 Sep 2012
Posts: 80



PostLink    Posted: Thu Dec 29, 2016 3:58 pm    Post subject: Reply with quote

Quote:
If not, do go into some detail as to why it would not work.


Usually the source image is turned into a collection of subframes with lower spatial resolution AND hard edges ,

and the temporal reconstruction of the high frequency detail actually relies on the presence of exactly those exact 1:1 hard edges (typical of fixed pixel panel displays),

CRT usually can't deliver said hard edges (low ,non-uniform MTF and problematic linearity ) , so no reconstruction of the encoded high frequency details. (there is no high frequency / MTF to begin with ).

Then the observer isn't still... it's a can of worms really.
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Thu Dec 29, 2016 4:56 pm    Post subject: Reply with quote

Ever seen a well tuned G90 or highly modded Ultra, or a Cine 9 in top form? They define edges a lot more precisely than you might think. Granted, they'll never achieve 100 percent MTF or even close to it on such a small scale of measurement, but if you are saying that the extra added detail is a consequence of the hard edges of fixed pixels, I beg to differ. I have quite clearly seen the difference between pixel edge related sharpness artifiacts and real sharp details due to the presence of high bandwidth detailed content in the image, and I do not believe that eshift requires hard pixel edges to exploit.

I do NOT understand your resistance to looking into an idea that just might work when it's the cheapest and almost certainly the simplest way to implement a means to push CRT past 1080p. I'm all about what CAN we try, not why it might not work.

And it might not work. It might be a wasted effort. But really we will NOT know without experimentation as I believe that your technical objections won't pan out if put to the test.
Back to top
barclay66




Joined: 27 Jun 2011
Posts: 1291
Location: Germany

TV/Projector: Marquee 9500 Ultra


PostLink    Posted: Thu Dec 29, 2016 5:41 pm    Post subject: Reply with quote

Hi,

In my opinion, the solution could be very simple. You'll need some kind of preprocessor which separates the 4k picture into the two lower res pictures plus delivering a status signal for identifying those pictures (unshifted/shifted). That's the complicated part.
Inside the PJ, You only need an additional circuit which will add a small bias current to each deflection coil which will result in shifting the image marginally (adjustable). The bias will be enabled through the status signal (shifted=enabled) and You're done. It could even work using the convergence coils...

Regards,
barclay66
Back to top
View user's photo album (25 photos)
xmob135lc




Joined: 15 Sep 2012
Posts: 80



PostLink    Posted: Thu Dec 29, 2016 6:05 pm    Post subject: Imaging Technology Department, Hewlett-Packard Labs Reply with quote

ON THE RESOLUTION LIMITS OF SUPERIMPOSED PROJECTION
Quote:
(...)we show that non-uniform sampling and pixel reconstruction functions impose fundamental limits on achievable resolution.
Quote:

Majumder [6] claimed to show that resolution enhancement is impossible except when the superimposed grid was uniform
and when pixel sizes are such that no overlap between pixels is allowed.



http://hpl.hp.com/research/pluribus/pluribus_icip07.pdf
Back to top
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Thu Dec 29, 2016 6:30 pm    Post subject: Reply with quote

Xmob135lc, The resistance you are showing to a new idea is frankly mystifying. I am sure that people thought man would never fly or walk on the moon, either.

Fortunately, I don't have to convince YOU to try something new. Just...somebody. It's obvious that your mind is closed to new ideas and the very idea of TRYING SOMETHING that hasn't been done before.

Quick, does JVC's eshift work if the projector is just barely out of focus enough that the defined pixel edges start to blend?

YES, IT DOES.

So if it works there then it WILL work on a projector that has adequate sharpness.

Once again, this is the lowest cost and technically easiest to give CRT enthusiasts a POSSIBLE path to enjoying a greater level
of image detail than 1080p provides. To resist the path of least resistance tells me that you have zero interest in resolution enhancements for CRT projection, end of story. Which means to me that you should abandon this topic as it is clearly not one you have any interest in making any positive contributions to.

This was intended to suggest something new to Lumagen. Not to have somebody else (I presume you do NOT work for Lumagen?) say he thinks it won't work and then points to documents of questionable relevance to "prove" his point.

I did not intend for this to turn into an argument.

So I'll say, dear Lumagen professionals, I have made my proposal. It's up to YOU to decide if it's worth pursuing. Really no other opinion matters. I've made my case. There's no reason for me to continue to argue for it.
Back to top
gjaky




Joined: 05 Jun 2010
Posts: 2789
Location: Budapest, Hungary


PostLink    Posted: Thu Dec 29, 2016 8:34 pm    Post subject: Reply with quote

barclay66 wrote:
Hi,

In my opinion, the solution could be very simple. You'll need some kind of preprocessor which separates the 4k picture into the two lower res pictures plus delivering a status signal for identifying those pictures (unshifted/shifted). That's the complicated part.
Inside the PJ, You only need an additional circuit which will add a small bias current to each deflection coil which will result in shifting the image marginally (adjustable). The bias will be enabled through the status signal (shifted=enabled) and You're done. It could even work using the convergence coils...

Regards,
barclay66


Hi Barclay,

This is what we done in the above mentioned setup as well for the vibrating mirror device.
There was a small script written in AVIsynth, and this was combined with the Stereoscopic player. The recognition of the "frame phase" is not very importand in the first hand because the software can flip the phase with one single button (F7), and still you have 50% chance you catch it on the right frame anyway.
Electronically then only a T flip-flop (or a binary counter) feed by the vertical sync pulse is what you need, modulating the position is a piece of cake, although it is projector dependent how it should be carried out.

In the attachement you can find the schematic of the 1st gen vibrating miroor driver, it even contains a delay, since with digital projectors you may never know how much the displayed picture is shifted from the input sync pule, this is not a problem with CRT though.



rezgo_tukor.jpg
 Description:

Download
 Filename:  rezgo_tukor.jpg
 Filesize:  196.46 KB
 Downloaded:  766 Time(s)


_________________
projectors in the past : NEC 6-9PG xtra, Electrohome Marquee 6-7500, NEC XG 1351 LC ( with super modified Electrohome VNB neckboard !!!)
current: VDC Marquee 9500LC
The MOD: VNB-DB, VIM-DB
Back to top
View user's photo album (1 photos)
cmjohnson




Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s


PostLink    Posted: Thu Dec 29, 2016 9:54 pm    Post subject: Reply with quote

Gjaky, what is YOUR opinion on the utility of this e-shift in software idea? I respect your input and knowledge.

Do you think it could provide a meaningful improvement in apparent detail for the best CRT projectors?
Back to top
kal
Forum Administrator



Joined: 06 Mar 2006
Posts: 17850
Location: Ottawa, Canada

TV/Projector: JVC DLA-NZ7


PostLink    Posted: Thu Dec 29, 2016 10:08 pm    Post subject: Reply with quote

Maybe I'm missing something completely obviously, but how can you do eshift in software and why would you want to?

On 1080p JVCs it's done in hardware to boost pixel density, to get rid of the screen door effect like so:



It makes a 1080p image effectively have the pixel density of 4K but of course doesn't give you a 4K image, very similar to what upscaling on CRT is done to fill in scanlines. Just like upscaling, eshift doesn't give you truly higher resolution however. Just like how on CRT 480p scaled to 1080p still looks like 480p (just smoother, without gaps), 1080p eshifted to 4K still looks like 1080, again just smoother with a better fill ratio.

To do 1080p -> 4K pixel shift in software you'd need 4K of bandwidth to begin with so why would you want to do that? That's just scaling. You can't apply eshift to a 1080p signal on a 1080p bandwidth limited system expecting to get more density out of it because there's no room. Eshift works because it's hardware only, applied over a 1080p signal to "mimic" 4K by increasing pixel density.

Or maybe I'm missing what you guys are talking about, as I'll admit not having read most of the above comments... Wink

Kal

_________________

Support our site by using our affiliate links. We thank you!
My basement/HT/bar/brewery build 2.0
Back to top
View user's photo album (18 photos)
Display posts from previous:   
Post new topic   Reply to topic   Printer-friendly view    CurtPalme.com Forum Index -> Lumagen Products All times are GMT
Goto page 1, 2, 3  Next
Page 1 of 3
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum