Blending, again. [Text View]

slartibartfasst

11-25-06, 07:23 PM

I'm very impressed with those blend pictures, and hope they give Andy some more optimism. The blend zones appear to be very well integrated. Are there any circumstances where the overlap is apparent?

Well done, Nick

Hi there, Nick.

I have to defer on that question, because the tubes in one of the projectors are quite worn. The array goes deficient in blue, on one side, at the high IREs. It is at these high IREs, specifically in the blue component of the image, that the blend irregularities are most apparent.

However, the blend zone is transparent in all of the material I've seen since applying this admittedly kludgey compensation. Previously, the vertical band of increased brightness in the middle of the screen when using nVidia's blending (Andy's "skunk stripe") had the effect of destroying the depth of the image and serving as a constant reminder that two projectors were employed to create the image. Now, however, the image looks continuous, if you will, across the width. Apart from an almost imperceptible drop in sharpness at the center 10% of the screen (the result of less-than-ideal convergence) I would say the image is indistinguishable from that of a single projector, at the very least with regard to colorimetry. In this regard, I would say that the nVidia approach is now at par with the VLC solution.

If anyone wants me to take any pictures, now is the time to ask. Lefty is getting a new set of tubes tomorrow, and there is a chance that the projector might not make it through the procedure. But barring that, I will be able to give a more confident answer once the new tubes are in.

RAPTORHT

11-25-06, 09:09 PM

Looking good, a couple of questions if you don't mind indulging. I am just learning about blending with a PC and so far it sounds like the next step for me to change from my stack to a blend.

Is the nVidia 6800 AGP or PCI-E? How does the PC handle HD material while doing to blend? Just wondering if there is any stutter/slowdown. What are your HTPC specs?

Gino AUS

11-26-06, 02:08 AM

They could not have gone to a better cause :D .

I sure hope the results are well worth it! But thanks again Andy! :)

Gino AUS

11-26-06, 02:11 AM

Hi Gino,

Those HQF lenses should settle the question, fairly definitively, as to whether a 2.35 blend with CRT projectors is possible with true transparency as it regards the blend-zone. If those don't cut it, all that's left is those Nikon lenses from the Sony HDIH series, right? If you don't mind me asking, what size image were you throwing, per projector, with the GT-17s?

Anyhow, I hope you're enjoying tuning out what appears, already, to be a top-shelf display.

Hi Slarti, yes, I sure hope it works, and I'm also interested to see the increase in lumens when able to use max phosphor usage for a 2.40 screen. 4:3 rasters with 10% overlap. With the GT-17's each projector was throwing a 63" x 72.6" image.

Don_Kellogg

11-26-06, 02:38 AM

Grrrr you guys..... The closer I get to finishing the theater the more you guys make me want to buy a second G90. I just have the feeling post Cliff Blend meet I'm going to be expanding. Of course I planned for this, there is plently of mounting space on the ceiling, I even put 6' of 2x10 under the drywall with an 800 pound winch to help lift them :)

Whining aside the pictures look amazing..

Chuchuf

11-26-06, 08:23 AM

Hi Tim,

The screen is 135" in the diagonal, so it's 120" wide. The picture below shows the un-blended overlap; I think it's about 35-40% per projector on the screen. The resolution per projector is 1208x1080 and the 2412x1080 resultant screen real-estate is overlapped by 512 pixels for a net 1920x1080 image. The rasters are as large as they can be in the vertical direction, and the top of the rasters are about as large as they can be in the horizontal.

Edit: Added picture.

Nice work,
Can you tell us more about the software setup you used to get rid of the white vertical line that was always present in the NVidia blending setup. Yoiu mention AVSynth.

Thanks,

Terry

welwynnick

11-26-06, 10:31 AM

Terry, did you mean a narrow vertical white line at the edge of the taper zone? That was something that Andy complained about until I showed him that was an optical illusion. I thought the fundamental problem lay with getting the gamma profiles right across the width of the blend zone, giving a farily broad skunk pipe effect.

Cheers, Nick

Tim in Phoenix

11-26-06, 10:37 AM

Hi there, Nick.

If anyone wants me to take any pictures, now is the time to ask. Lefty is getting a new set of tubes tomorrow, and there is a chance that the projector might not make it through the procedure. But barring that, I will be able to give a more confident answer once the new tubes are in.

I advise lowering G2s, contrast and brightness just before the tubes go in, set G2s to 55 or so, brite and contrast to 35 or 40 to start so you do not blast the new tubes.

YONEXSP

11-26-06, 11:05 AM

MadMrH

11-26-06, 11:18 AM

I'm guess Slarti used the Nvidia 6800 sofmodded to Quadros

I think you are right,

The Nvidia cards seem to have fairly similar boards, the drivers seem to create the differance of one card from the other - within a small range of course...........

This was why I moved away from the Quadro 4000 and over to the 7950 GX2 cards - However that move created some of its own issues, which I am still working with.........

RAPTORHT

11-26-06, 11:50 AM

I'm guess Slarti used the Nvidia 6800 sofmodded to Quadros, then used the overlap function in the driver. The Avisynth was used to do the actual blend with a custom GIF in over layed mode.. Just my best guess

So there is no active blending via software in this setup? Just gif images over the top of the blend zones?

Tim in Phoenix

11-26-06, 01:50 PM

Guys!

Arli from Chicago dropped in for a visit with his pal Adam; Arli (aka Bomrat) is Cliffy's sidekick for setting up the Chicago meet coming Dec. 9th, so Arli was wanting to see some BlendZilla setup moves, we devoted about twenty minutes to setup. I will let Arli describe his impressions of an hour watching BlendZilla.......

http://i15.tinypic.com/4dceivk.jpg

dokworm

11-26-06, 02:12 PM

I spoke to Slarti ages ago about using a "gradient bitmap" overlay with avisynth to knock out the visible line using the Nvidia blend, so my guess is he got that to work.

We should send the script to NVidia and see if it prompts them to fix it in the driver.

slartibartfasst

11-26-06, 04:36 PM

Hi RaptorHT,

The HTPC employs a Core2Duo processor running at 3Ghz, with 1 Gigabyte of RAM and an PCIe variant of the 6800U. This allows the computer to run the nVidia blend software (2% processor usage according to the Windows Task Manager), the video playback software (~40% of one core using ~20MB/s Mpeg-2@HL, using the DScaler5/IVTC audio and video decoders) and the Avisynth script (~50% of one core, using 2 instances of Overlay in a multithreaded setup). I don't see any stuttering with either MPEG-2 or with high bitrate H.264 variants. So the computer handles HD by software decoding, and isn't suffering from any stuttering, even with the blend apparatus in place. I haven't tried it yet with any HD acquisition cards, yet, though I don't think that acquisition will present a greater processing load to the system than decoding compressed material. I'm not sure that the PCIe bus will handle simultaneously porting uncompressed HD-SDI/component into the CPU and out to graphics card, but that's my next project.

Hi Gino,

63"x76", if I recall, is in the middle of the intended range of best-focus for the GT-17s. If the HQF lenses do yield better edge focus at that size, it would seem that you will have given a compelling data point regarding the focusing ability of the two lenses. So you're using a 10% overlap? I guess that explains needing the razor sharp lenses.

I've been interested, too, in the light output advantages of using the full raster. TSE mentioned, a while ago, that light output is a function of beam current. So, assuming no blooming is happening, if a smaller raster has the same current applied to the raster over the course of drawing the frame, will it have the same light output as the larger raster? I'll pull out my light meter and try and to get a grip on what's going on. At any rate, it seems that a larger raster will wear slower than a smaller one, and that less overlap per projector will yield more light.

Don, Cliff's meet sounds like it's going to be both an eye opener and a great time. I'd think, though, that with top-end 9"er the primary benefit to blending would be a legitimately cinematic 2.35 presentation, as 16x9 1080p isn't a problem. And what a fantastic idea, building with room for expansion. I'm taking notes...

Tim, thanks for the heads up. I had to replace all three tubes, though, so I ended up resetting the projector before I put the tubes in. This sets everything to 50 except for the contrast, which goes to 35, right? The Marquees are pretty amazing machines - even without a horizontal linearity control it only takes me about 2 hours to get the machine into a workeable state. They're really a great baseline for blending.

Hi Ken and Dokworm, you guys are spot on; the video card is a nVidia 6800 Ultra running Quadro 4000 drivers, and the compensation for the light vertical strip is an overlay. Thanks for all of the help with this.

Terry, I'm working on getting some pictures together to try and illustrate what's been done, so hopefully anyone who tries the nVidia and/or Avisynth route can avoid some of the pitfalls I got stuck on.

YONEXSP

11-26-06, 06:54 PM

I love my 6800 Ultra, and so chjeap now, even with 512mb RAM on board. Pity it's not HDCP Complinat then I would not have to go buy a 7900 e-VGA card for another $260CDN!! aghhh!!!

slartibartfasst

11-27-06, 03:23 PM

In brief, the method described here relies on the idea that a blend of two projectors should create a coherent image with regard to luminosity and colorimetry across the horizontal dimension of the image. If one were to split and overlap an image, with 60% on both projectors, the resultant image would have a band with twice the luminosity in the middle 20%. Simply halving the luminosity of both projectors on the innermost 20% of the blend would, assuming perfect colorimetry, yield an image that has the correct luminosity over the resultant image. In practice, with CRT projectors, this yield an image with three distinct color characteristics; one representative of the left projector, one of the right, and one that is the sum of both halves. Our eyes are sensitive to color change to such a level that no degree of manual calibration of the two projectors will yield a unified image with a simple halving of the intensity. To "blend" the images, therefore, takes a more complicated luma attenuation scheme, horizontally, across the overlapped region. Paul Bourke illustrates these blending concepts in "Edge Blending with Commodity Projectors," (http://local.wasp.uwa.edu.au/~pbourke/colour/edgeblend/)

nVidia, in their Quadro drivers, supplies an application called "nvKeystone" that effects such a blend, over and on top of a hardware function that splits and overlaps the image onto two outputs of a dual-head graphics card. The parameters for this split, overlap and blend combination can be modified using the "advanced" tab of the "display" control panel in Windows, once the appropriate driver set has been installed. I assume that the mathematical basis for this software implementation is sound; but in practice, with CRT front projectors, the package does not yield a visually unified image. This may be the result of CRT projectors having a different gamma response than is expected by the algorithm, or may suggest that the projectors are set up differently with regard to full white/black levels than is assumed. nVidia doesn't supply any guidelines with their software, so all that can be said is that the result is close, but certainly does not yield a watchable image.

Without getting too far into the nVidia control panel itself, there is generally a zone, vertically, of increased brightness at the horizontal midpoint of the screen flanked laterally by two areas characterized by a slight deficit in brightness.

So without knowing what the expected projection device characteristics that nVidia expects are, I simply left the projectors set at setting that were as close to one another as possible and yielded a good image for a home theater environment, and then went to work on attenuating the blend.

Tim in Phoenix

11-27-06, 04:42 PM

Guys!

Our DVX BlendZilla processor is purpose built with switching for eight inputs, the 60% left and right segments match nicely thru the center also:

left sixty percent

http://i12.tinypic.com/48pwjt3.jpg

right sixty percent

http://i12.tinypic.com/2dka5gh.jpg

and if it is aligned well we get this

http://i12.tinypic.com/29w3988.jpg

We can mix and match a wide range of sources from 480i to 1200p in composite, S, component, RGsB, RGBS, RGBHV, SDI and HDSDI and output at eight different resolutions for blend mode. The weather map is 1080i from Dish. The system, not PC-based, is perfectly stable once dialed in.

And this is the system that Cliff will be demoing.

Clarence

11-27-06, 04:55 PM

Guys!

Our DVX BlendZilla processor is purpose built with switching for eight inputs

And this is the system that Cliff will be demoing.
Yes, but this is slartibartfasst's nvidia blending thread, not Cliff's blendzilla demo (http://www.avsforum.com/avs-vb/showthread.php?t=729705) thread.

I'm interested in both. (Although I must repeatedly remind myself that a single G90 is sufficient for my HT).

slartibartfasst

11-27-06, 05:19 PM

The ffdshow DirectShow filter makes possible the use of a number of video manipulation algorithms, and access the "raw" video data being kicked around in DirectShow. Recent builds of this make possible the use of AviSynth scripts within ffdshow, and provide an easy way to implement AviSynth processing in real-time.

There is a lot of development going on in the Avisynth world, but for the purpose of blend-zone compensation the "overlay" function is what we're interested in. The goal, with this overlay, is to create a color-corrected, transparent overlay that is the inverse (in the horizontal direction) in intensity of the blend-zone discrepancy created by the nVidia solution. This compensation requires two components; one, a suitable graphic (in shape and relative intensity) to correct the blend-zone luminosity abberation and a method to port it onto the video stream in real time.

This document describes the method I used to create the graphic and overlay it:

http://www.scvi.net/logo.htm

This mechanical step, itself, is not hard; the difficulty lies in the iterative process of creating and tuning a suitable overlay that will effectively cancel the blend-zone irregularities created by your particular setup. If I figure out how, I will post the graphic overlays that I used, but I suspect that it will only apply to blend setting similar to mine.

The way I went about creating this graphic used Adobe Photoshop. My blend zone is 512 pixels wide, with an area on increased brightness approximately 120 pixels wide, and then about 150 pixels on either flanking side with a slightly decreased brightness.

Though I bumbled around in figuring out a method, I suggest that you download the MPEG-2 test patterns at www.w6rz.net, and play the "bluefield.mpg" pattern in your playback chain. This bluefield will illustrate, quite frankly, the deficits of your blending implementation, and the rough outline of what you’ll want to do to correct it.

With this image in mind, and perhaps on screen, you can start to sketch a rough compensatory graphic in Photoshop. I suggest using the “brush” function. Once you’ve selected the “brush” tool, pick an appropriate width for your stripe, and then go into the “brush options” panel. In there, and this is critical, select the “airbrush” tab. If you don’t do this, your transparency will not be transparent, it will have a white component in the image as opposed to a strictly black component with varying opacity. In Photoshop, if you hold the shift key while moving the mouse it will move in a straight line, in either the X or Y axis. In this case, you’ll want to move vertically down the image along the Y-axis.

Once you have this rough compensation graphic, use the procedure above to overlay it onto the video image. As tempting as it is to switch to video material to test the blend, if you simply work on the “shape” of the luminosity compensation in the monochromatic field you will get to the best result in the least time.

From there, go back and forth between Photoshop and the blue field, and try and get the luminosity of the blue field as consistent as possible across the horizontal width of the image. For me, this took two graphics using two functions: one for attenuating the brightness in the middle and one to increase the brightness in the flanking areas. For increasing the brightness in the outlying areas, instead of using the “blend” operator in the “overlay” Avisynth entry, use the “lighten” operator. Once you get satisfactory results in the blue field, switch to the green field.

In my setup, the blue channel required the most compensation. Thus, if you even out the luma compensation in the blue field, the green field image should actually show a dark are where the bright stripe used to be. What you want to do then is go back into Photoshop, and in the “blending options” tab, select the “color overlay” tab. Where the document above suggests making the value of the color white, or (255,255,255), what you want to do now is turn down the green component to limit the green attenuation. Once again, if you work iteratively between Photoshop and the green field, you should be able to achieve a similar degree of luminosity consistency across the green field as you did in the blue field. Repeat this process for the red field. The values that I ended up with were 255 for the blue, 48 for the green, and 8 for the red.

If you choose to (and it’s really not necessary) you may choose to repeat the process for increasing the luminosity of the flanking areas, using the “lighten” operator in Avisynth’s “overlay” function.

At this point you should have, in the red, blue and green fields, a reasonably consistent luma intensity across the individual color fields. At this point, you’ve effected the luma compensation, and corrected it for individual color aberrations. Hopefully, you’ve resisted the urge to watch video material thus far, and you’re thoroughly sick of the process. After stumbling upon this method, it took about two hours of work, between Photoshop and the color fields, to get the luma intensity even across the image.

At this point, you will have the nVidia blend apparatus in place, ffdshow in your video-chain, and Avisynth set up with the compensatory blend images. Throw in a movie, and sit back and enjoy.

slartibartfasst

11-27-06, 06:27 PM

Anyhow, I'm not sure if that covers it, but I'll probably go back and edit that for clarity and fill it out at some point. I'd be happy to answer any questions, and all criticisms are both requested and appreciated. I'm not really interested in being right, but I would like to collaboratively develop a really functional method for blending CRTs at a cost proportional to the cost of 8" EM machines.

At the end of all of this, the nVidia solution is closer than it was before, which was pretty damn close. I won't say it's videophile-class yet, because the compensation suggested above is just that, a compensation. It's as limited as the tool that I used as an optical comparator, my eyes, and as such doesn't really go beyond the limits of my time or patience to really rigorously ensure the desired result. That, and it's a kludgey patch, rather than an elegantly executed solution.

But, it works pretty damn well.

Hi Clarence and Tim,

Someone recently PM'd me the current asking price for the Blendzilla. For anyone looking for a turn-key blending solution with one-button simplicity, it's the only way to go, and for what it does, I think it's a great value.

But I think it exists in a different space from the computer blending projects, and this mirrors the general situation of CRT front projection in the home theater space. On one hand, you have the dedicated videophile set, for whom rapidly the diminishing advantages of the mature CRT technology still hold a edge that earns the our three eyed beasts a place in the no-compromise system. For the sub-set of these people who want large screens (very large screens?) the Blendzilla is the best end run around the light limitations of the CRT technology. Further, when you look at the whole package, if you're running a theater that's big enough to necessitate a Blendzilla, you're going to need an audio system that will fill that room. Once you get into the audio side of things in a room that size, at that level of fidelity, the Blendzilla's pricetag comes into focus a bit.

But there is the other side of CRT, represented by 8"EM projectors being sold at incredible prices. I'm not sure what the inventories of the CRT resellers that frequent this board are for 8"chassis, but I would think there would be significant incentive for everyone with a financial interest to see a workable blending solution out there. When you look at the investment that the various media industries have placed in the "HD" conversion, I think that 1920x1080 is going to be that standard for the forseeable future in the consumer space. Two 8"ers can light up a screen that would be more than appropriate for 90%+ of the rooms out there, while retaining all of the advantages of the CRT technology, and making an end-run around the vertical resolution problem that may or may not be an issue for the 180DxB22 class of tubes. You just have to find people crazy enough to hang two projectors.

This is a hobby for me. The payout is in the process; I frankly enjoy the technical aspects and design problems more than actually watching most of what passes as film these days. I think anyone with a 9"EM machine would agree that it's not really that hard to resolve a 1920x1080p image; in fact, using two blended 8" Marquees to display a 1080p image, to my eye, doesn't resolve a great deal more resolution than a single 8110. As an enthusiast, though, I sleep better knowing that there are scan-lines in the array.

You know what they say, horses for courses. I don't think PC blending is going to take a huge bite out of the market for the Blendzilla.

overclkr

11-27-06, 08:11 PM

Great post.

Cliff

RVonse

11-27-06, 09:12 PM

My hat is definately off to Slartibartfasst! A brilliant piece of work on your part and nicely done too.

The only other comments I can make is I'm very glad I stored away some very nice BG808's and that it is about time for me to upgrade the htpc to dual core.

dokworm

11-28-06, 03:13 PM

Just on the photoshop side I'd use different layers and then you can adjust their opacities on the fly as it were.
Instead of using the brush option you can rubber band out an area and then fill it with white on a new layer, then use the opacity slider for the layer to adjust the transparency to the desired level. The gradient fill tool can also be useful to get a desired result.

scorch123

11-29-06, 10:02 AM

Two 8"ers can light up a screen that would be more than appropriate for 90%+ of the rooms out there, while retaining all of the advantages of the CRT technology, and making an end-run around the vertical resolution problem that may or may not be an issue for the 180DxB22 class of tubes. You just have to find people crazy enough to hang two projectors.

slartibartfasst,

I still haven't seen a Marquee 8-series 1080P blended setup in person yet - where are you located? I hope you're not across the pond like MadMrH...

I agree with you - there are some key constraints preventing a lot of us from jumping on the 8" blended bandwagon. Twice the number of tubes, plus taking down and ceiling mounting two CRT PJs? That's a lot of free beer to give out to buddies for a weekend's worth of work :)

- Steve O.

hdtvme

11-29-06, 08:08 PM

Well done! If I could provide a way to get live 1920 x 1080i video into system memory could we then get the same results for the external sources?

Sinobi

12-12-06, 01:19 PM

Greate job, well done.

If I figure out how, I will post the graphic overlays that I used, but I suspect that it will only apply to blend setting similar to mine.
Any chance you can figure it out?
It would make it easier to understand the process.
Sidenote: What forceware driver version do you use for your modded 6800U?

Best regards
Henrik

Luke212

12-12-06, 05:54 PM

Greate job, well done.

Any chance you can figure it out?
It would make it easier to understand the process.
Sidenote: What forceware driver version do you use for your modded 6800U?

Best regards
Henrik

does this help?

http://www.watertrack.com.au/file/blend.gif

Saeid

12-12-06, 06:01 PM

That's funny!
Now can you do that 24fps?

slartibartfasst

12-12-06, 10:11 PM

Hi again,

Cliff and Bob, thanks. It's great to hear from fanatics wild enough to grapple with two of these projectors at once.

Dokworm; thanks for the pointers. In my efforts with this, I ended up with a series of layers that I would selectively activate/deactivate while taking notes on horizontal consistency and overall opacity. I played around a bit with the Gradient Fill tool, but I couldn't get the desired effect; though I don't really know my way around Photoshop. Any pointers?

Scorch123: I think two 8" Marquees are a sweet-spot for a 1080p blend. I've got scan lines over the entire raster (except the very top, due to my physical projector placement relative to the screen) enough to lead me to believe that I'm resolving the 1080 lines in the vertical plane but not enough to be seen at the viewing distance.

That said, blending doesn't really scratch the itch to upgrade the projectors, even though I'm satisfied that the image is resolved. I'm considering mounting the two sets of HD-145s I've got here to the Marquees. So, it really just makes the itch twice as painful to scratch.

Hdtvme: There are a number of 1080i ingest/acquisition cards out there. If you're familiar with the DirectShow architecture, it's not a problem to get an image. However, my nVidia card isn't up to par with dedicated hardware for 1080i60 video de-interlacing, and the software solutions are seriously computationally intensive. Which is to say, I don't know if they would work in real time and I'm not sure how they would look if they did. I will say, though, that the Radeon x1600 in the MacBookPro is a decided step above the 6800 that I'm using, and I don't have any problems watching 1080i60 video on that card.

Film is no problem.

Henrik, I would be happy to e-mail you the bitmaps I've used. Though they wouldn't help you at all if you weren't using the same per-projector resolution and overlap setting as me. However, you can export the blend settings from the nVidia card, so if it would help I can send you that file. The bitmaps themselves aren't too helpful, but I can send you the Photoshop .psd files so you can adapt them to your purposes.

I'm using the Forceware 93.50 drivers, but only because they are the last ones that have the "old" style of nVidia control panel. I don't think nVidia has changed anything in their blend software since the 5x.xx release. I haven't gotten familiar with the new style of control panel yet, and haven't seen any mention of a change to the blending software in the release notes for the newer drivers, so I'm not inclined to take the time.

Luke, last you posted, you had moved beyond DirectShow and trying to blend in the video card hardware. I'd be happy to test it for you, whenever.

EDIT: Added vertical sections of the bitmaps used for attenuating the blend irregularity. The actual bitmaps are 108x1080. The strong blue shade is a result of using a (16, 48, 255) (R,G,B) "color" for the blend attenuation.

EDIT2 and 3: For Luke: I didn't see the picture that you had posted. That's sort of an overview of the whole process. I'm not nearly as artistic as you, but what I'm describing here is a compensation to nVidia's blend, not the whole blend itself. So, in your last image, instead of "Happy" it would say "Close," and the last picture would be what is described in this thread, the compensation.

Sinobi

12-13-06, 12:26 AM

Great, thanx Slarti.

Sorry, but also forgot to ask, how exactly does the Avisynth script you use look like?
I'm dying to try it out on my 1271 set.

Forceware 91.36 seems to detect my Quadro softmodded pci-e 6800GT (NV40) as an 6800GT and thus doesn't enable the overlap menu.
Will try out other versions tonight...

Great site with test pattern videos by the way.

Henrik

Luke212

12-13-06, 01:02 AM

Hdtvme: There are a number of 1080i ingest/acquisition cards out there. If you're familiar with the DirectShow architecture, it's not a problem to get an image. However, my nVidia card isn't up to par with dedicated hardware for 1080i60 video de-interlacing, and the software solutions are seriously computationally intensive. Which is to say, I don't know if they would work in real time and I'm not sure how they would look if they did. I will say, though, that the Radeon x1600 in the MacBookPro is a decided step above the 6800 that I'm using, and I don't have any problems watching 1080i60 video on that card.

If you find a cheap 4-4-4 analog aquisition card, pls let me know. Im a bit hesitant to capture at 4-2-0 or 4-2-2 due to image quality concerns. On the digital front the Blackmagic HDMI is very cheap, but these will never accept HDCP protected material.

EDIT2 and 3: For Luke: I didn't see the picture that you had posted. That's sort of an overview of the whole process. I'm not nearly as artistic as you, but what I'm describing here is a compensation to nVidia's blend, not the whole blend itself. So, in your last image, instead of "Happy" it would say "Close," and the last picture would be what is described in this thread, the compensation.

Yes i know its not entirely accurate, but it was meant in jest! hehe.

slartibartfasst

12-13-06, 02:17 AM

Hejsa, Henrik:

The Avisynth script I'm using for the nVidia blend compensation is:

blend = ConvertToYV12(ImageSource("C:\logo.bmp", fps=23.976))
logomask =ConvertToYV12(ImageSource("C:\logomask.bmp", fps=22.976))
MT("overlay(blend, mask = logomask, opacity=.115, x=908)",3)

With the assumption that you're familiar with Avisynth, I can explain this:

The "Convert" function is less computationally intensive than whatever the "overlay" function does. This cut the processor overhead in my configuration significantly. The "fps" operator of "ImageSource" should be set to whatever the output framerate is. The "MT" function is a tweak, and is specific to my (dual-core) system to make sure that there were no peaks that caused stutter. The numerical value of the "opacity" value was perceptually arrived at, and follows from using a blue value of 255 in the Photoshop overlay. The X value of 908 is a result of using a 108 pixel wide blend-compensation in the center or a 1920 pixel-wide field. Mathematically, (1920-108) / 2 = 908. This places the 108 pixel wide image in the center of the blend, where the irregularity is. The method for arriving at the masks is clumsily described above.

But, you know, I'm not trained in computers, so perhaps someone versed in Avisynth or the programming behind it could give you a more revealing explanation.

Regarding enabling the Quadro features in the 6800, it appears that you're familiar with Mr. Ryben's procedures. It appears that you're stuck on the same step that I was. What enabled the Quadro features, for me, was to manually install the Quadro drivers using the Device Manager in Windows. Without going into depth, your OS will always report your PCIe card as a 6800x, but when you've manually installed the Quadro drivers, the Quadro feature set will be made available. So, instead of running the nVidia installer, use the Device Manager to install the drivers for a Quadro 4000.

Hi Luke,

I don't know of a 4:4:4 YCbCr capture card, but then again, I don't know of a 4:4:4 component source, either. There are certainly 444 RGB acquisition cards available in the Dual-link HD-SDI format, but you where are you going to find such material? You can pick that up off of a scaler/processor, but I've never seen any commercially available digital material sampled higher than 4:2:0. A decoder might blow it up to 4:2:2, but why not do the additional interpolation in your eye(s)? If you're source limited, why incur the processing/transport overhead unless you absolutely have to?

I'm with you, HDMI is a bitch. That said, I think the best we can do for video, on the commercially distributed side of things, is digital 1920x1080 4:2:0 YCbCr (SMPTE 292 (274?)). The HD-SDI transport does this just fine, and there are reasonable HD-SDI ingest solutions out there, so why worry about the rest? I'm guessing that full-bandwidth Component/RGB solutions are both boutique and outdated, and as such super-expensive.

You bring up a great point, though, which is how to tap the video data out there so we can muck with it. I think HD-SDI is the way to go, as far as acquisition goes.

Sinobi

12-13-06, 05:25 AM

Regarding enabling the Quadro features in the 6800, manually install the Quadro drivers using the Device Manager in Windows. So, instead of running the nVidia installer, use the Device Manager to install the drivers for a Quadro 4000.
Thanks, that did it.....onto testing.....once the sun goes down.....thats in 4 hours 29 minutes 50 no 40 no 35 seconds. :D

Henrik

Sinobi

12-13-06, 01:38 PM

I got it working (the overlap+blend) and has just finished setting up and calibrating the hole sjubang.
Although, still need to do a light/color calibration of the 2 1271 as they are very different....and your avisynth trick.
But even then it looks amazing.

Henrik

Sinobi

01-12-07, 07:09 AM

The Avisynth script I'm using for the nVidia blend compensation is:

blend = ConvertToYV12(ImageSource("C:\logo.bmp", fps=23.976))
logomask =ConvertToYV12(ImageSource("C:\logomask.bmp", fps=22.976))
MT("overlay(blend, mask = logomask, opacity=.115, x=908)",3)

The "MT" function is a tweak, and is specific to my (dual-core) system to make sure that there were no peaks that caused stutter.

Hi slart.

Could you do a recheck on the correctness of your posted MT setting.
When I try it on my P4 D920 (dual-core) I get an error.

Best reg
Henrik

dokworm

01-12-07, 08:16 AM

The $200 xbox HD-DVD drive is out and working now on the PC, anyone tried blending with one yet?

MadMrH

01-12-07, 05:33 PM

Well dont hold yer breath but the x box drive is here............

I have one 1209s to prepare for its new owner and then its FULL ON BLENDING here.........

Ive decided to remove my own 1209s from the ceiling and install 2x1209s ready to blend........

The brackets have been made, and the PJs are 90% ready to install.

BUT I must get the other 1209s completed and shipped first.............

I decided it was all or nothing this time so I will keep you all posted.

Slarti - fantastic work, well done, Cant wait to see pictures...........

Sinobi

01-12-07, 06:15 PM

Hi slart.
Could you do a recheck on the correctness of your posted MT setting.
When I try it on my P4 D920 (dual-core) I get an error.
Best reg
Henrik
Well of cause they were correct......Sorry for being too quick :rolleyes:
I use the script another way (I use mode="add" instead of the default mode="blend") as I only have to brighten and not dimm,
turns out that MT doesn't seem to accept the mode="add" parameter.
Without MT and with mode="add" it works fine, although only on one core.
With MT I get an error with mode="add".
Anybody knows why?
The damn (in a frustrating way) avisynth site has been down since before I discovered this thread...... :(

Henrik

slartibartfasst

01-12-07, 11:39 PM

Hi Henrik: I'm not well versed in Avisynth programming, so I can't help you with MTs incompatibility with the Add mode. I've been lucky enought to sidestep that particular problem; Add works fine with MT in my system. But how does the image look when you're running it on one core? What parts of the image are you trying to make brighter?

Dokworm: Does the HD-DVD drive work with non HDCP compliant graphics cards? Even if so, I can't imagine that the raw video data is made available to DirectShow (making this blend compensation irrelevant); but perhaps it is? I've been having good luck with a Decklink Intensity and a DVI Distribution Amplifier for HDMI input into the system from a Toshiba HD-A1 and my DirectTV reciever.

Mr Mad: Do let me know when you get going on the project; I'd be happy to help in any way I can. So far, Decklink HD Extreme has gone back in favor of the Decklink Intensity, which is a simplified HDMI type ingest board. I'll put some pictures up as soon as I install a set of modification plates to enable the use of HD-145 lenses on the Marquee chassis, at which time I hope to start using a 2538x1080 2.35AR screen. Apart from that, I've just been enjoying the image.

Happy New Year, to all.

v1rtu0s1ty

01-13-07, 12:33 PM

question guys about blending...doesn't crt send the black edges? I'm a bit confuse as to why/how the black gradient edges when overlapped together, it goes away.

Tim in Phoenix

01-13-07, 01:15 PM

Sinobi

01-13-07, 03:45 PM

I tried for the LAST TIME Nvidia - I cant add custom resolutions when in Horizontal span mode, I can add them in clone mode but cant them use them in H Span......
change the overlap and it changes the overall resolution..........
You end up in circles and my brain is fried!
SO Nvidia is even difficult to get just the overlap running correctly........
I felt like this when I began messing with Nvidia and in particular their driver blending solution.
Here is what I have figured out during my many hours of attempting to get Nvidias blending to work, hopefully this will help others:

It IS possible to add custom resolutions in Horizontal span mode,
You just have to click the "add" button, click "OK" in the denial message, then click the "add" button again. Hurrah for Nvidias lack of......something :rolleyes:

When changing something in the blending menu and clicking "apply", settings is accepted but the driver reverts to another resolution. This didn't happened at first, but after a short time trying different thing in the menus and setting everything back, thats what the drivers ended up with.
Changing the resolution and refresh rate back to correct setting and clicking the "OK" button (NOT!!! the "apply" button) will set everything to it's correct state.

Now, the blending menu is a mix of weird helptexts that really doesn't make much sens, and settings that also doesn't make much sens when the purpose is blending, but my brother (thanks) finally figured it out.
So here follows an explanation of Nvidias blend settings.

The blendingarea's fade to black is done by 2 linear rolloffs.
One that begins at the edge between the normal area and the blend area,
and another the same size that ends the blend area.
So the first rolloff begins at full normal luma,
and the second ends at zero luma.
The luma point where the first ends and the second begins is defined by the entered horizontal luma value.
all this can be seen here:
http://www.sinobi.dk/henrik/diverse/blending_explained.jpg

As can be figured out this enable the user to make a linear setup when the rolloff value is set to half of the horizontal value and the luma is set to 128,
but as we have talked about before no display device is linear with a gamma value of 1.
Thus the blending values has to be made something like the first example I have drawn above, where the luma value has been raised.
This enables the blending curve to be as close as possible to the gamma curve of the display device.
The result is a blending curve that starts out too low, gets too high close to the "breaking point" and finally becomes too low again.

We must conclude from this, that the german programming team from Nvidia has made the solution for linear gamma display devices, but with a possibility to adjust the curve close to the used display gamma curve.
Close but not acceptable.
As almost no display devices (I'm not aware of any) has linear gamma, the result must be: They failed.
Slartibartfasst has come up with a solution derived from the Avisynth blending script that compensates for this problem.

While Slart uses the script twice, one for increasing luminosity (mode="lighten"), another for decreasing luminosity (mode="blend"),
I stumbled upon the mode called "softlight" that can do both in one run.
You simply have one compensation picture that is gray (value 128) and do increase with lighter than default gray(>128), decrease with darker than default gray (<128).
So the script looks like this:

blend = ConvertToYV12(ImageSource("C:\compensation.bmp", fps=23.976))
overlay(blend, mode="softlight", opacity=0.150, x=864)

To not define a mask equals to 255, a white mask image, but is faster than using an actual image, so thats why this is missing, the opacity does almost the same thing although all correction now has to be done in the compensation.bmp.

The only thing I don't understand is why the MT("script", 3) only accepts mode="blend" on my rig.
Without MT(), when playing a 1920x1080 24fps mpeg2 .ts movie, I have 25-30% on one core and 96-100% on the other. The playback is smooth but I don't like the nonexcistant headroom and also would like to playback h.264 movies.

EDIT:

For some reason I have to use tripple quotes around the overlay function to make MT() work.
Like this:

blend = ConvertToYV12(ImageSource("C:\compensation.bmp", fps=23.976))
MT("""overlay(blend, mode="softlight", opacity=.075, x=864)""",3)

Also I have lowered opacity to 0.075 and raised the magnitude of the compensation image.
The reason for doing this is that when you mix the movie with an image with some magnitude of gray in it, the colours of the movie becomes grayer in the mix zone.
Fx. if you set opacity to 1.00 and mix with an image of gray-level 150, then the mix zone will only contain a band with gray (150) and no movie. The lower the opacity number the more transparent the gray image will be, thus giving clearer colours from the movie. So by streaching the magnitude of the compensation image so that either the white or black compensation part reaches it's max/min value and adjusting opacity to visual optimal setting you get the least colour reduction.

END OF EDIT

BTW. I run the set at 1920x817 with a 10% overlap.
Projectors in use is 2xSony VPH-1271Q
That is 1152x817 on each projector, and is very close to the actual max resolution they are able to produce.
One can say that they are "ideal" for a 2.35:1 blend rig with 1920 :)
PC specs are P4 D920 running at 3.5GHz and Nvidia 6800GT Rivatuner-softmodded to Quadro.
Screen is 104" 2.35:1

Henrik

v1rtu0s1ty

01-14-07, 12:30 AM

Say V

The process is called feathering and if done well, each side fades undetectibly into the other side; the more adjustiblity in the contrast modulation for this, the more liklihood of pleasing results.

http://i10.tinypic.com/2wckugz.jpg

http://i12.tinypic.com/2uhl1xv.jpg

http://i16.tinypic.com/4gsh27r.jpg

http://i18.tinypic.com/447e1rm.jpg

So going back to my question, am I correct that black are not really sent by the CRT projectors to the screen? It's because, if you look at the edges of the 2 images, the have black gradient to transparent. So if black are being sent, it will be obvious when you overlap 2 pictures together.

If I was right that the 0 black areas are not being projected, then overlapping each other will be nice. Please see attach picture of what I mean. I made a simulation in Photoshop where the background is transparent. Does a CRT projector work that way wherein the black isn't sent assuming the edges have been applied with the black blend mask?

Sorry if I have so many questions :D it's because, I'm also developing something but using linux. I had posted before about vlc but am not going to continue on that since it lacks fullscreen capability on 2 panoramix walls. I'm going back with the xinerama route but with modification on the compositor. This way, it's not only one application that has the edge blending capability but everything will be blended..

Gino AUS

01-14-07, 02:30 AM

Your picture is basically how it works... black is the absence of light

v1rtu0s1ty

01-14-07, 03:29 AM

Your picture is basically how it works... black is the absence of light

BADABING! :)

Sinobi

01-14-07, 10:21 AM

Made some small corrections to my explenation post above, both in text and in the picture to better reflect reality.

Henrik

Gino AUS

01-14-07, 10:22 PM

Henrik, so youve setup the blend for 1920x817 for 2.35 material... what happens when you watch 1.78? 1454x817 sacrificing resolution?

Sinobi

01-15-07, 06:30 AM

Henrik, so youve setup the blend for 1920x817 for 2.35 material... what happens when you watch 1.78? 1454x817 sacrificing resolution?
Yes thats correct.
The Sony 12xx series really don't resolve 1080 lines,
as the max resolution point, when perfectly adjusted, is 750-800 lines in 4:3.
Anyways, 1920x1080 1.78 material still looks amazing in 1454x817.
I guess thats one of the compromises one have to live with when not having the budget for a Blendzilla and 2x 9" ;)
At least 40% of my 1920 movies is 2.35, with DVDs the number is close to 80%
so most of the time the movie fills the screen.
and with the limited resolution of my Sonys I simply have to live with the situation :p
And I loooove it :D

Henrik

racerxnet

01-15-07, 01:12 PM

To all working on the PC blending solution, please post the links to the software you are using so that others might help in creating a solution. Be specific in the software number being used. This way we can all test and have some repeatability.

Has anyone created a graph with sliders for the blend zone to integrate with VLC or MPC? Luke 212, I thought that you were working on this solution per another web site discussion. If the latter can be done we have simplified the blend so that anyone with basic skills can run 2 projectors with a blend.

I have been layed off for now and will make some time to try and create this. I program in many languages: C, C++, .net (yuck) VB, and others. Please let me know what your thoughts are on how you would approach this, and I'll incorporate as much as I can in the code.

Thanks,

MAK :D

sarahstaar

01-16-07, 09:54 AM

would there be any advantage in using a 9" crt over an 8"crt in blending.

YONEXSP

01-16-07, 10:05 AM

would there be any advantage in using a 9" crt over an 8"crt in blending.

Brighter, I ran 2 8" NEC's and there is no Resolution advantage (for 1080p), but 9" ae gonna be brighter (assuming not Sony 1292's)

sarahstaar

01-16-07, 10:19 AM

i was thinking of marquee 8000 or 8500 they are bright

v1rtu0s1ty

01-16-07, 02:54 PM

Yes thats correct.
The Sony 12xx series really don't resolve 1080 lines,
as the max resolution point, when perfectly adjusted, is 750-800 lines in 4:3.
Anyways, 1920x1080 1.78 material still looks amazing in 1454x817.
I guess thats one of the compromises one have to live with when not having the budget for a Blendzilla and 2x 9" ;)
At least 40% of my 1920 movies is 2.35, with DVDs the number is close to 80%
so most of the time the movie fills the screen.
and with the limited resolution of my Sonys I simply have to live with the situation :p
And I loooove it :D

Henrik

I don't have CRT pj yet but do have 2 VGA out coming from HTPC. I was interested in Gino AUS question so I simulated it. I created a custom modeline for my linux at 1440x1080 72 hz non-interlaced for the 2 DELL ultrascan monitors. That means, I have a virtual screen of 2880x1080. I used mplayer to play a 1920x1080, 1920x817, 1920x784 mpg files. Using -monitoraspect "2880:1080" and -vf crop=1920:817 or -vf crop=1920:784, I am able to utilize all the 1080 pixel height even when playing a 2.35(1920x817) or 2.44(1920x784) like the Cars H.264 apple trailer. With 2.35 or 2.44, lesser black bars on the side than the black side bars on 1920x1080. But like what I said, mplayer's -monitoraspect does the magic and picture quality is still amazing.

I'm assuming that it would be the same implementation when I go with a real CRT pj.

Also, the metamode line makes the widescreen. The +1192 means a 248 pixel overlap for a total of 496 pixel.

Metamode "1440x1080 +0+0, 1440x1080 +1192+0"

I can take some screenshots if you want to see the 1.77, 2.35 and 2.44 on the 2 monitors.

Sinobi

01-16-07, 05:05 PM

ah hmm.
When Gino and I talk about 1454x817, we meen over the entire screen,
that is, using 727x817 on each projector and letting the projectors display black side-bars on each side of the image.
Meaning not using all the displayed pixels in the blend setup, as in down-resizing 1920x1080 to 1454x817 and displaying it on at blend-setup that is running 1920x817.

Henrik

YONEXSP

01-17-07, 09:28 AM

Damn, I'm getting the blening bug again! Does anyone have the latest on the blackmagic card experiments with Video capture?

YONEXSP

01-17-07, 09:34 AM

OK, the bug is definately Back, but what Nvidia card later than a 6800 using PCIExpress is capable of being turned into a Quadro for Blending? Or is that feature availbale for regular NVidia cards now?

I'm thinking SAT Receiver/DVD Player -----> HDCP Stripper-----> BlackMagic Intensity Card-------PC------>Software HD PPlayback? (PowerDVD ?)

racerxnet

01-17-07, 04:13 PM

slartibartfasst

01-17-07, 08:50 PM

Hi all,

Kudos to all of those pushing the blending efforts forward. I highly recommend the Blackmagic Intensity paired with a DVI repeater that will pass 1080i60 YCbCr data and VLC. VLC, if you're using a 16x9 screen, is as close to a computer video data pass-through as I've seen, and to boot it will effect a perceptually consistent blend. Further, if you take the time to learn the software (and command line program control), it can be a virtually transparent solution in the home theater environment.

I'm personally using a DirectShow blender that I've hacked together, but I'm not sure if it will port or if it's worth anyone's time to configure. The benefit of DirectShow is that you can use better scaling algorithms that those found in VLC, so for the 2.35 afficianado it's a nice choice.

Racer: If you're serious about taking the OpenGL path, I suggest you contact Prof. Cedric Cocquebert; he's written the fantastic Panoramix plug-in for VLC and might be amenable to sharing the source with you. I'd imagine that it would provide some good insights into the process of OpenGL blending.

Ken: If you have a HD-DVD player, I highly recommend trying out the Intensity. If I'm not mistaken, non-commercial (burned) HD files are sent over the HDMI connection without HDCP, so you should be able to test out the Intensity and HD acquisition without tooling around looking for a suitable repeater. I'd be very interested in hearing how you think the acquired/blended video looks through the Intensity compared to the Cyvix box you tested.

Henrik: I'm glad, to say the least, that your blending solution is working for you. The idea of using "Softlight" instead of two comparator functions ("Lighten" and "Blend") is a huge step forward; I'd imagine that it drops the CPU overhead substantially. Would you mind posting the "compensation.bmp" image that you are using? Do you have a digital camera? I'm sure it would be of interest to the whole blending community to see just what a pair of 127x can do in a blended configuration.

Chuchuf

01-18-07, 12:36 AM

I played with a Sony D50 today and if this ever gets worked out I'm thinking of experimenting with a couple of those as a blending rig.
Lets see D50 + D50 = D100........hmmmm better than a G90.......

Terry

Clarence

01-18-07, 08:01 AM

D50 + D50 = D100........hmmmm better than a G90.......Be careful with that logic, Terry...

Two 12-year-old girls would not be better than a 24 year old.
(But two 19-year-old girls vs 38 y.o., hmmm...) ;)

Sinobi

01-18-07, 10:07 AM

Henrik: I'm glad, to say the least, that your blending solution is working for you. The idea of using "Softlight" instead of two comparator functions ("Lighten" and "Blend") is a huge step forward; I'd imagine that it drops the CPU overhead substantially. Would you mind posting the "compensation.bmp" image that you are using? Do you have a digital camera? I'm sure it would be of interest to the whole blending community to see just what a pair of 127x can do in a blended configuration.
Photos of the setup will come tonight when it gets dark,
allthough I'll have to say that the compensation image is not finished yet (it is just a rough drawing that I have ended up with during trial of different methods).
As soon as I have the time I'll do it properly.

Here is a small version of the compensation for visualisation, the real bmp is attached further down.
http://www.sinobi.dk/henrik/diverse/blendcompsmall.jpg

I have edited my post on Nvidia blending explained with the following content:
---
For some reason I have to use tripple quotes around the overlay function to make MT() work.
Like this:

blend = ConvertToYV12(ImageSource("C:\compensation.bmp", fps=23.976))
MT("""overlay(blend, mode="softlight", opacity=.075, x=864)""",3)

Also I have lowered opacity to 0.075 and raised the magnitude of the compensation image.
The reason for doing this is that when you mix the movie with an image with some magnitude of gray in it, the colours of the movie becomes grayer in the mix zone.
Fx. if you set opacity to 1.00 and mix with an image of gray-level 150, then the mix zone will only contain a band with gray (150) and no movie. The lower the opacity number the more transparent the gray image will be, thus giving clearer colours from the movie. So by streaching the magnitude of the compensation image so that either the white or black compensation part reaches it's max/min value and adjusting opacity to visual optimal setting you get the least colour reduction.
---
I must say though, that the MT() function doesn't work good for me.
It actually slows the processing down some 5% on my system so that I hit 100% from time to time and the movie stutters a bit.
I suspect it's because MT() works by chopping the movie frame into smaller piezes and distributing the bits among the processors and then after, put them together again. This takes some cpu cycles and as the compensation image is very small/narrow the benefith is smaller than the cost.
So I continue to use the script without the MT() multithreading.

Henrik

Sinobi

01-18-07, 03:20 PM

As requested, here are some photos of my 1271 blending setup.
First, the projectors aren't colour/luma calibrated yet.
Second, the compensation bmp isn't the final result yet.
But here we go
All movies photographed are 1920x1080.
Photos are taken with a Konica-Minolta Dynax 7D, rotated (1.3 degrees) and cropped to edges of movie and resized to 1/4 of the original resolution.

5th Element
http://www.sinobi.dk/henrik/diverse/PICT0105c.jpg

5th element again
http://www.sinobi.dk/henrik/diverse/PICT0106c.jpg

Bourne Supremacy (this one is for Tim, who never thought it would happened (http://www.avsforum.com/avs-vb/showthread.php?p=7905596&&#post7905596))
http://www.sinobi.dk/henrik/diverse/PICT0103c.jpg

KungFu Hustle
http://www.sinobi.dk/henrik/diverse/PICT0113c.jpg

Mission Impossible 2
http://www.sinobi.dk/henrik/diverse/PICT0118c.jpg

As can be seen the right projector is a little more soft in the contrast and also a bit more green in the mid-tones.
Hopefully this will be possible to get rid of ones I get hold of Spyder2 at some point in the future.

Henrik

slartibartfasst

01-18-07, 07:20 PM

Nice work, Henrik. I found that a luma/color calibration was a great help in the final stage of tuning the compensation image; previous to the calibration I found myself using the compensation to try a correct image inconsistencies resulting from the color differences in the projectors rather than focusing on the R/G/B luma consistency over the screen. I'll try out the "Softlight" function as soon as I can.

So your image is 104" wide? Your pictures demonstrate a "punch" that I certainly never expected from an ES focusing set at that size. Further, the background detail in the "Kung Fu Hustle" shot really shows the resolving power of your array. That one really shows some depth of image, as well. Once again, well done.

Sinobi

01-19-07, 12:39 AM

Thanx.

Should have stated, 104" is diagonal
It's a standard size MDF board that I chupped off the top part to make it 2.35:1 so it's 2.44 meters/96" wide.
The board is of cause paited white, just a gain 1.0 at the moment.

Henrik

Briands

01-19-07, 11:00 PM

Could you take a picture of the Disney castle logo or the fbi warning, or rating screen. I think that was the one that Cliff had the most trouble with.link (http://www.avsforum.com/avs-vb/showthread.php?p=9127755&&#post9127755)

Also, my preference would be something in the driver so that it would work for anything on the computer display, not just video (as the VLC and direct show). I don't think I would pass WAF if it can't work with a front end as well as video sources. And the WAF is required to spend the green on PJ #2.

Sinobi

01-20-07, 03:46 AM

Could you take a picture of the Disney castle logo or the fbi warning, or rating screen. I think that was the one that Cliff had the most trouble with.link (http://www.avsforum.com/avs-vb/showthread.php?p=9127755&&#post9127755)
I'll do that,
but a week may pass before I'm there, as I don't have time untill next weekend.
Also, my preference would be something in the driver so that it would work for anything on the computer display, not just video (as the VLC and direct show).
I fully agree.
but a directshow solution would be second in line.

Henrik

v1rtu0s1ty

01-20-07, 01:07 PM

Here is a small version of the compensation for visualisation, the real bmp is attached further down.
http://www.sinobi.dk/henrik/diverse/blendcompsmall.jpg
Henrik

Hi Sinobi,

I would like to know more about the blend that we add. However, I'm actually working for the linux and they like the idea. I don't want to promise anything yet. Anyways, looking at some of our folks actual screenshots like those 60% left and 60% right, I noticed that the their video goes from with light to total black, it's like transparent to black gradient. Looking at the bmp that you attached, the left and right edges has color grey. I would just like to know why there should be color grey. I'm very newbie in this area and I'm just looking at the 60% left screenshots of others. So by looking at their screenshots, I think(but maybe quite incorrect), that overlaying the blendmask I attached will work. The attached is just a simulation since it is jpeg. We will be using png-24 with transparency or probably pure pixel coding with alpha.

Oh, and also, I was basing the blend mask displayed at http://local.wasp.uwa.edu.au/~pbourke/colour/edgeblend/

If that will work, I will be able to explain properly what we need to do.

Please let me know or anyone.

Thanks!

Awesome screenshots!!!!

Sinobi

01-20-07, 04:05 PM

Anyways, looking at some of our folks actual screenshots like those 60% left and 60% right, I noticed that the their video goes from with light to total black, it's like transparent to black gradient.
It's more like, fading the image to black, meaning reducing the RGB values more and more the closer the pixel is to the edge. Nothing about transparent or anything, just reducing luminosity.
Looking at the bmp that you attached, the left and right edges has color grey. I would just like to know why there should be color grey.
Have you been reading this thread?
That is what this specific thread is about.
It's about that Nvidia incoorporated a blending solution in there Quadro drivers,
but didn't do it properly so the blendarea has a bright stripe in the middle with darker areas flanking the bright stripe on both sides.
Beside Nvidias solution there is only 2 other choices (and they actually work correct), namely VLC with a special plugin, and an Avisynth script that is too demanding on the CPU to be used for HiDef.
If one prefere to use Zoomplayer or MediaPlayerClassic and Ffdshow one has to live with the dark-bright-dark area in the middel of the flick caused by Nvidia's lack of skill
OR
do a inverse compensation to the image so that the dark-bright-dark area is corrected with a bright-dark-bright area.
Now, ones again look at my compensation image do you see it's bright-dark-bright?
The middel gray means no correction, brighter than gray means - make the movie brighter, darker than gray means - make the movie darker.
This way we get rid of the imperfections in the Nvidia divers blending option.
So you see the grayish image is a correction, NOT a blending curve.
But most of this is deeper discribed on former pages in this thread, amongst other places also in this post (http://www.avsforum.com/avs-vb/showthread.php?p=9454348&&#post9454348)

Awesome screenshots!!!!
Thanx, will be posting more, ones I'm finished with the correction and maybe adjusting the projectors for luma/colour imperfections.
:o Next time I might even remember to do a white balance correction on the camera :rolleyes:

Henrik

v1rtu0s1ty

01-20-07, 05:02 PM

It's more like, fading the image to black, meaning reducing the RGB values more and more the closer the pixel is to the edge. Nothing about transparent or anything, just reducing luminosity.

Have you been reading this thread?
That is what this specific thread is about.
It's about that Nvidia incoorporated a blending solution in there Quadro drivers,
but didn't do it properly so the blendarea has a bright stripe in the middel with darker areas flanking the bright stripe on both sides.
Besides Nvidias solution there is only 2 other choices (and they actually work correct), namely VLC with a special plugin, and an Avisynth script that is too demanding on the CPU to be used for HiDef.
If one prefere to use Zoomplayer or MediaPlayerClassic and Ffdshow one has to live with the dark-bright-dark area in the middel of the flick caused by Nvidia's lack of skill
OR
do a inverse compensation to the image so that the dark-bright-dark area is corrected with a bright-dark-bright area.
Now, ones again look at my compensation image do you see it's bright-dark-bright?
The middel gray means no correction, brighter than gray means - make the movie brighter, darker than gray means - make the movie darker.
This way we get rid of the imperfections in the Nvidia divers blending option.
So you see the grayish image is a correction, NOT a blending curve.
But most of this is deeper discribed on former pages in this thread, amongst other places also in this post (http://www.avsforum.com/avs-vb/showthread.php?p=9454348&&#post9454348)

Thanx, will be posting more, ones I'm finished with the correction and maybe adjusting the projectors for luma/colour imperfections.
:o Next time I might even remember to do a white balace correction on the camera :rolleyes:

Henrik

I can make the image look like this but what do you think when it is projected to a real projector? I was told that black is absence of light therefore it will not be displayed. And if you look at the picture, they overlapped. By the way, I just did that in photoshop.

http://img262.imageshack.us/my.php?image=xinermabutnoblendmasktr5.png
http://img158.imageshack.us/my.php?image=xinermawithblendmaskha2.png

Chuchuf

01-21-07, 12:09 AM

So as I see this progressing, it would seem that the only viable solution if one want to run Theatertek or say PowerDVD 6.7 for say HD DVD playback, then one is limited to using a Quadro video card OR soft modding a 6800 video card in order to use the "broken" blending that NVidia provided for in their Quadro drivers and then running "compensation" which removes the thin vertical line. Is this correct so far?
If so, how well (if at all) do these video cards function with PowerDVD's HD DVD solution?
Playing back DVD in a blended 2.35 screen sounds good, but now that we are seeing the availibility to playback HD DVD using the XBox360 drive and a PC running Power DVD 6.7 or 7, I would think that would be an ideal 2.35 blended application.
I can tell you that I don't believe the 6600GT card will not play back HD DVD and I'd doubt the 6800 will either.....but I could be wrong.

Terry

v1rtu0s1ty

01-21-07, 01:02 AM

I can tell you that I don't believe the 6600GT card will not play back HD DVD and I'd doubt the 6800 will either.....but I could be wrong.

Terry

Don't know if this will help but I have a 6600GT and I can play 1080p trailer files without any frame drop issue using mplayer for linux via its H.264 decoder filter. From what I know, it's actually the CPU that is responsible especially with helping of the decoding process. Even other guys are able to play H.264 1080p files using FX5200. That's $31 card. So do we have to buy an expensive card like those 8800? That's when those high end card becomes real winners especially when using it for high intensive graphic games such as Far Cry since these cards does a lot of vector computation. Our HDDVD, HD broadcast, DVDs are just static frames.

Here is a link (http://wiki.multimedia.cx/index.php?title=Features) to the types of HD DVD/BluRay codecs.

I captured my 2 monitors in action that's using xinerama with overlap(preparing for edge blending :) ) and I played Cars 1080p(H.264 encoded) apple trailer. I think, I also played 24 in HD broadcast. Here is the link (http://www.4290greenfieldlanehome.com/xinerama.mpg) to the mpg file.

So yeah, the 6600GT will be fine but make sure you are using a decent CPU. Here is the spec (http://restricted.dyndns.org/pvr.html) of my mahcine. Mine is an Athlon 64 3000+ 1600MHz.

EDIT:
This is a nice link (http://ati.amd.com/products/pdf/H264_Whitepaper.pdf). It explains that CPU does the decoding. However, this is good news for PC users like us. There are newer cards that have H.264 decoders built in. :D

Chuchuf

01-21-07, 10:35 AM

v1rtu0s1ty,
I know that the 6600 and 6800 are certainly capable of doing 1080P material from a number of Codex, but that isn't the question here.
What I think is the case is that PowerDVD 7.7 or 7.1 (playback software for HD DVD's) won't allow for the playback of HD DVD using those cards with the NVidia drivers that support HD DVD playback? I just don't know because all the testing I did with PowerDVD and HD DVD playback was done with cards like the 7800 or 7950.

Terry

YONEXSP

01-21-07, 01:05 PM

I think 6600GT should be OK, but I believe the 6800 had a bug that did not allow h.264 etc playback

Sinobi

01-21-07, 03:56 PM

I think 6600GT should be OK, but I believe the 6800 had a bug that did not allow h.264 etc playback
No but the 6800 has a bug that prevents it from being able of h.264 hardware acceleration.
Playing h.264 material is possible, no problem, you just have to do all decoding in software codecs like CoreAVC or Ffdshow.

The 6600 has no problem with h.264 harware acceleration that I'm aware of.
It just can't be converted to a Quadro card.

Henrik

Briands

02-16-07, 10:55 PM

Seems AnyDVD has an HD version now that is allowing HD DVD playback w/o requiring hdcp hardware. Does this help the HD blending prospects?

overclkr

02-16-07, 11:16 PM

Seems AnyDVD has an HD version now that is allowing HD DVD playback w/o requiring hdcp hardware. Does this help the HD blending prospects?

Please let this turn into a STABLE WORKING app. This is too damn cool........

Cliff

Sinobi

03-02-07, 07:13 AM

Why don't we use the Nvidia SDK software and write the blend program in openGL to interface with the 7, 8, and Quattro series cards?? I think this would be an easier way to provide the desired results, without the hassle of several components to provide the end result. :cool:

I will be looking into this and post on progress made.

MAK
The lack of post...... is that the result of -no progress- being made?
The solution you talk about sound (for me as a non programmer) like the way to go.

Henrik

nashou66

04-10-07, 01:28 PM

Has anyone doen this on a mac computer? I hate pc's :-D

Athanasios

YONEXSP

04-10-07, 02:18 PM

Has anyone tried the latest Blending drivers for Quadro from Nvidia?

MTyson

04-10-07, 03:19 PM

Has anyone tried the latest Blending drivers for Quadro from Nvidia?

I'm very interested myself. :D Of course it seems too good to be true that it would actually work, so my guess is that it won't, but hopefully someone will try.

Sinobi

04-14-07, 07:15 AM

I'm using the 91.36 version (as far as I remember).
I've tried newer ones, but they have changed the driver user-interface to be more userfriendly for non tech people (wizards for almost anything) so now we don't have as much possibilities as before and sadly it looked like they had removed the blending feature.
So for now it looks like the 91.36 is the newest version that can be used (with all it's faults).
Sadly that still requires the overlay script in avisynth to get good results...

3 forum users have talked about programming a blending solution.
And did get some of the way through it.
Haven't heard anything more from them though...

Henrik

mtmelvin

05-01-07, 09:25 PM

I don't want this thread to die!!! I really love to see a simple, elegant solution for PC blending. Where are the guys that talked about programming a blending solution?!?!

Right now I'm dreaming of dumping the G90 and using two G70's instead. I've already got one and they're getting pretty easy to come by.

-Mark

nashou66

05-01-07, 11:01 PM

we all came over here !, well some of us......

http://www.avsforum.com/avs-vb/showthread.php?t=832362

"If you guys know so much about women,what are you doing at a gas and sip ,on a saturday night with no woman to be seen?"___Loyd Dobbler

Athansios

mtmelvin

05-02-07, 12:13 AM

Yeah man, I know about that thread. But over there you're hoping to successfully blend for under $6000. Over here I'm hoping to blend with the PC I already have. So there's about a $6000 price difference ;)

I'm following that thread too. It's still a little out of my price range, but it's interesting. The TV One units sound promising. Get it under $3k and I'm on board :)

we all came over here !, well some of us......

http://www.avsforum.com/avs-vb/showthread.php?t=832362

"If you guys know so much about women,what are you doing at a gas and sip ,on a saturday night with no woman to be seen?"___Loyd Dobbler

Athansios

Sinobi

05-02-07, 08:22 AM

Well, there IS a little movement on the pc front.
Yesterday I suddenly discovered that Ffdshow has exactly the same function as the Avisynth script I use for correcting Nvidias flaws.
It's called "Bitmap Overlay" and while the Avisynth script takes a considerable amount of cpu power,
the Bitmap Overlay takes close to nothing, so little that I can't se it in the cpu load graphs at all.

Henrik

mtmelvin

05-02-07, 10:00 AM

That's good news. Does that make it a little easier to implement the overlay? I would think so, because many people (like me) are probably not familiar at all with Avisynth.
Of course, using less CPU is a good enough reason already.
Nice work.

-Mark

Sinobi

05-02-07, 11:32 AM

There is still the work to do with making the correction image "match" the Nvidia flaw.
But beside that it's click'n'drag stuff.

Load in the correction image.
All position and alignments should be set to center,
and mode should be set to softlight.
Then it should just be a matter of adjusting Opacity to best suitable position, around 5-7%.

Henrik

Jamez

12-25-07, 07:55 PM

bringing up this old thread.
Have anyone of you tested the lates drivers from Nvidia?

This application note seem to be from June this year.
http://www.nvidia.com/object/IO_29134.html

from the figure on page 7 it doesnt seem to have a linear luma roll-off?
Maybe they've upgraded with more correct algorithms?

I dont have two projectors or a graphic card that would work, yet.
But I def plan on getting :)

will be starting out with a dual nec 9pg, and an quadro fx3450.
Dont exactly know what solution I would be looking at using, but I'm sure something can be made to work. Like you guys here have already shown :)

MadMrH

12-26-07, 03:43 PM

About 18 months ago I wrote the following.....

http://www.avforums.com/forums/showthread.php?t=352438&highlight=blend

There are some short films of my Nvidia blend in the post.

Also search for blend whilst you are there, thats my "home" forum and there is lots there.

Andy.

Jamez

12-26-07, 04:54 PM

very nice Andy :)

The short films actually show the problem very good.
But I was hoping the new drivers might be better...

I just won auction for a quadro fx 3400 card, 112€.

If I can just get another pj now, I could test it out in some time :)
if it doesnt work, I'll figure something else out...

Sinobi

12-26-07, 07:03 PM

I havn't tried out the new drivers yet,
and although the picture you refere to looks promising,
I doubt it holds any changes to the blend feature.
But the compensation solution that this thread is about works pretty well.
Not perfect, but pretty well.
I have done a color/light/grayscale calibration of my projectors recently and
I still optimize the compensation image from time to time and it is still getting better.
I am sure it will never be perfect, but it's an acceptable solution for a poor mans blend.
The error is slightly visible in 10-20% of a movie with a small error in very dark parts of a movie and a larger error in very bright parts, but most of the time it's invisible.

Henrik

Jamez

12-26-07, 08:13 PM

ElTopo

12-27-07, 09:59 AM

ElTopo

12-27-07, 01:02 PM

For those who want play a little bit with the control panel of the C2-260 unit of TVone.

http://tvone.crmdesk.com/answer.aspx?id=169&back=faq.aspx%3F

There's a demo mode but you can see all the features, and there a lots..

ElTopo

Sinobi

12-27-07, 02:45 PM

I've been in communication with a fellow that has been trying out the c2-260.
He is NOT that impressed with the results, I am not saying that it is not good enough, but he didn't find it to be the perfect way to do it.
If it was the controlpanel not being adequit or what it was, I don't know, but he was not satisfied.
They are chep, but may be not good.....I'm not sure.
Would be nice to hear a second opinion....
As I understand it you need a scaler or a quadro to do the split of the image before sending it to the 260's, and then you have fair chances of making a fair blend....
As I state; I havn't tried this out.....only heared second opinions about it.

Henrik

Sky042

12-27-07, 07:47 PM

Please let this turn into a STABLE WORKING app. This is too damn cool........

Cliff

I'm sure you guys know it by now but ANYdvd is dead stable now and using it with powerdvd you can play commercial HD-DVD's over non HDCP displays.

alan halvorson

12-27-07, 10:50 PM

As I understand it you need a scaler or a quadro to do the split of the image before sending it to the 260's, and then you have fair chances of making a fair blend....

The C2-260 is a scaler. The full image is input into the C2-260 and then it blocks the image per user settings and scales the result. There is a windows control program that can be loaded but I'm not quite certain how it interfaces to the C2-260.

Firmware, however, is supposed to be the same as for the more expensive, standalone models and MadMrH seems to have found that it's all good. It would be interesting if your friend would post about his trials and tribulations with this product.

Sinobi

12-28-07, 01:38 AM

As I understood it:
"you have to feed it a signal of 108 MHZ maximum and it does not like more than 1280 horizontal pixels on the input side"
That mean you would have to split the image before sending it to the 260's

Henrik

ElTopo

12-28-07, 02:11 AM

Here's what the manual says:

For use with dual-channel scalers such as the C2-7000 series, connect your video source (e.g. a DVD player or computer) to an input on the C2-7000 unit. It is not necessary to feed two signals – the C2-7000 unit can use the same signal for both outputs.
For 2 single-channel scalers, you can either use a distribution amplifier so that a single video source is split into 2 signals, or use the ‘loop’ output available on some units.

So you have to split the picture before.

ElTopo

Oliver Klohs

12-28-07, 04:36 AM

Hello,

I have been waiting to post anything about the C2-260 and its limitations but after I gave them my input and offered help with further product development I never got any feedback from them so apparently TV One just wants to sit tight and not do anything about their products weaknesses. In order to prevent others from going through the same problems I have been going through here is a short report:

I have tested the C2-260 and it has several limitations that negate the possible advantages of a blended setup.

Let's just say that it is not possible to arrive at a decent 1920 x 1080 desktop with it, no 1:1 pixel mapping is possible and the a/d conversion in the input section seems to hurt picture quality a lot when viewed on a large screen.

The Geforce solution is glitchy but at least it is transparent once it works obviously minus the problems in the blend zone and problems that stem from a PC solution (I could for example not get judder free playback)

I have tested this with a blended Setup of one NEC 9PG plus and a 9 Pg extra, both with almost perfect tubes and good raster usage. Let's just say the results where such that I could have gone back to 8" AC instead of 9" LC - two of these certainly top the resolutions capabilities of even the best 9" units :)

Here is a pic of the blend zone in the Geforce setup where it hurts the most, at the bottom of the screen which is in the outermost corners of the tubes. These are alternating lines of 1, 2 and 3 pixel wide.

http://img181.imageshack.us/img181/7980/blend070912008ml4.th.jpg (http://img181.imageshack.us/my.php?image=blend070912008ml4.jpg)

This is on a curved screen and not shot head on, all the lines looked like they do on the left of the pic, but I could not capture it that way due to camera placement and screen curvature.

Oliver

MadMrH

12-28-07, 07:51 AM

Hi all,

will go with the Quadro FX3400 and two brand new BG808s LC units.

I'm new to the blending theme but i have to take that challange :D

@Andy: You only have fed the two outputs from your Quadro FX to your Barcos ? Any add. SEMU units inside the Barcos ?

ElTopo

If you dont have the 808LC's already I'm selling my 1209s Graphics pair fully loaded with COG, Contrast Mod, Semu Advanced 7.22 firmware (7.23 if I get round to doing that)......

Any way - I have the SEMU systems in a number of projectors here, as its not that wasy to get them I decided to go with a more main stream system - for me the TVONE product.

So YES my 1209s have SEMU but NO i'm not using that system.

MadMrH

12-28-07, 08:36 AM

There is a windows control program that can be loaded but I'm not quite certain how it interfaces to the C2-260.

RS232 connection from PC to setup once set then can be left stand alone.

OR you should be able to control from a "remote rs232" system Crestron etc. I have not done this yet.

Firmware, however, is supposed to be the same as for the more expensive, standalone models and MadMrH seems to have found that it's all good. It would be interesting if your friend would post about his trials and tribulations with this product.

The "main brain" is the Corio2 - so this is really an amazing product.

I have three 260 units.

There are of course limitations with the 260 but what do you expect for a product of that value? its NOT a DiVentix or a TVONE C2-7200 - If you want those products then thats EXACTLY what you buy.

What this product seems to be is the closest to an all in one RGBHV in/out dedicated blend system.

The thing is it is NOT just about the product - The installation and intergration of the product with system setup are ALL major parts of the completed blend system.

alan halvorson

12-28-07, 09:06 AM

So you have to split the picture before

The signal is not split, it is duplicated. That's what a distribution amplifier does. The full picture is fed to each C2-260, then you block it as you see fit and the result is scaled. The C2-7xxx series does not need a distribution amplifier as they include dual scalers.

MadMrH

12-28-07, 11:24 AM

I have been waiting to post anything about the C2-260 and its limitations but after I gave them my input and offered help with further product development I never got any feedback from them so apparently TV One just wants to sit tight and not do anything about their products weaknesses. In order to prevent others from going through the same problems I have been going through here is a short report:

I have tested the C2-260 and it has several limitations that negate the possible advantages of a blended setup.

Let's just say that it is not possible to arrive at a decent 1920 x 1080 desktop with it, no 1:1 pixel mapping is possible and the a/d conversion in the input section seems to hurt picture quality a lot when viewed on a large screen.

Hi Oliver,

Send me a PM or email with your complete report on issues with the 260 product.

You only have to look down this years firmware releases for the 7200 and software to see how much work TVONE have done to perfect the blend and product. I send a full report on a regular basis to TVONE and have to say that to date 90% of my requests have been actioned, my final request is currently being considered I need to refine my request.

I have found TVONE to always be 100% responsive to my requests, Of the 19 points in the latest firmware list here are the ones I requested.

1. *Extra edge-blending option, providing 'side' adjustment where 2x2 (or more) blends are overlapped.
2. *Improved edge-blending guide-lines - more accurate.
3. *Fixed bug where macro store of output resolution would not store properly.
4. *Increased button hold time for Factory Reset using RESTORE button to 5 seconds (was less than 1 second).
5. *Wider range of gamma values now available when edge-blending (from 0.01 to 1.50in 0.01 steps).
8. *Changed adjusting routine so that brackets stop flashing (i.e. can no longer adjust) after 20 seconds of inactivity.
9. *Added menu item for 'Aspect change'.
11. Backlight on units with LCD can now be disabled.
12. Fixed issue where C2-7000 backlight could not be fully dimmed to 'off' state.
13. Added 'On source loss' to 'Adjust sources' menu, so that user can define what should happen when a source is lost. Default is blue screen. This replaces the 'De-glitch' menu item in 'Adjust windows'.
Added to / altered resolution database file. Interlaced resolutions now show as field rate and not frame rate. (e.g. 1920x1080i 29.97Hz now shows as 1920x1080i 59.94Hz.)

Known Issues:
1. Edge blending settings are not restored on power up.
(The above Known issue has also been resolved now)

Again send me PM or email with list of issues I can try and replicate.

YONEXSP

12-28-07, 12:02 PM

I'm also wondering if a softmodded geforce 6800 really will do the same as a 'real' quadro card. Seem to good to be true.
But for making the blending option possible, its great :)

the compensation method are elegant and simple... I really like it.
But then again, some progressive algorithm that would give a different compensation for brigth and dark scenes, and with full color adjustability would be even better :)

I'm not into the Avisynth scipts and all. But it should, even there, be possible to use different blending out from different intensity levels in the shown image?

It's true works like a charm.

Though because of the bright line issue witht eh Quadro drivers when doing the blend, I just used it for the overlap control.

With the external xed.1 box the results were amazing.

Though by now I was hoping for a Nvidia card with onboard HD support that can be softmodded, so I can use my old PC.

18 moths ago the xed.1 was $2500. Maybe they still have some old units available, if they were under a $1000, we would have a working solution with practically perfect results for $1500.

The arrival of AnyDVD & PowerDVD have made the difference.

K./