Increased resolution and colors on the prop?

DaveF

Hey all,
With all of the recent advances (increased memory capacities, higher resolution VGA driver, etc, I was wondering if it is now possible to increase the resolution of the composite (TV) output?

I am working on a product based on this, and have recently been thinking that the default resolution of the prop might not cut it....

Anyone?

If not, does anyone know of a chip that could be used in conjunction with the prop to display higher resolution images/animation?

Thanks!

Dave

AndreL

Its not the resolution of the prop, but the resolution of the TV, its finite and about 160-224 color clocks depending on the right/left overscan etc. The HYDRA demos and book show how to get 256 colors on the screen at once, so its possible to do with 256x192 resolutions, but very tricky programming is needed and 4-6 cogs to make it happen depending on what you want to do (tiles + sprites, scrolling etc.). But, you are never going to get VGA resolutions on NTSC unless you are dealing with a WEGA system. Another confusing thing about video is you might have a system that outputs 640x480 or higher that outputs to the TV, this signal is DOWNsampled to fit in NTSC, so the larger information is "there" in a sense, but its been blended into the surounding pixels, this is how a VGA to TV converter works, pixels are sampled with a 3x3 grid or larger matric with some blending function or convolution kernal then used as the output for the sub-sampled video stream.

There is a video floating around here with video from the HYDRA demos to give you an idea of the limits of the propellers graphics on NTSC if you can find it. Also, the HYDRA goes on sales tommorow and once the webpage is up, there is a link to the video plus a lot of screen shots to give you an idea of what you can make the prop do. I am sure someone will put a thread tommorow or saturday once all the hydra pages are up and its for sale, so you can get into the page(s) to see.

Andre'

DaveF

Hey Andre!
I have been reading a lot about you and the Hydra today. I must say, you got yourself quite an impressive resume! And even a bio on Wiki!! Nice!

I did find the video you are referring to earlier today. It shows a bunch of different games that were produced on it. They look cool!

I am not trying to create a game, but am interested in making a display that would be appealing to the eye. What I want to get away from is jagged edges in the graphics. I think that may be a problem with the resolution and color depth of the prop... Especially on larger screens (NTSC).

My application is not ment for VGA, and it close to useless if you are that close to a computer!

Maybe if there were a way to raise the amount of colors.... or display a static image that has high depth, and overlay text. That might work.

AndreL

Basically, all you need is 256 colors, and good graphiics that have smooth transitions, NTSC by its very nature anti-aliases due to bleeding, so you get free DSP from the analog nature of the signal and the TV rendering [noparse]:)[/noparse] So 256x192 bitmapped will look more than good, its just a matter of making good graphics. Of course, if you want really nice graphics that look photoreal, its not the color depth that kills you, but the shades, your brain can pick up such slight changes in shades, its ridiculous how much dynamic range your eyes have with shades. But,·with the prop you get about 6 shades or so of the colors, so not a lot of·range and banding is very apparent when representing non-artificial images like people, houses, whatever.

But, the bottom line is if you are rendering on large NTSC screens, the trick is to run a low pass filter on the image to slightly blur it, or at very least run the filter on high frequencie edges where the color/intensity changes rapidly, these are most noticeable to your eyes. You can make things LOOK REALLY good with NTSC if you just know a little about computer graphics and art, for example, in the 3D racing demo on the HYDRA, the signs where white with black letters saying HYDRA, this was really bad since the transition from white to black is basically a infinite frequency spike in the image space domain, but just by running an average filter on the image, it anti-aliases the image and it looks much better when rendered and scaled.

finally, you may want to look into "Mip-Mapping", for example, as you images get larger (if they do such a thing), you use lower frequency versions of the images whne they are smaller, the higher frequency images when they are closer. This minimizes pixel swimming and sparkles as the bitmaps are sampled and rendered.

Lastly, there is no reason why you can't create an interlaced display with the prop and get 525 lines of video, but interlaced looks TERRIABLE unless the image is moving, you will see it flicker and shake, so interlacing looks good ONLY with images moving with high color content and variation.

Andre'
·

DaveF

Yes, the limited number of colors (shades of the primary colors) is a problem with trying to make more real life looking images.

I guess there is no chance of getting more colors or shades out of the prop?

Anyway, It looks like I have a LOT to learn about how NTSC video works! I am happy to see that there are options to clean up the look. I can understand how a blur effect would help on larger screens.

I am definately going ahead with the prop anyway as the first phase of my project works fine with this kind of graphics

Thanks again!

Mike Green

Yes, there realistically isn't any way to get finer control of color out of the prop without adding significant external hardware and, even then, you're "pushing the envelope" of NTSC video.

DaveF

I am a bit confused...
I am not trying to do anything you don't already see on a tv. How many colors is a tv capable of? 16 bit?

AndreL

TV is analog and has nothing to do with bits, its capable of millions if not billions of colors, the voltage that drives each gun is analog in nature, there is some "resolution" to how fine each beam can be driven and the dynamic range of the phosphors and the finite energy level states of the phosphor the phosphors are ultimately quantum in nature and the intensities jump per each next energy level of bombardment, but we are talking about a LOT -- more than you need. The problem mostly is the bleeding pixel to pixel and other effects that decrease the contrast to a point where you don't get really crisp images. Now, of course on LCD screen etc. they ultimately are digital in nature and do have finite color abilities 8, 16, 24,32 bit etc.

But, the latest greatest graphics cards do color math in 128 and 256 bit widths to give you an idea of where things are going, that's a lot of color!!!!

Andre'

ciw1973

When it comes to VGA there are ways to get the maximum number of colours (64) out of the demo board hardware by using several cogs working in unison. With output to a TV, because of the way that the composite signal is produced, you are limited to the cog's native capabilities of 4 colours per 16 pixel wide piece of the screen.

OK, so that may not be a problem for your particular application, and getting 256 colours on screen at once isn't a problem either but as there can be no more than 4 colours in any 16 pixel horizontal block, you need to thinking about what you are displaying and where on the screen. At a basic level, the TV output from a cog isn't really any better than that of the Spectrum or C64, but there were some great looking games on these machines, you just need to be clever and work to try and get the best out of the hardware. It was the hardware limitations that pushed the artists and programmers back in the 80s and early 90s to create the great games they did. We've got lazy over the years working with PC hardware, and for me at least a return to our roots is long overdue.

LoopyByteloose

Andre mentioned shades.
It might help you to understand that the eye is composed of two types of cells -- rods for B/W and cones for color. One type sees shades and the other sees color. We all have many more cells that see shade than color. This adds to our night vision and demonstrates why B/W came along before color.

Additionally, in a 4·point / 2-axis·color circle, we tend to see more degrees of difference with reds and greens as they are in the mid-range between black and white. Blue is near the black extreme and yellow is near the white extreme when viewed as a pure color or greatest intensity.

I was quite surprised to hear Andre say that only 6 shades were available in the current software. It appears that the engineers have not looked into the physiology of the eye and just thought that the range of color is more important. Seems we have another frontier for the Propeller to explore.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"If you want more fiber, eat the package.· Not enough?· Eat the manual."········

···················· Tropical regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan

Post Edited (Kramer) : 11/17/2006 9:19:03 AM GMT

DaveF

Kramer,
That sounds like a volunteer/challenge to me! Are you going to be the one to break doen that barrier for everyone and be the big hero???

[noparse]:)[/noparse]

I for one would LOVE to see you do it!

Dave

LoopyByteloose

I wish I knew enough computing to handle the challange. I graduated with a degree in Fine Art - hence the knowledge of color. Video color output is currently way beyond me.

This might help someone get started.

If you look at the visible light spectrum [noparse][[/noparse]Red-Orange-Yellow-Green-Blue-Indigo-Violet] as a straight line; you may notice that red and green are not adjacent and neither are yellow and blue. In fact, they are considered opposites [noparse][[/noparse]or compliments] on a color circle. This fact seems useful as the two pairs are also 90 degrees out of phase.

So it seems that a curve similar to a Sine Wave might mimic the range from light to dark on a straight line of pure hue at optimal intensity. Let's say white is positve, black is negative.

Pure Red and Green would be at or near the crossing of the axis. Pure Blue is closer to Black and Pure Yellow is closer to White. On a grey scale, the axis would represent a midway between pure black and pure white have a value of 0; pure black would have a value of -1; and pure white would have a value of +1.

If the distance from the axis is logrithmic and normalized to the human eye rather than linear value, you might find a model that would fit for efficent computing.

You would have one variable for the amount of black or white that would produce a value for grey content.

And you would have a second value for hue.

You combine the two variables to get a specific color.

The two components are already recognized by others. But the big difference would be to have the scales normalized to linearly mimic the response of the human eye. So you could forget about having 64 colors if only some of them are useful as being visibly different to the human eye.

We already do this with db and audio volume and the Richter scale with earthquakes. And the Propeller does logs!

Hopefully I am making sense. It the meantime, I think I better shut up and just read more of what others are contibuting.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"If you want more fiber, eat the package.· Not enough?· Eat the manual."········

···················· Tropical regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan

flapdoodle_dinghy

I experimented with 16 color mondrians on the Commodore 64 based on work by the Polaroid Corp. research. Polaroid discovered that adjacent colors affect how the brain interprets the color of a pixel. Other major corporations picked up on this.

This is probably more than most people want to know about how mondrians work:
http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JEIME5000011000003000293000001&idtype=cvips&gifs=yes

Color Management Concepts has an extensive list of applications: http://72.14.209.104/search?q=cache:UBJbm63eensJ:download.microsoft.com/download/1/6/1/161ba512-40e2-4cc9-843a-923143f3456c/ColorMgmtConcepts.doc+polaroid+adjacent+color+research&hl=en&gl=us&ct=clnk&cd=9

It is used in digital cameras, printers, monitors and HDTV.

Adaptive fuzzy color interpolation or color demosaicing here:
http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JEIME5000011000003000293000001&idtype=cvips&gifs=yes

Has an algorithm for smooth color transition.

AndreL

The reason why there are only 6 shades is that there is only a 3-bit DAC generating the analog signal, this was just how chip gracey designed the propeller. So with those 3 bits you have to represent black and sync, and have enough room for the color signals to modulate on the luma +-1, so 6 is actually not true, its more like 5, but in some cases you can get 6 if you do it right.

Andre'

Phil Pilgrim (PhiPi)

DaveF,

With composite video, you actually can increase the number of intensity levels by superimposing DUTY-derived grayscale output from a separate cog onto the video signal. But you don't get realistic color this way, since there's no way to adjust the saturation (except for using the "super-saturation" dodge). What results looks more like a tinted black and white photo than true color. Also, you're pretty much locked into the sixteen available hues, since they derive from the sixteen available phase shifts in the chroma clock.

-Phil