p. not so great 2D/TVout quality (c't review)

[xortam]

DVI matters less and less if you're not allowed to display the full resolution.

[Greebe]

or color spectrum

[KvHagedorn]

ah but the color spectrum is infinite.. even if we had some supercard with 256 bit color we would still not have ALL colors...

[Dr Mordrid]

Yes it is infinite, but the human eye's ability to discerne all those colors is not. Most research indicates we can perceive about 10 bits worth of color, dispite what Carmack deems "necessary"

A quite good article on Parhelia's color system vs. human perception is on AccelerNation.com;

http://accelenation.com/?doc=145&page=1

and they are spot-on about 10 bit color.

Dr. Mordrid

[Yannick]

It's worse than that. Color is not a combination of three values, it's a continuous function across the interval of visible light, and we see three zones of it on three different series of captors... and IIRC our computer screens only try to match those three zones as accurately as possible.

Now there's the problem of all those people that do not see the same three zones as the average population... and that includes 8% of the male population, IIRC...

Now that's a challenge...

[Jon P. Inghram]

Yes, I demand all new video cards implement analog color! The display should be represented as a mathematicly defined waveform, not an array of pixels.

[DGhost]

Originally posted by Dr Mordrid
Yes it is infinite, but the human eye's ability to discerne all those colors is not. Most research indicates we can perceive about 10 bits worth of color, dispite what Carmack deems "necessary"

A quite good article on Parhelia's color system vs. human perception is on AccelerNation.com;

http://accelenation.com/?doc=145&page=1

and they are spot-on about 10 bit color.

Dr. Mordrid

i believe what Carmack was refering to is mostly a higher precision rate when dealing with the internal calculations so that you don't "loose" accuracy when dealing with several light sources + fog + alpha channels being used in the scene, not so much the frame buffer. from what i understood, it was a problem to the developer and the program, not so much what the person sees on the screen...

kinda like 3dfx's '22-bit' color thingy...

[Yannick]

Originally posted by Jon P. Inghram
Yes, I demand all new video cards implement analog color! The display should be represented as a mathematicly defined waveform, not an array of pixels.

If you want to represent the true color, you'll need even more, since the color for each dot is a function. Which means, if I am correct, that a completely analog color image should be defined as a function defined on a 2D plane and having its values in a space of functions...

Now what 10 bits are these ???

[Wombat]

Bah. How about "indistinguishable by human eyes?" That should keep you happy, and your 32bpp display covers that pretty well except for parts of the green spectrum.

[bobby2]

How the human eye works, never mind how monitors represent signals coming from computers.

Our eyes have no stepping - as pointed out - analogue is infinite. 10 bits is just more colour per channel - 256 for RGB with 24 bit, and 1024 for each channel (RGB) for 30 bit - or:

GIGACOLO[U]R - 1024x1024x1024 = 1 Gig

Matrox provides us with more steppings in our gradient (4x more), which gives smoother representation, and indeed, better representation of colour space.

Our eyes perception of colour works in a very different manner - cones and rods on our retina percieve colour and brightness components - the colour coming from the analogue spectrum, and the brightness coming from the intensity of the waveform.

Yes, research into our ability to percieve has shown that 10 bits per channel is about maxing out - but its limited to a static saturated test zone which doesnt measure how we percieve levels of saturation in bright/dim components - thus it can compress to very few colours, or expand to a colossal 16+ bits in 'perfect' levels. Count how many colours you can see in dim light/ bright. Same scene ... The reaction speed to levels also affects the number of colours we see.

We have a staggering range in the brightness component - but its not a linear or indeed mathematically simple curve on how the eye responds - imagine the lights going off in a dark room - it takes time to adjust - light to dark and dark to light.

We also see reds/ greens/ blues with varying degrees of quality, and some women have a chromosome that allows them to percive more colour (it actually might be that men evolved to see less colour to facilitate hunting/ + illustrate poor taste in clothes while women evolved as gatherers - colour being important in picking fruits etc)

Carmack was talking about the usable leftover bits from 32 - 8bit RGB leaves 8 bits. 10 bit RGB leaves only 2.

What Gigacolo[u]r provides is gorgeous non crappy 1024 components for each of RG and B. Not a crappy 256 ... although differentiating is hard and may require a decent monitor that is calibrated well.

Puff

[bobby2]

PS - do I win an award for that much text in the little end of thread textbox posting?

[xortam]

We're starting to see some real-world feedback from AVS members with regards to Parhelia's use in HT. Check posts starting with Tom Strade's post today.

[Greebe]

Hey xortam... wanna get with the program and point him in our direction ... some of what he's having problems with has already been discussed here (ie how to set custom resolutions)

That would be better than effectively asking us to comment in every forum under the sun inorder to help those that need it or you allowing him to flounder.

[xortam]

I had just read their updates. I was going to point out the custom res solutions to Tom once I was sure I understood it. I think the answer is that you can hack the Registry for now and Matrox is working on a more elegant solution. I wanted to recheck the MURC posts before responding. Their other comments are interesting, i.e. Ashley's status, Gigacolor, iDCT, VIP 2.0, component output crippling, etc.

P.S. I guess I want to point them to this thread but I think they're trying to set up custom T&Rs that aren't available through PD-HF.

[xortam]

Andrew Chilvers of TheaterTek Inc. (a highly regarded DVD SW player among HT folks) just posted a review of HTPC DVD quality between the Parhelia and an ATI Radeon 8500. The Parhelia didn't fare very well.