What makes the 2D quality of Matrox better then other ?

[Slashhead]

Rubank,

Thanks for the link. It looks very usefull to me. I think I can find the answers I'am looking for.

[Jorden]

Slash, one thing about the difference between the NV and Herculus: They might look the same, smell the same and might be the same, but did they (Herculus) explain the price difference then?

Trouwens, wil je nu echt via een Herculus 2D kijken?

Jord.

[Elie]

Slash,

That's Matrox's secret only! if they let it all out guess what'll happen?

So if you have a Matrox card, enjoy it's quality, if you don't have one go out and buy a G400 or G450 and enjoy, even the G200 has better 2D quality than Nvidia's set of cards.

Regards,
Elie

[Randy Simons]

They did some thourough tests at tecchannel.de..

However, the image quality of the G450 was not as good as it could be according to them. The Radeon scores much better at this point.

I've only a G200, so I can't tell the difference, but this looks to me as myth-shattering.

[Slashhead]

Jorden and Elie,

My interest is pure technical. It is not my intention to buy a Matrox card to have probably the best 2D available. I just want to know what causes the image quality differences between the different companies.

[rubank]

Yes Randy,

that is their conclusion. I read it a couple of months ago but I thought it best not to mention it in this forum, you see - I have no asbestos suit

rubank

[arbymo]

i don't believe it's just better components with lower tolerance, these will affect the charpness and lower the blurriness of the pic.
but guys just look at the colors, they r sooooo saturated and shiny and bright, it's something in the chipd design itself.
i have a G200 with 8 MB SD, and i just stare at the monitor when i used to play quake2. GREAT COLORS !!!
i saw quake2 on voodoo 2/3 and TNT2 ULTRA, a BIG difference.
there's a secret level in quake2 that's full of bright colors (red, green, purple) it just makes u love ur PC.

[Slashhead]

I have read the article and i doubt if their tests are reliable. IMHO measuring the output signal of the DAC has nothing to do with image quality. It's just finding something what you can measure, but it's very doubtfull that it has a direct relationship to the overall image quality. Perhaps you can intercept one particular signal, but what happens when there is a random output signal, which looks not very strange to me in operational environments.
Also the importance of the blockwaves makes no sense to me.

I am not a vision technician, so it is hard for me to say what IS important

The RAMDAC is a digital to analoque converter, so it has to be of a very highresolution to make nice waves without aliasing. If it is important to have sharp edges, why not keep the signal digital ??
They aren't explaining anything, just measure some "important" parameter (what about the resolution? ) and that's enough for them to draw conclusions.
But I believe what I see on the monitor and not wat my signal analyzer is showing. (as far as image quality is the isue)

[This message has been edited by Slashhead (edited 18 April 2001).]

[Slashhead]

Yes, i did some thinking myself and found something.

Asume you have a resolution of 1280x1024@85Hz. This means a total of pixel refreshes of 111M/sec for one of the RGB channels. (1280x1024x85)
Perhaps it is important that the DAC reaches the desired voltage in a small time table. For my example 8,9ns (1/111M)is the available time for that particular pixel.
What really important is (IMHO) how does the data from the first pixel effect the next one ? If the other pixel is infected by data from the former one, you may notice a certain aliasing effect, since that particular pixel is not showing the data it must show.
So, how is a DAC reacting on different voltage levels, how accurate is it ?
You have 2^32 different voltage levels, so it is easy that your signal can be infected by poor components or other external factors. What you may expect in a multiple color environment.
Is it necessary for the DAC to reach the zero voltage level ? I don't know.
I can't get a clear answer reading their article, also why are they measuring direct behind the DAC ?

[rylan]

Look, DAC output is almost _everything_
The company I work at designs very high resolutions cards for medical imaging/diagnosis. High quality dac + high quality output stage = high quality image.. period.

One of the ways we test our board quality is to measure the output from the board. We display several test patterns, and measure the rise/fall time of the signal (from black to full white), high/low level, and signal jitter. Fast rise/fall times translates to the pixels on the screen looking more defined between differant colors and brightness levels. Low jitter (very important for high res clarity) means that the pixels are stable. If you go to high res and get a magnifying glass and look at the pixels on a monitor, you can see them 'wiggle' a little.. that is jitter. Less jitter the better, obviously, as it increases image 'crispness'.

Also important is low crosstalk between the RBG color channels in the DAC.. you don't want some voltage on the reg channel feeding through to blue or green.

As your resolution increases, you need tighter and tighter jitter and rise/fall time specs, becuase the pixel clock rate skyrockets. We use 1GHz capable Dacs on our display boards, and usually run them around 560MHz for 5MPixel output (2560x2048).

Point is that Matrox uses a higher quality DAC for their boards.. it has lower jitter specs and less crosstalk, resulting in better color saturation and crispness of the image and higher resolutions.

And yes, unless you are going to use a digital LCD, you have to go to analog, since thats what CRTs are. With a full digital signal, you don't have the jitter problem found on CRTs due to the electron beam sweep. However, with digital signaling you don't have the bandwidth like you do with analog. The new generation of PanelLink chips are higher bandwidth so you can get past 220MHz pixelclock, which you'll need for 1600x1200 true color.

Anyway, feel free to ask anything else you like.. this is some of the stuff I do, so I can explain it more if ya want.

[Guru]

RAMDAC info at http://grafi.ii.pw.edu.pl/gbm/matrox/ramdac.html

[Slashhead]

Thanx Rylan.

I am looking for answers for a view weeks now, and I must say, at this board, it looks very promising. I've never been this far.
I will first read the whole link posted by Guru. If there are any questions left, or new ones born, I will post it.

[Slashhead]

My Questions:

1) How can you calculate the necessary RAMDAC speed for a certain resolution and refresh rate ?

2) How can you calculate the available bandwidth of your entire 2D circuit. I mean the DAC and the RF filter. Or is it published by the board manufacturer ?

3) How can you calculate the necessary bandwidth for a certain resolution and refresh rate ?

3) Why are they placing RC filters to prevent the DAC-output signal from fast rice/fall properties ? (I know it is for the CE mark, but it is ruin your signal output from the board. That's why I was surprised that they were measuring direct behind the DAC signal)

This makes absolutely no sense to me. You want sharp edges, putting lot's of energy in a very fast DAC to have sharp rice/fall behaviour. And then you connect it to a RC filter.

4) What kind of signal is used Hsync/Vsync to synchronise the DAC with the monitor. And how important is the accuracy of these signals in relation to image quality ?

5) Where do you expect RGB cross-talk at board level ?

6) What is the best way to keep the fast rice/fall from the DAC and also keep an eye on the CE.


Rubank,
I re-read the german article and it handles a lot of quality isues. My reaction was indeed not completely correct. But I still have the feeling that they were throwing with numbers, without showing how they were getting or calculating them. So i felt a bit left with empty hands. Also their measurments brings more questions to me. See Question 3 (above). Feel free to discuss this with me.

[This message has been edited by Slashhead (edited 18 April 2001).]

[rubank]

Slashhead,

my german skills might be inadequate, but I don´t read the article in the same way as you do, obviously.
The drawing you have pasted doesn´t show any filter they apply while taking measurments, as I understand it, it´s a sketch of the principal load and dampening effect of the cable between the card and the monitor, and/or the filter the manufacturer puts on the card. I haven´t been able to find any reference to their measuring directly after the DAC, as you have pointed out several times, I have taken for granted that they measure the actual output from the card. Are you absolutely sure you have got it right?? If so, I stand corrected but for now I´m in doubt (other times often in error but never in doubt ).

I also find that some of your Q:s are A:ed in the article.

Since all modern cards have the RAMDAC integrated into the graphics processor, one could argue that RGB crosstalk appears in the core. Many have said that an external RAMDAC is a better but more expensive solution, perhaps it is so.

rubank

[rubank]

Slashhead,

maybe you´re right.

But at least they are measuring something, whereas others just give their subjective view on the matter.

The traditional way of testing e.g. Hi-Fi stereo amps includes maeasuring the output signal through oscilloscope to determine the amount of distortion, noise, ringing, frequency ratio, eff. bandwidth and so on.
While these results don´t tell the whole story, they usually are a good indication of the quality of the sound reproduction.

I can´t see why evaluating video signals, in much the same manner, wouldn´t give a good indication of the resulting picture quality.

I don´t agree with you that they are not explaining anything. I think they explain in detail what they are measuring and how different values affect the visual output.
Why not keep the signal digital, you ask. Well, CRT monitors are analogue devices, and if you want to go all digital you must use an LCD (flatpanel) monitor with digital input and a video card with digital output. And even if you use a standard LCD monitor, which have a built in AD converter, most vidcards give pleasing results. The discussion is more or less a CRT matter only.

Maybe you´re wrong.

rubank