What should I make of Yousuf's post, here?

Yousuf Khan wrote:
Yeah, DVI doesn't improve your picture one iota. About the only
advantage I've seen from it is that with an lcd monitor, it allows you
to scale the non-native resolutions a little better, closer to a CRT
monitor's scaling.

* MISS CHIEVOUS:

Don't know. It's up to you. He even didn't write what setup (gfx card,
monitor) he has and where he didn't notice any difference. My experience
tells me otherwise.

Benjamin

Output filters, sound right.

"Benjamin Gawert" <bgawert (AT) gmx (DOT) de> wrote

I realize this is showing up as a reply to the wrong posting, but I missed
the
following comments when they were first posted, and want to correct a
couple of misconceptions:

Actually, "DAC" (digital to analog converter) is correct; a "RAMDAC"
was simply a DAC chip which included the color look-up tables
("color map" memory, as RAM), before BOTH functions were
integrated into the graphics chips. There's no sense in keeping the
term "RAMDAC" around at all any more, since the two are completely
separate functions.


This is a common misconception, but it IS a misconception. The
main reason that any "digital" interface provides improved image
quality with LCDs or other fixed-format displays is that such interfaces
provide an explicit pixel clock, so that the data can always be properly
mapped to the physical pixels of the screen. Analog interfaces such
as the "VGA" connector do not provide such information, and instead
the sampling clock has to be derived from other timing information
(typically the horizontal sync signal) provided by the interface.
Creating a sampling clock in this manner, though, can lead to some
errors (and it's why many analog-input LCD monitors include controls
which permit the user to fine-tune the sampling clock frequency and
phase).

The notion that avoiding a digital-to-analog conversion is responsible
for whatever quality improvement occurs comes from the common
(but also mistaken) notion that LCDs are themselves somehow
"digital." Fundamentally, though, the LCD is an analog-drive
device, and a digital-to-analog conversion occurs within the LCD
panel, at the drivers. In fact, LCDs have been made which preserve
an analog video input all the way through to the pixel level - these were
sometimes used back when analog monitor interfaces were all there were.

Bob M.

* Bob Myers:

It makes sense because the CLUT table is still there. Most modern GPUs
work with a fixed color depth (usually 24bit, some also with 32bit). The
CLUT is still needed and used when modes with lower color depth (16bit,
12bit, 8bit) are used.

No, it isn't.

This is of course correct, but doesn't change a yota to the fact that
the main reason for the improved image quality of digital connections
via DVI simply is the absence of A/D- and D/A conversion, and the fact
while with analog video signals the image quality is directly
proportional to the signal degradation of the transmission line with
DVIs digital TMDS signalling the image quality remains constant until
degradation reaches a certain point.

LCDs are pixel-mapped devices with fixed resolution while CRTs can be
pixel-mapped, line-mapped (i.e. TVs) or vector-mapped (i.e. data
displays in most aircrafts). It simply doesn't matter that the LCD
pixels are driven analog. The fact that the digital signals provide an
image that is "pixel-matching" while the resulting signal after
conversion into an anlog signal doesn't is one of the main factors
invfluencing the image quality the user notices.

Benjamin

"Benjamin Gawert" <bgawert (AT) gmx (DOT) de> wrote

Yes, it is. How long shall we keep this up? :-)

Not at all. This is far from the MAIN reason for improved
image quality, as evidenced by the large number of current
analog-interfaced LCD monitors (and other fixed-format display
devices which provide analog inputs) in which the resulting image
quality is indistinguishable from the same situation but with a
"digital" input. The impact of instantaneous noise in the analog
channel on the quality of the resulting image is generally negligible,
unless it gets REALLY noisy - owing to the fact that the display and
the eye will average out noise-induced errors over multiple frames.
The classic problem with analog-connected LCDs, etc., has
always been instability in the video data with respect to the display
panel's physical pixels - or, in other words, incorrect and/or unstable
sampling. The digital interface in this case has the distinct advantage
of providing unambiguous pixel-level timing information, and so
avoids such problems. (For a fairer comparison, try disconnecting
the clock signal in a digital interface - and then try to regenerate
THAT clock from, say the horizontal sync signal, which is exactly
what analog inputs on LCDs are doing. Let me know how well
that works out for you...:-))

If by the above, you mean LCDs are fixed-format (i.e., possessing
a fixed physical array of pixels), that's precisely what I said above.
However, CRTs have NEVER been "pixel mapped," in that there
has never been a CRT-based display made (well, with the exception
of some extremely low-volume, niche-market designs) where the
"pixels" of the video data were in any way constrained to map to any
physical structures (the phosphor dot triads, say) of the screen.
Pixels as such simply don't exist for the CRT in the sense meant in
this discussion.


Bob M.