red-blue, 
red-green,
blue-yellow 

odd-lines: half res right image 
even-lines: half res left image

(works for HMD's, no stereo vision with shutterglasses)

(no stereo vision)

The refresh rate is doubled in the process.
(In practice some lines in the middle are lost because of the retrace time. There has to be a black border between the two images)

(best quality!)

Page-Flipping

There are 3 major ways to do page flipping. 

(1) Have 1 screen in video memory.  After every refresh, copy the other 
eye view onto the screen.  This is possible for small windows on the 
screen, but it is impractical for full-screen images. 

(2) Have data for both L/R screens in video memory.  Have software change 
the Start Address on the graphics card (the pointer to the first byte 
to display on the screen) between the beginning of the L or R buffer 
after each refresh.  This is software page flipping. 
The fact the software is doing the flipping means that it consumes 
CPU time. 

(3) Have data for both L/R screens in video memory.  Give the Start 
Addresses for both views to the graphics card, and tell it to automatically 
flip between these 2 Start Addresses on every refresh.  This is hardware 
page flipping, and it has no CPU overhead.  BUT it requires hardware 
functionality on the graphics card -- which is now popular due to the 
new VESA 3.0 spec.  It is also does not co-exist well with serial/parallel 
port glasses, because you still need software to signal the glasses. 

Interlace

Note that interlaced screen modes are actually a special case of 
hardware page flipping. 

...the graphics card sends a vertical sync pulse 
to the CRT once per refresh to signal the vertical retrace.  In interlaced 
modes, the width of the vertical sync pulse is different for the even and 
odd scan half-fields. 

Line-Doubled Page-Flipping

Technically, the term line-doubling means that the graphics card 
scans every row of data from video memory twice.  So, if you enable 
line doubling in 640x480, each of the first 240 lines are scanned twice 
to the CRT.  In effect, you get 640x240 but at the same CRT parameters 
(and hence same refresh rate, since you are still sending 480 scan 
lines to the monitor) as 640x480.  This mode only requires enough 
video memory as 640x240. 

Another way to achieve 640x240 is to create a new graphics mode. 
This is very complicated, since you need to recompute the CRTC parameters 
(H and V parameters).  You also need to reprogram the CRTC clock -- 
note that any given graphics chip (like the ATI Mach 32) can be 
on a board with any one of many different clock chips.  This mode 
would only scan out 240 lines, though, so you could use very high
refresh rates. 

The final way to achieve 640x240 is by using interlaced graphics 
modes... 

There are 2 disadvantages of line doubling.  First, your refresh rate 
is the same as your 640x480 mode.  Second, not all graphics cards can 
do line doubling in the VESA or accelerated graphics modes (like the 
ATI Mach32). 

There is one big advantage of line doubling -- the screen looks better 
on high-performance monitors.  When a monitor syncs to a graphics mode, 
it adjusts the spot size of its electron beam.  For low-res modes, it 
uses a bigger spot.  For hi-res and interlaced modes, it uses a smaller 
spot.  For very lo-res modes (?x200) on a large (15" and up) monitor, 
you will see thin black lines between each row of pixels, because the 
CRT can not make the spot any bigger.  By double scanning at low res, 
the CRT will remove these black lines. 

Note that all VGA resolutions below 350 rows use line doubling. 

Donald Sawdai 

LCD-BIOS page

There's a new technique not discussed there: 4) Line Blanking - synthetic interlace