mboen@nixpbe.UUCP (Martin Boening) (05/11/89)
Since some people asked for it, here it is. But caution: since I'm no
electrical engineer, the terms may be jumbled AND the descriptions
sometimes may be entirely off (which is then partly the fault of the ST
68000er magazine, as I quote from them).
---------------------Start-of-Text--------------------------------------------
This is a 'summary' of an article in the German 'ST68000er Magazin' which
described a possibility to extend the graphics ability of a normal ST to
up to 800 x 500 pixels for only 3 DM. Programs like Calamus, the newest
Version of 1st Word Plus as well as GFA Basic, Version 3.04 and many more
are supposed to support this extended graphics display (no warranty, of
course).
Here goes:
Basically the Atari ST Computers don't use the whole screen but only a
reduced portion of it. By making a little hardware extension the borders are
used as well thus giving a bigger picture and more pixels. The video RAM
is 67024 Bytes when using this extension.
The parts that are needed are a changeover switch, two diodes, two resistors,
a transistor and a capacity.
After the change you have 420 x 284 Pixels in low res mode (16 colors),
and in mid res mode (4 colors per pixel) 840 x 284 pixels. In monochrome
high res mode you get 840 x 568 pixels. Another advantage is that the
whole screen is used. There are some uses mentioned which I won't list here.
After some introductory remarks some theory to how the Atari video works:
(This is, of course, for 50Hz Mode, having been described in a German mag)
An ST video in 50Hz mode consists of 313 lines that are displayed 50 times
per sec. There is no interlaceing, the two halves are exactly on top of each other. The first 39
lines are always displayed in background color (0 in the palette). After
that follow 200 lines that begin with background color, contain (depending
on the mode) 320 to 640 pixels of graphics display and end with background
color on the right border. The following 45 lines are in background color
again. That way you get the green desktop ona white background. The last 29
of the 313 lines are blanked by the appropriate GLUE signal but aren't
visible on the screen.
Some remarks about hackers and the nifty demos they wrote follow. Then the
article goes on to say that an extended picture format has something to do
with the DE (Display Enable) signal of the GLUE chip. This signal is passed
on to the MMU, the shifter and the input for timer B of the MFP. The article
describes the way a screen is displayed in color mode:
The vertical sync signal is passed by the GLUE chip to the monitor AND to
the MMU. At the same time the start-address of the videoram containing the
first half picture (313 lines) reach the counter of the MMU (this address
is derived from addresses in FF8201 and FF8203. If at this point DE is high
the signal 'Display-Cycle-Clock' is generated which causes the transfer of
display information from the RAM to the shifter in 16-bit quantities via
the load port of the shifter. At the same time the counter chain in the MMU
is incremented. The current value is stored in the video-address-counter
at FF8205/07/09. If DE stays HIGH for 'too long' the MMU will address more
than 32 KB video ram and send it to the shifter as graphics data. If there
is to be a change in the color palette registers of the shifter, the signal
CMPCS (Color Map Chip Select), generated by the MMU, changes. Lastly the
Blank-Signal from the GLUE chip blanks the color signal from the
RGB-Resistance-D/A-network for the horizontal sync and for the 29 lines
vertical sync.
Software that generates extended pictures uses a trick to hold the DE signal
high a bit longer than necessary for displaying a line by switching the
display frequency in certain lines (interrupt-driven routine). This could
be done quicker by sending a different signal to the MMU instead of DE, a
signal that already has the appropriate pulse-pause-ratio (high-low-ratio).
ST6800er magazine now went and looked for a signal that might be abused.
The connection supporting MMU and shifter with the normal DE signal was
cut. The old DE inputs (Pin 52 of the MMU and pin 37 of the shifter) were
then connected to HIGH for a first test.
This seemed to work. The upper half of the screen contained randomly colored
pixels and had no border. The lower half contained strange 16-pixel-groups
that constantly changed in color. These were signals from the data bus due
to the fact that the videram is usually placed in the last 32K of RAM.
Therefore the video address in FF8201 and FF8203 was adjusted.
Another thing was that the desktop had disappeared. Since the bitplanes did
not fit anymore, only undefined pixels appeared.
There was also a problem with the old RAM-TOS GEM because the lines could
contain a maximum of 255 bytes video ram info and the number of bytes has
to be divisible by 4.
After a few experiments it was decided to use the composite sync signal
of an AND of HSync and VSync as a replacement for the DE signal.
This composite signal fulfills the requirements BUT it can only be used for
the old DE inputs of Shifter and MMU. If it is also hooked up to the input
of timer B on the MFP (Pin 20) the ST won't boot anymore. So for counting
lines the HBL interrupt is used.
Now for the description of how to modify your ST:
The guys from ST 68000er Magazine modified a 520 STM with a Version F
motherboard. (The printed circuit that supplies shifter and MMU with the
old DE signal should be the same for 520, 520+ and 260). For motherboard
revisions other than F the circuit should be checked with an oscilloscope
to see whether the signal is the same as on pin 37 of the shifter. If it
is, that circuit has to be cut. (There were pictures of the motherboard
but the usual vt100 has a rather low resolution so I didn't include them :-)
On the 1040 and the Mega ST (and when reading this summary) you have to
find the circuit that has to be cut yourself.
After cutting off the DE signal from MMU and shifter you can't just take
the composite sync signal because it is weakened too much by the monitor
to supply both shifter and MMU with a valid signal. A new composite sync is
needed. This is achieved by connecting two diodes, two resistors, an NPN
transistor and a capacity.
The circuit should be placed near the inputs of the HF modulator since all
signals needed (HSync, VSync and +5V) are available. After you're finished
you should insulate the whole setup against the cover.
The resistor for the emitter is connected to ground near the modulator
since the modulator itslf does not provide a ground on it's input pins.
For the whole setup the H- and VSYnc signals for the modulator should be
used even if there is no modulator since these signals have the correct
amplitude. There's some more about why the modulator input signals
should be used. For instance for the D revision of the 1040 ST the
capacity C114 should be removed since it dampens the signal.
There was a circuit diagram. The following is a sort of reproduction. (its
rather wide)
Schematic diagram for increased resolution graphics:
6 5 4 Pins at HF modulator (Old ST 520)
Switch
7 5 4 Pins at HF modulator (520 STM)
+5V HSync VSync
Overscan <----> normal o o o
/ | | |
/ |------o------------o | |
------- 100nF| | | --- ---
| | --- | --- ^ ^
------- --- | | | / \ / \ 2 * 1N4148
| | | | | 10K | | --- ---
| | | --- \ Ohm | | | |
| | | \ --- | |
| | | \| | | |
| | | |---------o------o-------o
| | | /|
| | | |/_ BC 555 or similar
| | | wire 3 |
| | |-----------------o
| | composite | Motherboard from above:
| | sync ---
| | | |240 Ohm
| | | | |----------------------|
| | --- | ------------ | ------------
| | | | | MFP | | L26 | | | |
| | --- | | 68901 | | | O | | | |
| | wire 2 (to DE entry) | | | | o cut | | MMU |
| |-------------------------------| ------------ \ / C16 | |
| | / | |
| wire 1(old DE signal) / / ------------
|-------------------------------------\ / /
\ | |
-------- \ o |
| GLUE | \ /
| | \ |
-------- --------O-/
solder here
Legend:
o connection points (usually displayed as 'bullets')
O soldering points on motherboard for the DE signal
(in the original these were marked with thin arrows
labelled 'soldering point x)
cut this is where the old DE is cut off from shifter and MMU
(in the original this was marked by an arrow labelled
'cut printed circuit here')
Caution: this diagram taken from a figure in the original article is
reproduced without warranty and of course way off the true scale.
The connections for wires 1 and 2 can be seen from the above diagram
(or can't they :-). one end of wire 3 is connected to the right pin of
the changeover switch on one side and to the connection of the emitter
resistor (240 Ohm) and the emitter of the transistor (BC 555 or BC 537)
on the other side. The changeover switch switches either the old Display-
Enable (for normal display mode) or the composite sync (for hyperscreen
mode) to the DE inputs of Shifter and MMU.
Now the bad news: For patching a TOS for hyperscreen a GFA Basic program
was written that patches an old RAM-TOS to a new beta TOS. This new TOS
supports both screen modes and even the switching between them during
operation. However, this program can only patch a disk version TOS. Also,
I didn't see the listing in the article.
For the ROM TOS a 'patch program' has to be put into the AUTO folder. This
program sets up everything for hyperscreen mode on startup. All it really
does according to the description is to 'notify' GEM of the bigger video
ram needed for hyperscreen mode. This program was supplied as a hex
dump only and I'm not going to hack that into any machine.
The next issue of ST 68000er computer is supposed to contain specifics as
to the screen operation of hyperscreen as well as a list of tested programs.
--------------------------End-of-Text----------------------------------------
This has been a lot of writing and undoubtedly I got many things wrong
(translating German electrical engineering terms into English ones when
you don't know very much about electrical engineering is very hard) But
anyhow this might give some people out there an idea of what the article
was all about that had the atari netters buzzing. I'm assuming that now
that this text floats around the net there'll be others who have read the
article and can add to or correct what I've written here. Maybe someone
even posts the program to adapt the ROM-TOS to the new screen-size on
startup.
Take this for what it's worth. The above doesn't reflect the opinions of
my employer, since it doesn't even reflect my own opinions. Anybody who
ruins his ST by trying this out is entirely on his own!
Martin
--
Email: in the USA -> ...!uunet!philabs!linus!nixbur!mboening.pad
outside USA -> {...!mcvax}!unido!nixpbe!mboening.pad