[comp.sys.atari.st] More detailed instructions for screen-size change

ivan@rzsin.sin.ch (Ivan D. Reid) (05/15/89)

	Several people have asked me to expand on the circuitry from "ST-
magazin" to allow an expanded ST screen.  The trick is this: a composite signal
is generated by ANDing together the HSynch and Vsynch signals and supplying
this signal to the MMU and Shifter chips instead of the normal DE (display
enable) signal from the GLUE chip.  Note that the original DE signal must still
be passed on to the Timer B interrupt (pin 20) of the MFP68901.  Additionally,
a switch is used to enable either the new DE signal or the original signal to
be used, allowing normal operation to be resumed when desired. 

	So, to describe the circuitry: the heart is an NPN transistor, type
BC555 or equivalent.  They give a BC237 as an equivalent; offhand I don't know
what 2N series would be equivalent -- perhaps someone with access to a D.A.T.A.
handbook can comment.  Next are two resistors: a 240 Ohm emitter resistor to
ground, and a 10K pullup on the base to +5 V. The collector is also connected
to +5, so it's essentially a common-emitter circuit; a 100 nF capacitor from
collector to ground takes care of the switching transients.  Finally two
diodes, 1N4148, are also connected to the base, with their cathodes going to
HSynch and Vsynch respectively.  Thus when either synch signal is low, the base
is pulled low via a diode and the transistor switches off giving a low signal
at the emitter; when both are high, the base is pulled high by the 10K resistor
and the transistor switches on, pulling the emitter high to ~4.4 V. 

	As for the actual connections, the original was built in a 520 STM
where the synch and +5 signals were conveniently located together at the
modulator.  Here +5, HSynch and VSynch are pins 7, 5, and 4 resp. For older
520s +5 is on pin 6 instead of 7.  My internals book gives H and V synch
signals at pins 37 and 38 of the GLUE chip, but there's a comment that the
signal should be taken from the modulator input connections even when  it is
not installed. There's also a comment that a Version D 1040ST sometimes has too
weak a synch signal to properly drive the transistor circuit, and I think it
says you have to buffer the signals through two 7404 inverter gates (but it
might be saying that these gates already exist -- I had to quit my German
classes just when we were learning "werden"). 

	That takes care of the inputs; the output from the transistor emitter
is taken to one side of a SPDT switch, while the original DE signal from the
GLUE chip (pin 39) goes to the other.  Now, the PC trace for DE is cut so that
the it still goes to the timer chip but not the MMU and Shifter chips; the
common pin of the switch is connected somewhere on the MMU/Shifter side of the
cut.  The three wires to the switch have to be long enough to allow the switch
to be mounted in some convenient location on the case of the ST -- personally
I'd prefer something a little more complex using DC switching so that the
signals weren't taken over such a long route (nearly 12 inches judging from the
photographs).  The original was also constructed in a "fliegenden Aufbau", i.e.
suspended in mid-air by the component leads, and wrapped in insulating tape.  I
would suggest a bit of perf-board or something so that a neater and more stable
construction can be made. 

	There's also a comment that you might have to remove C114 in a version
D 1040ST, I think to get a better risetime ("Flankensteilheit"); this capacitor
is at the upper left of the shifter shield (? "Shifter-Abschirmgehauese"). 

	As for the software, I'm still baulking at typing in 2900 bytes of
hex-dump, plus I'm not sure of the legality of posting it to the net. However,
the introduction states that when a substitute DE signal has been arranged
"then compute-time intensive synchronisation tricks are unneccesary and a
modified TOS or BIGSCREEN (ST-Magazin 11/88 edition) runs with the new
Hyper-Screen resolution under GEM".  Keith at UH Atari Computer Enthusiasts
(uace0@uhnix2.uh.edu) informs me that he has already modified the BIGSCREEN
code to suit, and is waiting on this detailed circuit description.  I trust
that he will make the code available once it has been tested.  By the way, it
is supposed to work in all resolutions: 420x284x16 colours in Lo-rez, 840x284x4
colours in Med-rez and 840x568 in monochrome Hi-rez.  I hope this has answered
all the questions... 

	ivan

Disclaimer: all of the above may be wrong.  We Aussies never have much reason
to learn German!