Xref: utzoo comp.sys.ibm.pc:15877 comp.arch:4969 comp.graphics:2547
Path: utzoo!attcan!uunet!husc6!mailrus!ames!amdahl!dlb!ardent!sleat
From: sleat@ardent.UUCP (Michael Sleator)
Newsgroups: comp.sys.ibm.pc,comp.arch,comp.graphics
Subject: Re: How did this program burn out two monitors?
Summary: It's a common problem.
Message-ID: <366@ardent.UUCP>
Date: 24 May 88 21:19:39 GMT
References: <10244@agate.BERKELEY.EDU> <17460@glacier.STANFORD.EDU>
Reply-To: sleat@ardent.UUCP (Michael Sleator)
Organization: Ardent Computer
Lines: 39


(Apologies if this is excessive detail.)

Typically, the horizontal scan circuit in a video monitor uses a "flyback"
technique in which the horizontal output transistor is turned ON to cause
the beam to sweep from left to right, and turned OFF to allow the magnetic
fields in the yoke and flyback transformer to collapse, quickly returning
the beam to the left edge.  During this time a pulse of, say, 400 to 600
volts appears on the collector of the output transistor.  If, in the middle
of this pulse, you attempt to turn the transistor back on, there will not
be sufficient base drive to saturate the transistor.  The result is extremely
high power dissipation which turns the transistor junction into a puddle of
silicon in a matter of microseconds.  A well designed monitor will not allow
the sync input to change the phase of the horizontal oscillator so abruptly
that this condition can occur.

The moral is that you don't have to set the horizontal frequency to zero
to cause a problem with some monitors.  Simply glitching the horizontal
sync in just the right manner is enough.  In fact, I'm a little skeptical
that turning off horizontal sync could do it.  Most monitors have a horizontal
oscillator that simply free runs in the absence of sync, and even those that
don't seldom have direct coupling to the horizontal output transistor, so
it shouldn't be possible to just turn the transistor on and leave it sitting
there in that state.  After all, if your product (the monitor) self destructs
if it's powered up without drive signals, it's very difficult to even build
it and get it out the door.

Of course, setting that register to 0 in the crt controller chip may in
fact produce a maximum frequency signal, rather than DC, which might wreak
havoc by the mechanism described above.

I vaguely recall rumors of a "letter bomb" program that blew up those silly
Commodore machines with the toy keyboards, way back when.  It zapped the
monitor in precisely this fashion.


				Michael Sleator
				...!uunet!ardent!sleat
				...!hplabs!ardent!sleat