18 Dec
2005
18 Dec
'05
9:37 a.m.
On Sun, 18 Dec 2005, Bob Shannon wrote:
The 12555 generates its Z axis signal based
(indirectly) on the I/O backplane
timing.
In a faster machine (E and F series) a new set of X/Y values can be loaded
into
the registers before the Z axis signal is done. This will produce a glitch.
If your using this board in an E or F series processor, I'd recommend one of
two approaches to work around this issue.
1. use programmed I/O rather than DCPC (DMA) transfers.
2. Double up each point sent, sending the same X/Y values at least twice.
I'm not using DMA, nor do I use the Z axis signal for blanking. We've
scoped some signals on the card, and I really don't know what is going on
on this card. The Z signal appears just when the glitch starts and stays
active just before the new value arrives. So either you have a long Z
signal and always the glitches at the beginning, or you only have a very
short pulse which is not long enough to have something useful on the
screen. (The 12555 is used on the HP 2100S)
The glitch I am talking about depends from the distance the point has to
go, and is opposite of the direction! That means e.g if the dot is
at the center position of the scope and I let it move to the bottom center
position, I get a very strongly visible glitch to the top, the longer the
distance the longer the glitch (verified by Y-T scoping the signals using
the Z signal as trigger). This must come somewhere from the digital part
of the D/A card, and that's why I need the schematics to understand how it
works.
I hope the following will help a bit
(O=current position, | and - transition, X=final position, .. glitch)
.
.
O
| ..O----X X----O..
|
|
X
Christian