Yes, I understand that this can be troublesome. But to know which keyboard
to use among a table full of them is troublesome to....

Anyway I have tested a simple mechanical switch for 8 computers
 controlling five NT servers and two linux machines ) and it worked quite
fine. The only problem we had was that we needed to switch to the server we
booted otherwise we lost the mouse on the NT machines. A "feature" that was
a little annoying after a power failure.....

Mark Willis suggests using 4051 doing the job and I think that sounds like a
good idea. The question is if it can handle the video signals or if one must
use some other means for that.

Most mouse devices today is compatible with a well known mouse from a
Redmond based company so a PIC that handles that protocol should solve the
"mouse problem".



/Ake

> -----Original Message-----
> From: pic microcontroller discussion list
> [mailto:PICLIST@MITVMA.MIT.EDU]On Behalf Of Peter L. Peres
> Sent: Monday, July 27, 1998 00:31
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [~OT] Switching for 1 monitor to many machines
>
>
> On Sun, 26 Jul 1998, Ake Hedman wrote:
>
> > I'm in real need for a monitor switch. One of those that are capable of
> > switching between several PC machines using one mouse, keyboard and VGA
> > monitor. The obvious solution would of course be to go out and
> buy one but I
> > kind of like to build things.....
> >
> > Have anyone one the list done a design of this kind or have any
> thoughts on
> > the subject? What special considerations should one have
> regarding switching
> > video streams?
>
> I have designed/built several prototypes. It looks easy but it is NOT
> easy. The monitor switch is relatively easy. The safest way is to switch
> all 15 poles, with the added problem of not wanting crosstalk induced by
> stray capacitance between R,G,B and anything else. Thus, get a large
> switch and use the (stacked) contacts such that between each R, G and B
> switchover stack segment, there is a GND-siwtching segment. The case
> grounds of all the units connected must be wired together (beware !).
>
> Some OSes/SVGA cards get confused when the feature bits are interrupted
> during operation. This may mean lock-ups and lost work or even crashes.
> The simple solution is, not to switch the feature bit signals through the
> changeover switch, and to provide a small DIP switch for each computer's
> SVGA feature bits. Set the switches to suit your setup BEFORE starting
> either computer. You will lose any VESA monitor control functionality like
> this. Also, if you happen to be using certain video gaming software aka
> mainstream OS (I will quote no names here), you will likely have to use
> 'safe mode' boot and set up the video/monitor mode after starting to use
> this setup (i.e. turn off the VESA SVGA monitor support). Also, spend some
> *money* on cables. I mean it, unless you like ghosts. For monitors with no
> internal 'smart' control, frequent switchover between sources that require
> multisync (H freq.) switchover spells early and spectacular destruction.
>
> Now, to more complicated things:
>
> Keyboard switchovers are straightforward, but one must make arrangements
> for both machines to be able to boot without a keyboard, and to make sure
> that the keyboard receives power when either computer is powered, and
> stays powered all the time (diodes on Vcc). The signal lines in all
> directions must have defined states while a switch changes sides (pull
> ups). Some software constantly updates LEDs and the like in the keyboard,
> which implies continuous transmission to the keyboard. Such software
> cannot be used with a simple (mechanical)  switch, it will lock up the
> keyboard at least, and may crash the machine(s) connected. Of course you
> don't want to switch while holding down any key. This is a PIC job really.
> Did I mention that the lines to/from each computer and the keyboard need
> to have defined states during switchover ?
>
> Mouse switchovers are straightforward for serial mice, if an only if, both
> machines use the mouse in the same mode. This is not very easy to do, as
> some 'extended' and 'improved' mice are used with proprietary drivers.
> Also, you cannot practically use a mechanical switch for this. If the
> mouse moves even one pixel when you switch, it being on the same table and
> attached by a wire, you are likely going to break a mouse protocol packet,
> and lock up or crash both machines immediately. Even Linux X servers die
> with this treatment. This is also a PIC job, a protocol buffering PIC job,
> to be more exact, so you have to know the protocol. Steer clear of
> multi-mode mice and 'extended functionality' for best results. Plain Mouse
> Systems protocol serial 3-button works ok, and covers UNIX [tm] (3
> buttons).
>
> So, now that I've said so much, I'd like to recommend the system switches
> from Rose, which I know to work well, and seem to be durable, or that you
> embark in a medum size PIC project, that will likely be more expensive in
> the end, than commercial solutions, due to the amount of mechanical manual
> work required to make such a box look good ;)
>
> My conclusion on the 'homebuilt' is, that it is not worth your while. A
> mechanical monitor switch is about the limit for easy homebuilt, and 2
> good ethernet cards plus a reasonable networking OS that lets you work
> remotely (not made in Redmond) are cheaper than the time to do all this in
> one-off mode (and cheaper than the commercial switches too).
>
> hope this helps,
>
>         Peter
>