Re: emc success

--- In CAD_CAM_EDM_DRO@egroups.com, Jon Elson <jmelson@a...> wrote:

>
>
> Wally K wrote:
>
> > Hi Jon, i have a sugestion for cleaning up the signals did you try
> > using a schmidt trigger to buffer the parallel port. Also you can
> > try putting the schmidt trigger at the paralell port as a

transmitter

>
> Not so easy, as there is no +5 V power available at the parallel
> port connector.

Well i guess you will have to run a 5 volt wire down there. If the
signals are as bad as you say you have no choice.

>
> > and have another one located at the input to your circuit.
>
> Well, I used a Schmitt trigger circuit on my hand-wired prototype,

but

> didn't have the right chip (need an SO-14) to fit the production

board -

> my goof! I used a 74HC04, which is obviously not as good. I will
> have to get some 74HC14's soon.
>
> > Make sure
> > the schmidt triggers have bypass caps at least .1uf and also 1uf

to 5

> > uf tantalum caps if located a long ways from power supply.
>
> I refuse to use Tantalum capacitors, due to the fact they burn up
> in a rather nasty way. Multilayer ceramic caps are fine for the
> same purpose, and even cheaper

In 25 years of electronic work i have never seen a tantalum burn up.
They are made for this kind of stuff. high volumetric efficency and
good service life.

>
> > If you
> > are running long lead lenghts use ribbon cable with every other

line

> > a ground.
>
> Well, ribbon cables may not be a good idea for this. Just using

ribbon

> cable on an IDC DB-25 plug won't work. The cable has to be made
> with hand wiring to put the noise sensitive (strobe) signals on

their

> own twisted pair. This would be a pain, except that IEEE-1284
> cables are already made to this spec! So, I'm using a 1284-

compliant

> cable. The real problem is that the drivers used are VERY weak, and
> so there are all sorts of glitches, wiggles and slow rise-times at

the

> end of the strobe/ pulses. These can be deep enough to cause
> additional pulses to be detected, which causes the address counters
> inside the chips to bump twice, and get out of sync with the

computer.

>
> > only connect 1 of these ground wires to your circuit the
> > others should only be connected on one end.
>
> Well, that sounds right, if only one conductor is carring the

signal.

> Unfortunately, every signal wire has a fast-transition edge signal
> on it at some time, and therefore needs a close, full-continuity
> return path. So, you really need a separate ground on every fast
> signal. In this case, I think they get away with a twisted pair

ground

> for all strobe signals.

If i remember right the 66Mhz IDE drives achieve there speed by using
the ground method i suggested.

>
> Jon