Hi Ake,

Identify the current path of the noise through your board and either bloc=
k it=20
or make it as easy for it as possible to pass through your board. It only=
=20
becomes a problem if the current is converted to a voltage as seen by you=
r=20
circuits (through an impedance of some sort, perhaps a bad ground plane).

Remember that the noise (or transient or surge) wants to go back to it's =
source=20
and it probably does so through earth. Either in your powersupply (direct=
ly or=20
capacitively coupled) or capacitively coupled directly from your board to=
 an=20
earthed groundplane.

One effective way to keep the noise currents away from your board is to p=
ut it=20
in a metallic enclosure and have shielded cables where the shield is dire=
ctly=20
connected to the enclosure through a very good connection such as an EMC =
cable=20
gland. Connect the shield at both ends. Block the power supply with induc=
tors=20
and decouple with capacitors to the enclosure (not the ground plane) in o=
rder=20
to make the impedance for the noise much lower through the enclosure than=
=20
through your board. You could also decouple the sensor signal to the encl=
osure.

If you don't have a metallic enclosure, mounting the board on a chassis p=
late=20
where the shield is connected to the plate and the powersupply decoupled =
to it,=20
could have the same effect.

If you want to filter the signal directly you have to do this through a g=
ood=20
groundplane. Knowing what frequency your board is most sensitive to the n=
oise=20
will help you pick the right filter components. An EMC test will tell you=
 that.=20
Either way, a good very ground plane is always helpful.

Good luck / Ruben

> Jose,
>=20
> my main problem appeared to be a ground problem just as Jim pointed out=
.=20
> I still need to follow you  advices though for he final card.
>=20
> I have no decoupling on the sensor lines today. Only on the power to th=
e=20
> sensors. The sensor in this project gives out a square wave of 1 kHz - =
4=20
> kHz and having a cap. on it will ruin the signal.
>=20
> Would something like Murata BLM21AG102SN1=20
> (http://www.murata.com/catalog/c31/es0100.pdf) do this job? I know I ca=
n=20
> test this but just now I feel I need to hold the hand of someone who ha=
s=20
> done this more than once....
>=20
> Another path in on the board is the CAN bus which has a very long cable=
.=20
> Should I do the same thing there?
>=20
> Regards
> /Ake
>=20
> Jose Da Silva wrote:
>=20
> >On Thursday 03 February 2005 09:35 am, Ake Hedman wrote:
> > =20
> >
> >>I have two identical boards on my desk  and have some noise problems
> >>with one of them. Both have a sensor input and a cable of about two
> >>meters attached to the input plus are connected to a CAN bus. The=20
> >>sensors are decouples on the board with a 0.1uF cap. Above my bench I
> >>have a fluorescent lamp and one of the boards get very upset (watchdo=
g
> >>reset) when this lamp is turned on/off.
> >>   =20
> >>
> >
> >If it is not a shielded cable, then you have a "nice" antenna to colle=
ct all=20
> >that RF noise. The 0.1uF caps by themselves are not enough. You need t=
o=20
> >think of adding either a resistance or an inductor on the sensor line =
before=20
> >the cap and, preferably you add them to both lines (assuming 1 pin for=
=20
> >gnd/vcc and 2nd pin for sensor input).
> >
> > =20
> >
> >>First it is strange that only one board does this but my question is =
how
> >>do one approach a problem like this?
> >>I have also tested a piezo gas lightning device close to the boards b=
ut
> >>it does not disturb the systems.
> >>   =20
> >>
> >
> >As you have found out, not all frequencies appear to affect it, or aff=
ect=20
> >every board the same way however, you are still dealing with an RF sou=
rce=20
> >and an antenna. If you look at some of the tests done to pass FCC or E=
U or=20
> >other electrical/electronics agencies, you will note that the tests ha=
ve=20
> >some very high voltages which will affect cables like yours of only 2 =
meters=20
> >length.
> >1st, you specify that the sensor should be on a shielded cable to make=
 your=20
> >circuit more immune from noise, 2nd you design your board so that peop=
le who=20
> >seem to know more than they think and still use plain wire won't affec=
t your=20
> >circuit.... so basically, you make your sensor input a low-pass filter=
=20
> >input. RC might be cheap, but LC may do nicer.
> > =20
> >
>=20
>=20
> --=20
>  ---
> Ake Hedman (YAP - Yet Another Programmer)
> eurosource, Brattbergav=E4gen 17, 820 50 LOS, Sweden
> Phone: (46) 657 413430 Cellular: (46) 73 84 84 102
> Company home: http://www.eurosource.se     =20
> Kryddor/Te/Kaffe: http://www.brattberg.com
> Personal homepage: http://www.eurosource.se/akhe
> Automated home: http://www.vscp.org
>=20
>=20
> --=20
> http://www.piclist.com PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>=20

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D
Ruben J=F6nsson
AB Liros Electronic
Box 9124, 200 39 Malm=F6, Sweden
TEL INT +46 40142078
FAX INT +46 40947388
ruben@pp.sbbs.se
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D

--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist