RE: Single ended termination?.

OK,
I'll try again with a message I sent on Monday but never showed up.
I've removed the photo of CAN reflections as outlined in Steve's paper. =20
I'll post it on a web page.
John

=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=3D=3D=3D
You know of course the reason you need at least one resistor on the bench i=
s because the drivers require that return path or they just won't work. =20

Now I've run all sorts of bench implementations with only one 120 ohm resis=
tor.  I've also used 300' of the non twisted 4 conductor solid tan jacketed=
 telephone zip wire for demos but there I did have two resistors.  So the e=
xperience side of me says it will probably work. =20

But think about this for a second.  The point of using a transmission line =
with 120 ohm impedance and a transmission line termination at each end is t=
o absorb reflections and create a bus with square wave signals that don't r=
ing and aren't upset up reflections. =20

The actual bit rate you use doesn't matter because the rise time of the sig=
nals is still the same regardless of 10kbps or 1Mbps.  The late Steve Corri=
gan, designer of the TI drivers, wrote an interesting paper on bus spacing =
along with stub length.  He showed that if the nodes are too close together=
 you can get a reflection that superimposes on other data bits.  I had trou=
ble believing that until I took a scope photo showing exactly that;  my nod=
es were too close together.

The other problem is your data is somewhat random. You can send a nice symm=
etrical 20kHz square wave through a bus driver and look at the signal to se=
e if an open end results in reflections on your bus.  But you aren't sendin=
g that sort of data.  It's all over the place with anything from symmetrica=
l to bit stuffing resulting in 5 dominant or 5 recessive broken up by the s=
tuff bit. =20

Again, if there is a reflection that changes the bit level enough to change=
 the data value you might not be able to reproduce it but oddly the error c=
ounters cycle above 0.  Sometimes once per month to bus recessive or even b=
us off.  But it's one of those problems that a 3 fingered salute on the key=
board fixes.  Or appears to fix.  Or it just goes away.

So the short answer is probably yes.  But a human life depended on the resu=
lt would you then follow the rules? =20

I'm rather anal that way so I follow the rules and then I don't have to thi=
nk about bus issues and focus on how I screwed up my software.



From: [email protected] [mailto:canlist-owner@vector-info=
rmatik.com] On Behalf Of John Dammeyer
Sent: February-10-21 11:18 PM
To: [email protected]
Subject: RE: [CANLIST] Single ended termination?.

Test email. =20
John


--
Subscribe, unscribe, and archives at www.canlist.org - report any problems to <[email protected]>