Searching \ for 'Reflections in RS485 PIC Network?' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/microchip/ios.htm?key=rs485
Search entire site for: 'Reflections in RS485 PIC Network?'.

Truncated match.
PICList Thread
'Reflections in RS485 PIC Network?'
1998\09\18@151623 by N Steenkamp

flavicon
picon face
Hi,

I have designed a network of PIC driven units that will be installed in
an injection moulding factory.  The network is a muti drop RS485 with a
simple master-slave protocol.  The whole thing works beautifully, but I'm
a bit concerned about the effects of the network wiring when it is
installed in the factory.

I am planning on using an Ethernet-like structure where each unit taps
from a center connection.  My baud rate is 19200 and the taps can be
quite long 10's of meters.  What would be the effect of all these taps
(since each tap is a discontinuity in impedance)?  Would I have problems
with reflections?  Also, I cannot put a termination resistor at each
slave since the DC load would be too high.  Would AC termination be a
good idea to help minimize reflections from each slave back into the
network?

The alternative is to snake the central conductor to each node, thus
elliminating the long tap wires.  This, though, makes the wiring more
difficult and more expensive.

Lastly, I have 6V8 bidir-tranzorbs from each of the two differential
lines to ground in each slave and make use of a shielded twisted pair
cable.  Is this sufficient (in terms of interference and transient
protection) for such a network?

I would appreciate any suggestions!
Thanx
Niki

1998\09\18@165822 by Keith Doxey

flavicon
face
Hi Niki,

I havent actually tried this solution but I have considered it.....
Run your cables from a central point as you are proposing, but run CAT5
or similar and wire as follows....

Main Controller -----Blue pair in cable to 1st node then connect the Orange pair
to the blue at node 1

This sends the signal from Master to node 1 and ends up on the Orange pair
of Node 1 cable back at the central point.

Join the Orange pair of Node 1 to the Blue Pair of Node2
At node 2 join Blue and Orange

Join the Orange pair of Node 2 to the Blue Pair of Node3
At node 3 join Blue and Orange

etc...

At the final node terminate the Blue pair with a terminating resistor and also a
t
the Master.

The net result would be an effective continuous network without any taps TWICE t
he
length of your total cable runs but with only single cables to each location.
As mentioned earlier I HAVE NOT ACTUALLY TRIED THIS but I see no reason
why it shouldnt work.

Of course others may know better :-))

Best Regards

Keith

http://www.btinternet.com/~krazy.keith
Krazy Keith's World of DIY Home Automation

----------
From:   N Steenkamp[SMTP:spam_OUTSTEENKMPTakeThisOuTspamFIRGA.SUN.AC.ZA]
Sent:   18 September 1998 22:14
To:     .....PICLISTKILLspamspam@spam@MITVMA.MIT.EDU
Subject:        Reflections in RS485 PIC Network?

Hi,

I have designed a network of PIC driven units that will be installed in
an injection moulding factory.  The network is a muti drop RS485 with a
simple master-slave protocol.  The whole thing works beautifully, but I'm
a bit concerned about the effects of the network wiring when it is
installed in the factory.

I am planning on using an Ethernet-like structure where each unit taps
from a center connection.  My baud rate is 19200 and the taps can be
quite long 10's of meters.  What would be the effect of all these taps
(since each tap is a discontinuity in impedance)?  Would I have problems
with reflections?  Also, I cannot put a termination resistor at each
slave since the DC load would be too high.  Would AC termination be a
good idea to help minimize reflections from each slave back into the
network?

The alternative is to snake the central conductor to each node, thus
elliminating the long tap wires.  This, though, makes the wiring more
difficult and more expensive.

Lastly, I have 6V8 bidir-tranzorbs from each of the two differential
lines to ground in each slave and make use of a shielded twisted pair
cable.  Is this sufficient (in terms of interference and transient
protection) for such a network?

I would appreciate any suggestions!
Thanx
Niki

1998\09\18@182823 by Ryan Bloom

flavicon
face
I just finished implementing a 32 device RS-485 Network.   They are all in
close proximity, not through out a building but I found one problem which
affects the solution below if you use RJ45 connections.  If you go the 'in
series' approach with RJ45's beware of the voltage drop through each
connector (it's greater than you might think - I know first hand).

I had a lot of communication problems because of this, finally resolving it
by creating 'ring' type networks where either side of the net is connected
to the main bus.  Each ring supports 6 devices.  My RJ45 problem was
unusually bad because each one of my boards had an RS485-In and a RS485-Out
RJ45 jack, doubling my loss per board. The series approach is also kind of
scary because if one of your boards gets unplugged you lose all devices
after that point on the net.  The ring approach works very well and if a
device gets disconnected the other half of the ring is connecting the
devices.

Amazingly enough I haven't had any problems with reflections in this type of
network.  I have put two termination resistors at two ends of the net which
are not on the rings, balancing the net.

FYI - I am using CAT 5 cable unshielded and for RS485 com I am using the
National DS96176 transceiver.

Hope my $0.02 helps.

-Ryan

-----Original Message-----
From: Keith Doxey [keith.doxeyspamKILLspamBTINTERNET.COM]
Sent: Friday, September 18, 1998 1:51 PM
To: .....PICLISTKILLspamspam.....MITVMA.MIT.EDU
Subject: Re: Reflections in RS485 PIC Network?


Hi Niki,

I havent actually tried this solution but I have considered it.....
Run your cables from a central point as you are proposing, but run CAT5
or similar and wire as follows....

<snip>

1998\09\18@184136 by David Blain

flavicon
face
Hello,

I have never tried this either, but what about creating a hub/repeater.  By
using 2 devices like the MAX-487 back to back you would in-effect create a
repeater.  Put a pair per port on a PCB and each port would service 1 or
more devices.  The main bus would run on the PCB connecting all ports
together.  You could also have the master (PIC) on the same PCB.

This would allow you to customize the wiring/termination on each port to
meet the needs of the load.

Just a thought.

David.

>{Original Message removed}

1998\09\18@185207 by David VanHorn

flavicon
face
Running such low speeds, you aren't likely to see anything but a tiny bit of
jitter on the edges of the bits.
For the reflections to matter (especially short stubs) they have to be
electrically long relative to the wavelength of the bits, in the wire.  If
you could get reflections out to say 10% of a bit time, then I'd be
worrying, but this would require REALLY long lines, or much faster
bit-rates.

1998\09\19@012858 by Chris Eddy

flavicon
face
N Steenkamp wrote:

> Hi,
>
> I am planning on using an Ethernet-like structure where each unit taps
> from a center connection.  My baud rate is 19200 and the taps can be
> quite long 10's of meters.  What would be the effect of all these taps
> (since each tap is a discontinuity in impedance)?  Would I have problems

Case study one:
A past job involved an RS485 like bus (refer to national DS8831,
discontinued) for use in coal mines.  I have done field service on nets that
run split at the mining office, run 500 yards in one direction, also split
and run 1200 feet down a shaft, then split at the bottom of the shaft and run
5 miles in both directions, and go through a miriad of splits.  Trust me,
Cletus in west virginial has no respect for the bus versus star topology.  It
worked at 300 baud!  I would show up to troubleshoot bad comms, and when you
put the scope on the line, use differential mode, you could barely discern
square waves in the ocean of junk, yet enough messages get through in tact to
make most nodes remain in comms on the screens.  After tracking down problems
such as partly damaged but working comms cards, bad pieces of wire (?!) and
runs of wire that are wire tied along a conduit carying a few tens of KVA, I
would have it cleaned up such that you got a very respectable signal.  The
keys to operation were these:

We didn't bother with termination resistors (a hush falls over the crowd..).
They would have never figured it out.  I would find them tacked on the line
by a former troubleshooter, and they were just masking the real problems.

Each node had an isolated DC/DC and three optocouplers (RX, TX, HS) for the
total isolation of the RS485 driver and reciever.  I saw situations where if
you connected local ground to the isolated driver ground, it just stopped
working.  They LIKED having that isolation, and like a kid, it got mad when
you bridged it.

Each node had a weak (IE100K) pull up on one line and pull down on the
other.  This may have acted as a pseudo termination, but clearly not to the
RS485 specifications.  RS485 (ref National databook, with a fabulous app
note) should have these pull ups/downs.  When a driver releases the line, it
keeps ambient noise from triggering the recievers which when the two wires
are close in potential cause receivers to spout reams of garbage pulses.  The
bias resistors pull the line to that default space state.  They are more
important than one would think.

Why not build a network with spools of wire, and put a signal generator on
the line to check it out.  Remember that in the office, you have built a
textbook perfect example.  But you can check those nasty reflections.  You
might see some useful discussion on the Echelon LON web site under free
topology.  They do have a study covering the effect of stub lengths on their
topology.  To be honest, I can't remember if they explain why, or just tell
you that the max stub length is __ and not to exceed.  Worth a look.

Best of luck.
Chris Eddy
Pioneer Microsystems, Inc.

1998\09\19@082234 by Russell McMahon

picon face
I have no practical experience of this, but ....
.
At 19200 bps a wavelength is C/19200/K
where C is 3 x 10^8metres/second and K is a velocity modifier.
For K = say 3,  a wavelength is about 5000 metres.
I'd say your  eg 30 meter stubs are going to be fairly harmless.
.
But then, as Carl Sagan was won't to say, I may be wrong.


       Russell.
{Original Message removed}

1998\09\19@091302 by David VanHorn

flavicon
face
>I have no practical experience of this, but ....
>.
>At 19200 bps a wavelength is C/19200/K
>where C is 3 x 10^8metres/second and K is a velocity modifier.
>For K = say 3,  a wavelength is about 5000 metres.
>I'd say your  eg 30 meter stubs are going to be fairly harmless.
>.
>But then, as Carl Sagan was won't to say, I may be wrong.



We installed a lot of 485 networks for banking equipment that ran 485 at
19200. (or slower)
Never saw a problem related to reflections, and actually, it appeared that
the termination resistors were hurting rather than helping. The reflections
were over with so fast, that the resistor's main effect was to cut the
signal level in half. I think they may help damp out noise pickup though.

If you don't have real uarts, then a weak pull to +5V on one line and ground
on the other will help cut down on the idle-state "chatter".

1998\09\20@085330 by N Steenkamp

flavicon
picon face
Hi,

> >I have no practical experience of this, but ....
> >.
> >At 19200 bps a wavelength is C/19200/K
> >where C is 3 x 10^8metres/second and K is a velocity modifier.
> >For K = say 3,  a wavelength is about 5000 metres.
> >I'd say your  eg 30 meter stubs are going to be fairly harmless.
> >.
> >But then, as Carl Sagan was won't to say, I may be wrong.

Doesn't the rise/fall times also play a role in the severity of
reflections?  I remember something about a rule of thumb saying if the
rise/fall time is shorter than twice the propagation time then
transmission line theory must be applied.
{Quote hidden}

Do you mean by real uart a uart that oversamples the input and decides on
some majority rule?  If all transmission line effects die out within the
first 10-20% of each bit, then one should be safe.  As long as the
tranzorbs can taken care of the over and undershoots.

Thanks,
Niki

More... (looser matching)
- Last day of these posts
- In 1998 , 1999 only
- Today
- New search...