In comp.arch.embedded, Paul Keinanen <keinanen@sci.fi> wrote:> On Wed, 07 Oct 2009 11:04:01 +0200, Stef ><stef33d@yahooI-N-V-A-L-I-D.com.invalid> wrote: > >>In comp.arch.embedded, >>Paul Keinanen <keinanen@sci.fi> wrote: > >>> My understanding is that a 2x pair has a better balance than between >>> two randomly selected wires in a 4x cable, thus having better >>> differential mode noise rejection. >> >>When using a star quad, it is not advisable to randomly select wires. >>You still need to pick 2 oposite wires for the "pairs". At least that's >>what I found in my web searching last night. ;-) > > My practical experience with 4-wire RS-422 (2x2), the installers seems > to be quite capable of making split pair errors :-(. > > I am not so sure, if they could pick the correct wires from a 4x > cable.But that also goes for dual twisted pair cables, easy to pick the wrong wires and it really happens. Fortunately they are color coded, so just specify which color to use for what. I doubt an installer would even spot the difference between the 1x4 and 2x2 cable, both have 4 wires, they can have the same color coding and the wires are a little curly in both. -- Stef (remove caps, dashes and .invalid from e-mail address to reply by mail) You have a massage (from the Swedish prime minister).
Experience with Star Quad cable for CAN bus
Started by ●October 6, 2009
Reply by ●October 7, 20092009-10-07
Reply by ●October 7, 20092009-10-07
On Wed, 07 Oct 2009 14:20:05 +0200, Oliver Betz <obetz@despammed.com> wrote:>Paul Keinanen wrote: > >>>But IMO the star quad should give higher EMC immunity because there is >>>less common mode noise between data and power (GND) wires. With 2x2, >>>GND and data has a larger loop area and therefore is more susceptible. >> >>If you use the ground wire at all. A correctly terminated line with >>floating receivers and transmitters should work well even without the >>Gnd connection. > >Do you have any real world example of a "floating receiver" designed >to operate without GND connection?The CANbus transceiver is quite similar to RS-422/485/Profibus-DP transceivers and I have never have had any problems using these with properly terminated lines without ground connection. With the termination resistor present, this is a nice bipolar current loop circuit and the floating receiver input transistors are biased to a reasonable potential with the internal pull-up resistors (in the order of tens of kohms).>Unless your isolation barrier between the transceiver and the "rest of >the world" has extremly low capacitance, common mode noise will also >appear at the transceiver terminals and cause problems.The stray capacitance reactance and the receiver bias system resistance is an issue. One alternative would be to use a single shielded 2x twisted pair with a proper termination for data transfer. The twisted pair will handle the data communication without problems. The cable shield could be connected at both ends with a 100-200 ohm 1-5 W resistor to local signal ground (possibly bypassed with small capacitors). Those resistors will reduce the "ground loop" 50/150 Hz current to a reasonable level and hence reduce the magnetically induced noise voltage to a reasonable level. Signal cable shields are really nasty in big industrial plants, since quite a lot of the mains current neutral current will flow though these cables, unless special precautions are used. Paul
Reply by ●October 7, 20092009-10-07
Stef wrote:>>>> Our bus will be aprox. 40m, 500kbit/s and the environment is noisy.[...]> That is something to consider as maximum bitrate in CAN depends upon > propagation time. But as we are not using maximum cable length and not > maximum speed, I suspect it's not that critical here.You're not hitting either of the absolute limits, true. But the actual, relevant limit is ~50 Mbit*m/s for the product of bus length and baud rate, and you're within roughly a factor of two from the ceiling on that: at 500 kbit/s the theoretical maximum CAN bus length is ~100 meters. So you're not exactly in the critical regime yet, but neither are you comfortably far away from it. Any single design choice would have to be pretty bad at this stage to break the design all on its own --- but several mediocre choices could break it, too. So you'll have to take care.
Reply by ●October 7, 20092009-10-07
Paul Keinanen wrote:> > The CANbus transceiver is quite similar to RS-422/485/Profibus-DP > transceivers and I have never have had any problems using these with > properly terminated lines without ground connection. With the > termination resistor present, this is a nice bipolar current loop > circuit and the floating receiver input transistors are biased to a > reasonable potential with the internal pull-up resistors (in the order > of tens of kohms).That's fine so long as the common mode voltage between the ends is within the range of the receiver. Usually, each end is grounded, so you effectively get the third gnd wire anyway, but if the ends are separated by a large distance, or on different line phases with gnd significantly above zero, you can worst case burn out the line receiver, or at best get lost data. That's why most comms equipment has both a frame gnd, going to power earth, and logic gnd, with the two separated. In the classic case of 10BT ethernet, you always have a transformer to provide the gnd isolation... Regards, Chris
Reply by ●October 7, 20092009-10-07
In comp.arch.embedded, Hans-Bernhard Br�ker <HBBroeker@t-online.de> wrote:> Stef wrote: > >>>>> Our bus will be aprox. 40m, 500kbit/s and the environment is noisy. > [...] > >> That is something to consider as maximum bitrate in CAN depends upon >> propagation time. But as we are not using maximum cable length and not >> maximum speed, I suspect it's not that critical here. > > You're not hitting either of the absolute limits, true. > > But the actual, relevant limit is ~50 Mbit*m/s for the product of bus > length and baud rate, and you're within roughly a factor of two from the > ceiling on that: at 500 kbit/s the theoretical maximum CAN bus length is > ~100 meters. > > So you're not exactly in the critical regime yet, but neither are you > comfortably far away from it. Any single design choice would have to be > pretty bad at this stage to break the design all on its own --- but > several mediocre choices could break it, too. So you'll have to take care.Thanks for that warning. This actually confirms my "need" for a good cable. I know there are at least a few mediocre (high speed signal wise) connectors in the bus that can not be easily avoided. Using a good cable at least reduces the number of probable problem spots. The idea for now is to do some measurements on the completed system to check how the CAN signals behave. We'll probably use a CAN monitor, but that will only give us a hit/miss condition for the bits, not how close or far they are from the limit. So in addition we want to see the signals on a scope to get an indication of the signal integrity. Any ideas on suitable methods? -- Stef (remove caps, dashes and .invalid from e-mail address to reply by mail) Zero Mostel: That's it baby! When you got it, flaunt it! Flaunt it! -- Mel Brooks, "The Producers"
Reply by ●October 7, 20092009-10-07
On Wed, 07 Oct 2009 14:20:05 +0200, Oliver Betz wrote:> Paul Keinanen wrote: > >>>But IMO the star quad should give higher EMC immunity because there is >>>less common mode noise between data and power (GND) wires. With 2x2, >>>GND and data has a larger loop area and therefore is more susceptible. >> >>If you use the ground wire at all. A correctly terminated line with >>floating receivers and transmitters should work well even without the >>Gnd connection. > > Do you have any real world example of a "floating receiver" designed to > operate without GND connection? > > Although I would be able to design such a circuit, I would not consider > it if I had a chance to get a GND wire.It's RS485, not CAN bus, but here is one example of a transceiver designed to work without a ground connection that allows kV of common mode signal: http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1007,C1017,P1746 I'm glad that parts like this exist now; I had to roll my own back in the day. Regards, Allan
Reply by ●October 8, 20092009-10-08
On Wed, 07 Oct 2009 22:08:23 +0200, Stef <stef33d@yahooI-N-V-A-L-I-D.com.invalid> wrote:>In comp.arch.embedded, >Paul Keinanen <keinanen@sci.fi> wrote: >> On Wed, 07 Oct 2009 11:04:01 +0200, Stef >><stef33d@yahooI-N-V-A-L-I-D.com.invalid> wrote: >> >>>In comp.arch.embedded, >>>Paul Keinanen <keinanen@sci.fi> wrote: >> >>>> My understanding is that a 2x pair has a better balance than between >>>> two randomly selected wires in a 4x cable, thus having better >>>> differential mode noise rejection. >>> >>>When using a star quad, it is not advisable to randomly select wires. >>>You still need to pick 2 oposite wires for the "pairs". At least that's >>>what I found in my web searching last night. ;-) >> >> My practical experience with 4-wire RS-422 (2x2), the installers seems >> to be quite capable of making split pair errors :-(. >> >> I am not so sure, if they could pick the correct wires from a 4x >> cable. > >But that also goes for dual twisted pair cables, easy to pick the wrong >wires and it really happens. Fortunately they are color coded, so just >specify which color to use for what. I doubt an installer would even spot >the difference between the 1x4 and 2x2 cable, both have 4 wires, they >can have the same color coding and the wires are a little curly in both.Having only 1x2 plus an optional shield will reduce the wiring errors considerably. Paul
Reply by ●October 8, 20092009-10-08
On Wed, 07 Oct 2009 22:00:52 +0100, ChrisQ <meru@devnull.com> wrote:>>Paul Keinanen wrote:>>>>If you use the ground wire at all. A correctly terminated line with >>>>floating receivers and transmitters should work well even without the >>>>Gnd connection.By "floating" I refer to galvanic isolation.>> >> The CANbus transceiver is quite similar to RS-422/485/Profibus-DP >> transceivers and I have never have had any problems using these with >> properly terminated lines without ground connection. With the >> termination resistor present, this is a nice bipolar current loop >> circuit and the floating receiver input transistors are biased to a >> reasonable potential with the internal pull-up resistors (in the order >> of tens of kohms). > >That's fine so long as the common mode voltage between the ends is >within the range of the receiver.Typically 500-2500 V in case of galvanic isolation.>Usually, each end is grounded, so you >effectively get the third gnd wire anyway, but if the ends are separated >by a large distance, or on different line phases with gnd significantly >above zero, you can worst case burn out the line receiver, or at best >get lost data.I use the following rule of thumb when specifying communication systems for large industrial plants: * RS-232 is OK within the same equipment rack * non-isolated RS-422/485 is OK within the same equipment room * isolated RS-422/485 is OK within the same building * fiber optic connections are required between buildings with separate ground electrodes due to the ground bounce created by a lightning stroke into the lightning rod of one building.>That's why most comms equipment has both a frame gnd, going to power >earth, and logic gnd, with the two separated. > >In the classic case of 10BT ethernet, you always have a transformer to >provide the gnd isolation...The only thing I like about ethernet to serial RS-232/422/485 converters is that you get the galvanic isolation for "free". Paul
Reply by ●October 8, 20092009-10-08
Paul Keinanen wrote: (floating receivers and transmitters without GND connection)>>Do you have any real world example of a "floating receiver" designed >>to operate without GND connection? > >The CANbus transceiver is quite similar to RS-422/485/Profibus-DP >transceivers and I have never have had any problems using these with >properly terminated lines without ground connection. With theThe product(s) passed all the EMC immunity tests, especially ESD, EFT and BCI?>termination resistor present, this is a nice bipolar current loopA (multidrop) current loop is something completely different.>circuit and the floating receiver input transistors are biased to a >reasonable potential with the internal pull-up resistors (in the order >of tens of kohms).Do the math: If the internal resistors - both in parallel - are 40kOhms and the capacitance of the isolation barrier is 50pF, 1kV/us will result in 20V common mode voltage on your receiver - likely too much. And you will get much more than 1kV/us in ESD and EFT tests and in real world operation - guess why good couplers specify a common-mode transient immunity of >10 kV/us. Or consider the BCI test (EN61000-4-6). At 1MHz, 50pF is i*3kOhms, so you will see the full common mode voltage even at this low frequency.>One alternative would be to use a single shielded 2x twisted pair with >a proper termination for data transfer. The twisted pair will handle >the data communication without problems. > >The cable shield could be connected at both ends with a 100-200 ohm >1-5 W resistor to local signal ground (possibly bypassed with small >capacitors). Those resistors will reduce the "ground loop" 50/150 Hz >current to a reasonable level and hence reduce the magnetically >induced noise voltage to a reasonable level.we are talking about _isolated_ transceivers. 50Hz and 50pF give tens of Megahoms impedance.>Signal cable shields are really nasty in big industrial plants, sinceTherefore you use isolated transceivers. But with most isolated transceivers, you need a GND wire. As I wrote - it is _possible_ to make a true two wire isolated transceiver, but I'm not aware of real world products. Oliver -- Oliver Betz, Munich despammed.com might be broken, use Reply-To:
Reply by ●October 8, 20092009-10-08
Allan Herriman wrote: (floating receivers and transmitters without GND connection)>> Do you have any real world example of a "floating receiver" designed to >> operate without GND connection? >> >> Although I would be able to design such a circuit, I would not consider >> it if I had a chance to get a GND wire. > >It's RS485, not CAN bus, but here is one example of a transceiver >designed to work without a ground connection that allows kV of common >mode signal: > >http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1007,C1017,P1746I can't read anywhere in the data sheet that it is designed to work without GND reference. Where is this stated? Anyhow, the data sheet is rather poor. They neither state a coupling capacitance nor a maximum spec for common mode dV/dT. The receiver common mode range (referred to GND2) is also missing. And see figure 12 - they _do_ connect GND2 to the cable shield. Well, they have a huge 10nF capacitor between the grounds in this example so there definitely need a GND connection there. I'm curious whether anybody successfully made a two wire (without GND) floating RS485 interface with this part and surviced EFT and ESD tests or even current injection. Oliver -- Oliver Betz, Munich despammed.com might be broken, use Reply-To:
