60hz interference with CANbus?

CAN is quite robust to the air coupled junk. Most likely you are experiencing the strong common mode interference along the CAN line. Make sure there is no bouncing between the ground potentials on the both ends of the bus, and there is no grounding currents via CAN.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

If it is common mode noise Paul could try some clamp-on ferrites on the lines. Either order from a place like Amidon or, with lots of luck, a Radio Shack here or there might have some. A couple years ago I lucked out when I had a similar situation at a client. Their local Radio Shack had four of them (down to zero when I left...).

For hardcore CM noise: Get some large #43 and some large #77 or J ferrites. 1" O.D. or more. Then pull the cable through as many times as possible without force or cable damage. If that doesn't change a thing then most likely it ain't CM.

Generators produce a lot more than just 60Hz. Besides lots of harmonics there may be commutation spikes etc.

The CM input impedance of CAN is at the order of kOhms. Besides, the proper termination of CAN on the transceiver should have a decent common mode choke. The external ferrite can't really help.

CAN is very reliable indeed. However there are two typical mistakes when it works with glitches:

1. No common ground connection between the ends of the bus at all.
2. The huge current along the bus ground.

I believe it can be either one of those two reasons. An optocoupled CAN can help both cases however it looks like an overkill for this application.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

top posting corrected

try 100nF to ground on the unterminated shield

martin

They can. I have used them successfully on other kinds of buses that have inherent CM rejection. Ethernet as well. I would definitely try.

The 2nd problem can be helped with additional CM chokes if it is noise at higher frequencies that rides on that ground. I agree on the opto-solution. It's often done in RS232 but that is almost like using a Smith&Wesson to kill a fly. I think in "Once upon a Time in the West" a guy did that...

CAN is *not* differential. It is pseudo differential. CAN requires that the both ends of the bus should have the common reference ground. CAN is three wires, not two. CAN-H, CAN-L and GND. That makes it very different from the TP Ethernet, for example. The most common reason for the problems with CAN is the bouncing between the reference grounds at the ends.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

Supposedly it is to be ground-bounce insensitive. Here is a link:

About half way down that article: "Information is carried on the bus as a voltage difference between the two lines. If both lines are at the same voltage, the signal is a recessive bit. If the CAN_H line is higher than the CAN_L line by 0.9V, the signal line is a dominant bit. There's no independent ground reference point for these two lines. The bus is therefore immune to any ground noise, which in a vehicle can be considerable."

Now "immune" is maybe a bit of a stretch here but that should leave available options for common mode debugging. of course, it's not really isolated like LAN but at least differential. And a 0.9V hysteresis is pretty good.

The typical CAN transceiver is operable if the CM is in the range of

VLV

Yes, when it hits the rail it'll fall apart. You seem to be quite knowledgeable about CAN: Why didn't they select an architecture like Ethernet where it is transformer-coupled? From a cost POV those LAN transformers are a dime a dozen these days.

I work for the automotive.

CAN is a common bus architecture. CAN is pretty cheap and simple, and it runs via any pair of wires. The neat feature of CAN is the bus arbitration and the collision avoidance at the hardware level. This implies the DC coupling. With the AC coupled architecture, the star configuration with hubs would be required. Also, that would require some sort of network management. So it would be rather expensive and complicated, especially at the time when CAN was introduced.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

P.S.

The initially proposed CAN standard was also about the fail safe operation. The bus was supposed to be fully operable if either CAN-H or CAN-L is broken or short to the GND or Vcc. That implies the connection to the ground.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

Traditional 20 mA current loop, properly terminated RS-422/485 and CAN bus systems are very closely related and work without a separate ground wire.

In all these systems, one end of the transmitter sends a current into the loop, which flows through some kind of load or termination resistance and return to the other end of the transmitter.

At the other end the voltage drop across the load/termination resistance is sensed (either directly or as the light intensity in the optoisolator in an isolated 20 mA current loop).

RS-422/485 are bipolar current loops, while the "20 mA" and CAN are unipolar with nominally 20 mA in the Mark state or 10-70 mA Dominant state and practically 0 mA in the Space or Recessive state and hence no voltage drop across the load/termination resistance.

The only thing that might justify the separate ground wire is the receiver common mode voltage problem and the receiver transistor biasing. By using galvanic isolation with a floating transceiver power supply, the high resistance receiver transistor biasing resistors will force the floating transceiver power supply around the actual line potential.

Some clamping diodes may be needed between the line and the floating transceiver Vcc and Gnd, if there is a large stray capacitance between the floating electronics and some kind of physical ground, since if there would be a very fast common mode interference on the line, the receiver biasing transistors would not be able to charge that stray capacitance fast enough.

Some RS-232/422/485 converter manufacturers have even removed the signal ground terminal from the RS-422/485 side of their never converter versions.

While the Ethernet hardware these days is dirt cheap, the situation was quite different in the 1980s. The original Ethernet vampire tap transceivers were very expensive, the first Ethernet I built used AUI cabling to avoid the transceivers. Even the first external Thin Wire Ethernet transceivers were quite expensive (but at least comparable to the actual Ethernet AUI card in the computer).

The CANbus is from the same era.

The 10baseT Ethernet became popular much later.

Paul

Try to take two CAN nodes and make them work together without connecting the GND. That may even work at the bench top conditions and for the moderate speed and the short wire length. Practically, the GND has to be connected for good, and the GND problems is the most common reason for the CAN not working properly.

CAN transmitter is pulling in one direction only. If there is a CM, there will be a huge CM glitch with ringing when the dominant level is changed to the recessive.

CAN is bidirectional bus with the arbitration at the bit level. It is fairly sensitive to the fast CM glitches. The OP question was about operating CAN in the proximity of the powerful alternator.

VLV

I was almost going to ask whether you could build cars with less electronics in there but now I better not ... ;-)

Why does it require DC coupling to allow a hub-less design? NRZ would mean more signal analysis to detect a contention but that's cheap these days. It's been a bit long ago but at my first job (80's) we had a coax LAN. There was no hub, just a really long line of RG58 coax which everyone tapped into. No DC-coupling. The only time this LAN ever collapsed was when someone needed a 50ohm barrel terminator and "found" one under the table and the end of "some" piece of coax. He was sentenced to bringing popcorn for all the next day.

Ok, this was in the 80's: We needed ECG trigger inputs for our ultrasound machines so we figured we might as well design a simple ECG unit and put it in there. Must be isolated to IEC/EN (and the old UL544 back then). As usual, this extra feature was supposed to cost next to nothing. So, someone had the idea to use what was then called "Star-LAN" transformers. I had a hard time talking the group out of that because they were only available up to 1.5kV breakdown which isn't enough for medical. But I remember those were really cheap, a few dimes at the most and that was in low quantities.

Custom-made could be much cheaper, provided you need some large 6-7 digit quantity. Around 1990 I designed a switcher where I needed one. Not happy with the 50c price tag we asked a Taiwanese company. They made a custom part for us at a fraction of that.

If you would know how all of that software looks like, you would never drive a car of fly an airplane :-)

Detecting of the collisions of the AC signals at the bit level wouldn't be very trivial.

It would be difficult to distinguish the dominant and recessive levels with the transformer coupled signal.

In the coax Ethernet, the tranceivers are DC coupled at the bus side. It detects the collisions by the DC level. There were bus-type LANs with the AC coupling, however they detected the collisions by the packets destroyed, and/or used the centralized management to eliminate the contentions. You can afford that if you have a lot of bandwidth and computing power, however this is not the case with CAN.

I can remember that too. The coax ethernet was a hassle: every time a cable was jammed somewhere, the whole network was down. I had to check the cabling section by section. An enlightened installer used some pieces of 75 Ohm cable intermixed with 50 Ohm, which also contributed to the reliability...

VLV

After an engine-out over the Atlantic (in a two-engine aircraft) I became a bit hardened in that domain ;-)

But I do prefer cars with the least amount of electronics. The others tend to be in the shop too often.

Ideally that would have to be at the signal signature level. I guess that could be called "sub-bit level".

Yes, that is certainly true.

I have yet to see a client where I don't find a mix of 50/75 coaxes in the lab that look deceptively similar.

That seem to be peculiar to European vehicles... Mercedes, in particular, has big problems here in AZ.

...Jim Thompson