1

We have a display and a sensor that are installed in vehicles. The sensor is in a metal TO Can, and the manufacturer shorts the ground pin to the TO Can body for shielding. We mount that can in a metal probe, which of course touches the TO Can and so is also grounded. The metal probe is mounted to a plastic cover in the vehicle. However we have found some vehicles have metal covers now, which have electrical continuity to the chassis. The result is a ground loop which can cause intermittent failures of communication:

Block diagram of ground loop in vehicle through chassis and sensor probe

I have seen circuits where they separate chassis and signal grounds using resistors and capacitors. Could that mitigate the ground loop in this case, if one or both of the PCBs isolated their digital grounds from the chassis ground with a resistor and capacitor in parallel? I have also heard that resistors/capacitors to isolate grounds is a bad practice, so I'm confused why I see it on some automotive circuits.

I also thought of using opto-isolators so that grounds need not be shared between sensor and display. However, the problem with that is that if the sensor gets mounted in a plastic cover, then it has no connection to the chassis ground anymore. In that case the sensor would need to use the display's ground as a return path.

Luminaire
  • 355
  • 1
  • 8
  • 1
    You have to provide the can sensor data. and what circuit. Some can sensors are differential output (balanced connection) (signal + is a positive going signal, and signal - is a negative going signal) vs single wire output (unbalanced) (just a signal + out and ground). They are distinctly different, and in how you would interchange them. – David Mikeska Apr 23 '21 at 21:28
  • 1
    You show 2 twisted pairs (4 wires) in the cable. Can another one be added? Where is the shield connected? Is the entire circuit (including display etc.) independent of other circuits in the vehicle? – Bruce Abbott Apr 23 '21 at 22:00
  • 1
    My guess is that you are not connecting the sensor board with I2C (as seems to be indicated in the rough schematic) but with some other communication medium (CANbus?). Could you please edit the question to add which medium you are using for communication? Or am I wrong in my guess? (I2C is really not meant for "remote" communication) – Math Keeps Me Busy Apr 23 '21 at 22:00
  • @DavidMikeska The sensor is a TPiS1T 1086L5.5/7452 . The output is an I2C bidirectional open-drain configuration, so the signal lines are either driven low by a bus device, or float high via pull-up resistors. Does that help? – Luminaire Apr 23 '21 at 22:01
  • @BruceAbbott adding another wire is possible. The shield is connected to the Display PCB ground. Yes aside from the battery and chassis ground connections, the circuit is independent from other circuits in the vehicle – Luminaire Apr 23 '21 at 22:05
  • @MathKeepsMeBusy yes this is I2C, and you're right that it is not ideal and CAN Bus would be preferable for this scenario. But since we communicate fairly slow (20 kHz), only have 2 devices on the bus, use shielded cable, and only transmit a small amount of data (10 bytes, once per second) I2C performs well. – Luminaire Apr 23 '21 at 22:10
  • @user263983 That DC/DC converter is interesting, maybe that paired with a signal isolator could do the trick. It's pretty large and expensive – Luminaire Apr 23 '21 at 22:11
  • 1
    How are you connecting the chassis to ground? Are you running on steel wheels without rubber tires or do you have one of those dangly straps (never were very effective either)? – Solar Mike Apr 23 '21 at 22:12
  • @SolarMike haha I guess that symbol is technically "Earth Ground" I should have used the upside-down triangle for "chassis ground" – Luminaire Apr 23 '21 at 22:13
  • 1
    Series inductor may work depending on other issues. – Russell McMahon Apr 24 '21 at 00:51
  • 1
    I take it this is a pretty long run on the shielded twisted pairs going to the sensor, and more than 2-3 inches from Display pcb, correct? When you are wiring an i2c on wires, its recommended that SDA and SDC are not a twisted pair rather SDA and power is one twisted pair, and SDC and ground be the other twisted pair. Beyond this, you could star ground your devices together to remove any differences in ground. – David Mikeska Apr 24 '21 at 01:25
  • 1
    As far as the resistor cap isolation, its not needed here. There are some circuits that benefit from this type of bonding, but there are only rare cases where it is needed on com interface lines, and most of the time, there is always ways around that. So I don't explain where that is truly needed for circuit operation. – David Mikeska Apr 24 '21 at 01:25
  • @user263983 I know right? It doesn't seem like much but yeah that would be a big hit, but obviously we'll do what it takes to fix the issue so I'm looking in to it – Luminaire Apr 24 '21 at 17:39
  • @DavidMikeska yes the cable is up to 12 meters in length. Great suggestion, yes we have SDA and SCL in separate pairs to isolate them somewhat. So by star ground, do you mean adding an additional grounding point to the system somewhere to help even things out? – Luminaire Apr 24 '21 at 17:43
  • @RussellMcMahon I was wondering about that, if we need an inductor to also mitigate high frequency ground loop currents. Maybe in series with a small resistor? – Luminaire Apr 24 '21 at 17:46

2 Answers2

3

If it's automotive, consider EMI for your solution as you likely need it qualified.

You have to prevent a large loop between battery positive, display PCB, LDO, sensor PCB, sensor, chassis, back to battery negative.

This happens if you cut the chassis bond of the display board, although it will solve the ground loop. The large loop will remain and cause a lot of stray radiation pickup and voltage spikes in your circuit.

I can think of two solutions instead:

Option 1:

Use an isolated 12->5V converter to create power for the sensor and run the power and data lines each through common mode chokes over to the sensor. Ground the isolated sensor power negative on the sensor board and leave it float on the display board (cap-attach it to the display ground as per dc/dc converter design rules). Bond the chassis, to both boards and the the cable shield on both sides.

To power the display board, you must not use the isolated sensor power. You can use a local non-isolated psu for that (e.g. your LDO).

Option 2:

This is without an extra PSU, so potentially lower cost. It is also a variation of Bruce's answer. Remove the 12 V connection from the Display Board, and feed 12V power into the sensor board instead. Pass this 12V and a return wire (with Common mode choke) through the cable to the display board's LDO.

In this case, too, keep both boards' end of the shield attached to local chassis. But the Display Board's ground must not be DC-bonded to chassis. Only RF-bonded.

tobalt
  • 18,646
  • 16
  • 73
1

You need to get the display PCB and sensor PCB grounded at a single point, and that can only be at the sensor itself. So disconnect battery ground from the display PCB (LDO etc.), feed it through a separate wire to the sensor PCB, and 'ground' it as close as possible to the sensor. Then run that ground back to the display PCB through the existing ground wire and/or shield.

This will eliminate voltage dropped across the ground wire between PCBs due to the ground loop, but there could now be some ground noise caused by operation of the display circuit. However with good MCU bypassing it should be OK because I2C has relatively high noise immunity.

Another way to do it is insert a resistor in the ground wire from the battery to the display PCB. This will increase the ground loop resistance and reduce the current flowing through the ground wire between PCBs. The resistor value needs to be low enough that the LDO still gets enough voltage to work, so it cannot eliminate the ground loop current between PCBs - but might reduce it enough to solve your problem.

Bruce Abbott
  • 55,540
  • 1
  • 47
  • 89
  • 1
    This will form a huge loop through 12V, LDO, display PCB, sensor PCB, sensor back to battery.. Difficult EMI situation. It will be much better to consider a solution in which both sides remain chassis bonded. E.g. an isolated 5v psu in the display board. – tobalt Apr 24 '21 at 06:00