Picking up where we left off in Part 1, we need to explore some methods to mitigate problems caused by unpredictable ground currents.
Let’s start with some simplification, although we may need to first complicate things to then simplify them. First, let’s arrange a massive ground plane around the analog circuitry on the PC board. Then we can connect the #5 ground connection from the sensors to that ground plane near the op amps processing the sensor signals. See Figure 1, a repeat of Figure 2 from part 1. The #2 ground connection is gone.

For that op-amp circuitry, given that it is powered from a dual supply (the ±12 VDC), theoretically, there won’t be any supply current flowing into the ground. Realistically, there will be some due to (1) the op amp supply line decoupling capacitors to ground and (2) any impedance networks between the op-amp outputs and ground. You could route these ground connections through separate traces (rather than the ground plane) back to the point on the PC board where the ±12 V enters the board. This will minimize voltage disturbances in the ground plane, but there’s a tradeoff. The inductance of those separate ground traces we added for items (1) and (2) will not be trivial and may introduce unexpected problems. Those include ringing in the op amp output signals (if they are high-bandwidth devices) or parasitic oscillations. Thus, it’s probably safer to return items (1) and (2) to the ground plane.
For the two analog supplies connecting to our PC board (±12 VDC and +5 VDC), their return lines (±12 V AN_RTN and the +5 V AN_RTN) should be connected together at only one point, and that point should be at the ground plane described above.
Having considered some of the grounding issues, let’s not go any further until we consider the power (not ground) routing. If there are multiple PC boards located some distance apart that would keep you from having a single point return for the two analog supplies, you may want to add an isolated DC-to-DC power supply (circuit or module) to help isolate the power return connections. In fact, if you do this, you could change the analog supplies to just a single positive voltage supply. You could distribute simply +12, +24, or +48 VDC and create all necessary voltages at each analog PC board using (for example) a low-power flyback inverter topology (Figure 2). That would provide isolation and multiple voltages and polarities. Doing this would allow the power supply return connections to be placed very close to the analog circuitry drawing current.

Connecting a sensor
Look closely at the sensors and the analog circuitry that process their signals. Recall from Figure 1 that pesky #6 and #6a (multiple) ground. If possible, isolate the sensor from any ground connection. Don’t ground either sensor connection at their remote location.
If the sensor is a two-terminal device, such as a thermistor, photocell, 4-20 mA device, inductive proximity device, piezo-electric device, etc., you can already think of it as providing an isolated, differential signal. Thus, you can bring both of its terminals to the DAS through a shielded, twisted pair cable. Connect that shield only to the DAS ground plane, but leave the end by the sensor unconnected, or you’ll create a ground loop. Be sure to keep the sensor signal (mostly) noise-free, processing its output with a true differential input amplifier. Use an instrumentation amplifier, not simply an op amp configured with a differential input.[1] There are multiple articles written that explain the difference between these and detail what can go wrong.[2] The short version of the problems with the diff-amp: The bias currents and the input impedance looking into the (+) and the (–) inputs are not the same. See Figure 3 for a comparison of the topology of these two amplifiers.

Some sensors have active circuitry integrated into the housing; the connections may include power, ground, and the signal. In this case, the signal is “single-ended” rather than the two-wire differential version previously described. It can be handled in a similar way, sometimes referred to as pseudo-differential. See Figure 4 for an example. This is an electret microphone with a built-in FET amplifier. It needs power and ground and has a single connection for the output. We can use a cable with a twisted pair for the ±OUTPUTs, a separate wire each for +5 VDC analog and +5 V AN_RTN, and a shield connected as shown.
For the digital lines, if a portion of the digital circuitry (perhaps a remote module) is powered from an isolated supply, they can probably share common ground connections without causing problems. If the digital circuitry is not powered from an isolated supply, you can break ground connections associated with the control lines by interposing a digital isolator. Such devices use various technologies such as optical (LED + photodiode), AC (modulator + transformer + demodulator), magnetic (electromagnet + Hall-effect sensor), or capacitive (modulator + capacitor + demodulator).

Figure 5 shows an agnostic version of this with a generic symbol for the isolator device. Note that these isolator devices contain active electronics and therefore need their own power and ground connections.
For the isolator receiving a data line on Board B from Board A, that device has connections for power and ground brought in from Board A. Thus, use three wires or PC board traces: power, ground, and data. The data is a digital signal with respect to the same ground-carrying power. Unlike the analog signals previously described, small amounts of ground current or noise shouldn’t cause signal-integrity issues.

For the isolator receiving a data line on Board A from Board B (running in the opposite direction), that device has connections for power and ground brought in from Board B. Again, use three wires or PC board traces.
Keep in mind that significant current draw or rapidly changing current draw (typical in a digital system) may move the ground potential around (ground bounce) with respect to the ground potential at the power supply. Proper use of bypass capacitors, as previously described, is essential to minimize this condition.
If you follow these general guidelines, you’ll have a higher rate of success and less need for multiple iterations of PC board layouts.
References
[1] What are the advantages of a three op-amp instrumentation amplifier?
[2] What’s the difference between instrumentation and precision amplifiers?





Leave a Reply
You must be logged in to post a comment.