AI support on current sens pcb layout

[mortenlund]

Hi All

After consulting an AI on pcb layout there is a little unclarity.

The AI warns about ground loops. On the attached foto there is a huge shunt and an current sens amplifier TLC2020 not mounted yet. The foto shows the ideal layout from the manufactures viewpoint. The AI suggest to split the ground plane (both sides of the pcb) into an island for the shunt side that is not gnd to the amplifier.  Notice the pcb on the foto has solid groundplane on the backside and vias stiched all around the outer perimeter. On the top side, the huge areas around the shunt is all gnd.

All suggestions and viewpoints are highly appreaciated.

[golden_labels]

Hello,

Consider consulting a Magic 8-ball to get the answer on PCB design. Follow its advice, that will end well.

What problem did you encounter, that made you disregard manufacturer’s recommendations? Where are the ground loops you’re talking about? Show us the schematic of your circuit, because in the picture attached ground is attached only to the amplifier. What are the currents involved?

[mortenlund]

@golden_labels

If understanding you right, it is a little like antenna design. Some woodo is involved.

PCB (the "floating part in the middle and main plus wire and main GND wire to the left and at bottom.

Only ground loop - if understanding it right, is the connection to the NTC from the pcb?

[mtwieg]

@golden_labels

If understanding you right, it is a little like antenna design. Some woodo is involved.

No, not really.

"Ground loops" are poorly-defined concepts which usually can't be identified based on a schematic alone, and usually only become relevant when multiple boards are connected together with a common reference/ground. I don't see any possible ground loop in your schematic.

LLMs often mix useless/unrelated information with their responses, this is one of those cases.

[mortenlund]

@mtwieg

Thanks for clearing that out. Concerning the full ground plane on the back side of the pbc. Would you make an island that start in the middle of the shunt and follow the trace on the top till the connector (it will be a tht connector), or should there just be a smaller island on the back side where the connector pokes through? Does it matter in dc environment, high current?

[golden_labels]

If understanding you right, it is a little like antenna design. Some woodo is involved.

No, there is no antennas involved at all in this design.

If you mean the word “antenna” used in the datasheet, where they discuss PCB layout recommendations, it’s about creating an unintended antenna. The chip is intended to work with very low currents, where electromagnetic interference can easily induce unwanted signals. Hence nothing connected to this chip should act as an antenna, and current loops should have possibly small area. You can see an example of this approach in how sense lines are laid out. The red part is the loop area being minimized, with the loop arae itself marked in a fragment of your schematic:



Other loops would be formed by having long power line traces, but area is already minimized with the ground plane being used.

Only ground loop - if understanding it right, is the connection to the NTC from the pcb?

A ground loop is any current loop that involves ground.

From Ohm’s law (V = I·R), any current in a conductor is going to produce voltage. So real ground conductors are never truly at 0 V, if they carry current. These are usually minuscule voltages. But if we work with equally tiny signals or signals are amplified, that matters: that tiny offset is added to our tiny signal or is going to be amplified. Now observe that the current draw from nearby circuits is rarely steady, but instead is fluctuating. Which basically means it’s injecting a ton of noise into what we’re trying to actually process. Simply because current goes through the same ground conductor.

Does it matter that much? Is it as scary as the previous paragraph makes it sound? In general: no, not really. With most circuits it’s not going to matter, ground planes already work great for eliminating the problem, and the shift to using higher speed (means: higher frequencies) circuits makes the problem even less pronounced than it would be in the past. But it is a problem one should certainly be aware of.

This goes beyond my detailed knowledge, so I’ll leave this part to more experienced forum members, if you have more specific questions.

The datasheet mentions separating grounds. I believe the authors just handwaved the entire topic, assuming the reader knows the problem already. It’s about separating this chip’s ground from a ground that powers digital section. That is: if you have some heavy digital circuitry in your design, its ground should not be connected in a way that would make currents affect this amplifier’s ground. It does not seem to be the case in your schematic.

[mtwieg]

@mtwieg

Thanks for clearing that out. Concerning the full ground plane on the back side of the pbc. Would you make an island that start in the middle of the shunt and follow the trace on the top till the connector (it will be a tht connector), or should there just be a smaller island on the back side where the connector pokes through? Does it matter in dc environment, high current?

I don't really grasp what you're suggesting here. But when you say "island" I presume you mean separating the plane into two completely isolated nets.

But the battery can't be completely isolated from the current sense amplifier, as the shunt amplifier has a limited input common mode voltage range for it to operate properly (it's not an isolated amplifier). So there needs to be some sort of connection between the current sense amplifier circuit GND and the battery itself itself. Your schematic shows the shunt connected on the "low side", i.e. the shunt is in between the circuit GND and the battery negative terminal. That's a fine way to do it. An alternative way is with a "high side" shunt, in which the shunt is connected to the positive battery terminal (the TSC can measure the shunt so long as its common move voltage is within 0-100V). You could then connect the battery negative terminal to circuit GND directly on the board, or at some other point in the larger system.

There are tradeoffs when choosing between high and low side shunts, but which one is right for you depends on the specifics of your system ("the big picture"). There really is no general rule for this sort of thing. For all we know, there are several other boards which interact with this battery and shunt amplifier.

[mortenlund]

@mtwieg

Thanks for your explanation. It is highly appreaciated. Am a beginner to this.

Will try to be a little more precise. Have attached a png. Marked with red, is suppose to show where a cut out (island) could be made on the back side. In blue is showen a line to mark a much smaller cutout on the backside that is just the size of the solder pad. Some text with which side is which and that the whole back side is one big system ground - except the area marked on the png.

Would you make the big cutout or the smaller that just include the solder pad?

I actually consider moving the connectors closer to the shunt. What do you think about that idea?

[golden_labels]

The area marked with red and blue is a trace that carries current to be measured. What line of thought brought you to a conclusion, that increasing ground impedance is bringing any improvement? What even brought you to putting that subject under consideration? I don’t see anything in the documentation or this thread, that would.

I do not see any reason to move connections closer to the shunt. What is yours? What problem are you encountering?

There is no need to send notifications (with the @member feature) each time you reply in this thread.