Choosing the Right Ground Point for PCB Design

[amonraa]

In designing a circuit what point should I pick as the ground point? I know it's the reference node in schematic to calculate node voltages and could be anything, but when designing the PCB there should be a physical point that adhere to some conditions as I guess that make it a valid ground point. so is there one and only one point in any circuit in the world that can be designated as the ground?

Thanks.

 

[Simon Bridge]

You can make the actual physical connection anywhere on the ground rail of the PCB. It makes no electrical difference. It is usual to design the ground rail to be quite fat though.

There will be practical considerations like how the ground-lead exits the case etc.

 

[berkeman]

In designing a circuit what point should I pick as the ground point? I know it's the reference node in schematic to calculate node voltages and could be anything, but when designing the PCB there should be a physical point that adhere to some conditions as I guess that make it a valid ground point. so is there one and only one point in any circuit in the world that can be designated as the ground?

Thanks.

Can you say more about your circuit? Different circuits generally dictate different grounding schemes.

In general, you want to use a "star ground" scheme, with the center of the star located with the IO and power connectors, on the edge of the PCB. That gives you the best transient immunity.

In mixed signal circuits with analog & digital areas (and especially with RF mixed signal circuits), floorplanning a good star ground will be critical to the circuit's performance.

 

[berkeman]

This is a good Google search for you -- PCB "star ground"

 

[meBigGuy]

Actually star grounds can have as many problems as other kinds. You need to be aware of ground currents and ground impedances. If you are dealing with voltage sensitive analog systems and high ground currents you need to be careful where the currents flow. If you are designing logic systems with >.5V switching thresholds then it becomes less important.

The proper way to design a ground system is to understand that currents through ground create voltages that act as signals.

Read about ground loops if you are doing sensitive stuff. Make a nice ground bus if you are doing logic. If you have mixed analog and digital, create analog ground and digital ground and connect them carefully (where depends on system architecture).

I can't find a good picture. When signal1 from a module on a crummy ground goes to a module that is well grounded, the crummy ground's voltage appears on signal1.

Braided wire is good for making ground busses. Low impedance is good.

 

[berkeman]

@meBigGuy -- could you elaborate on problems with star grounds? Thanks.

 

[amonraa]

I did some searching on the PCB topics and design guidelines, but I still don't understand why a randomly picked reference point which we call the ground has to be any different than other lines on PCB? For example it has to be low impedance, why not other lines or all? I could pick one of those lines as the designated ground in the schematics.

 

[jim hardy]

...but I still don't understand why a randomly picked reference point which we call the ground has to be any different than other lines on PCB? For example it has to be low impedance, why not other lines or all? I could pick one of those lines as the designated ground in the schematics.

Yes you could pick any point.
It is common practice to use one of the power supply terminals because the power supply goes most everywhere.

Imagine yourself in the shoes of a circuit designer.
When working on the circuit you'd like to pick some common point for your voltmeter's black lead.
The point in power supply to which all current returns makes a lot of sense because when tracing the path of current through a circuit, it's a pretty safe bet that's where it is going. So voltage at any point gives a clue as to how far along its Kirchoff journey that particular current has progressed.

You'll most often see power supply negative, or in multi-voltage systems the common point of the main power transformer's secondary windings. Often but not always that point is tied to the chassis of the appliance and to Earth through the power cord. That keeps technicians safe while working with chassis exposed.
If you study the circuit you'll figure out the reason behind designer's choice.

Modern TV's I've seen have complicated switching power supplies with a high frequency transformer that provides both voltage step down and isolation from mains power..
The power supply circuit board is clearly marked as to which side is connected to power mains , and which side is connected to secondary of the power transformer and the rest of the circuitry.
Mains side is isolated from chassis and not "grounded".
On Secondary side, the common point of the several different low-voltage windings is connected to chassis and all voltages are referenced to that . Those "ground " tracks are large and usually the circuit board's mounting screws go through them .

any help? (I hope I understood the question...)

 

[Averagesupernova]

Actually this is a discussion that can go on and on and on. Always remember that no conductor is perfect and current through any conductor will cause a slight voltage to develop across it. The key is to make connections in such a way so that important things are not affected. Also remember that two PCB traces that run to a common point and are running side by side are acting as primary and secondary windings of a transformer. This can be the result of a star point ground. When you design something a certain way to solve one set of problems you can end up with a new set of problems. But, that's just engineering.

 

[amonraa]

I see...now I'm getting closer...but have problem understanding the chassis. Why is the ground rail connected to it?
Sorry for asking a lot of stupid questions but I cannot recall if I ever learned this at college.

 

[Averagesupernova]

There are hosts of reasons for connecting a chassis to 'ground', or circuit common. One reason is shielding. If the metal chassis is just floating about then it can capacitively couple lots of noise into higher impedance parts of the circuit. Also, a mistake I often see, at least a mistake in my opinion, is that the chassis is allowed to carry ground currents. Imagine getting a batch of circuit boards and there are some defects in some of them where the ground trace is not complete. Now the chassis carries it. This can cause problems because now there is a part of the product that is carrying current that normally doesn't. It may induce noise into parts of the product where it normally does not. Imagine trying to troubleshoot this.
-
Another thing I have seen that is not cool is grounds are laid out in such a way that when something goes wrong a ground trace blows open. That in and of itself is not such a big deal. But the example I am thinking of the ground trace that opened was the ground reference for the 5 volt supply. Now everything on the 5 volt supply is running at 12 volts or so. This comes from multiple paths for the ground. Those multiple paths have the 5 volt devices still grounded yet but not the regulator. A bad idea.
-
Amonraa, don't feel guilty for asking these kinds of questions. This sort of thing is usually learned the hard way.

 

[meBigGuy]

An example of a problem with a star ground is the example in my post. When signal1 from a module on a crummy ground goes to a module that is well grounded, the crummy ground's voltage appears on signal1.

The definition of a crummy ground is one with a voltage drop, generally because it is switching current. If any leg of the star has a voltage drop, it will be seen as signal at any other starred component it goes to. The only perfect ground is a perfect ground. You always need to understand where the ground currents are flowing because they cause voltage drops that are imposed on signals.

So, even a star ground can have issues if it isn't "starred" with knowledge of where voltage drops are occurring.

Let's make up an extreme example of a ground problem to help the OP visualize the issue.

Imagine an optically coupled linear isolator that is detecting a 1 amp sine wave current. Assume the poorly designed isolator has 1 ground pin, 1 input pin, and 1 output pin that magically outputs a voltage proportional to the rms current at the input. There is a one amp sine wave flowing out of the ground pin to a good ground through a wire, say 0.1 ohms. This means there is a 0.1V sine wave on the isolator's ground pin, and the output's dc signal is referenced to that bad ground. So a receiving op amp that is connected to the good ground sees a 0.1V sine wave superimposed on the dc rms signal.

So, reduce the 0.1 ohm ground connection, but there is always a common mode voltage imposed on the output signal.

One could also fix this problem by choosing an isolator with 2 grounds, one for the high current, and one for the output signal (which is normally how it is done).

 

[meBigGuy]

If you don't connect the chassis to ground it could develop a voltage. Also, the chassis sometimes provides the best low impedance ground plane for the system.

There are two schools of grounding:
1. Isolate grounds and ground currents and controll everything carefully.
2. Ground everything to everything with as low an impedance as possible.

The second works OK for digital systems, and the first is required for sensitive analog systems.

 

[amonraa]

This is very helpful, thanks for all!