Chassis and multiple voltage grounding

[wesg]

I'm currently designing a robot and want to sort out my electrical system grounding.

The planned circuit is powered by 2 12V SLA batteries in series fed to a main fuse block and split using BECs to onboard 5V and 12V power. Those other voltages are provided by 2 more fuse blocks, for a total of 3. Each block has a ground stud, but each BEC also has a negative wire for the converted voltage, so should each fuse block ground be connected to the battery negative in addition to the BEC negative wire? I don't have a schematic handy yet but will be working on one from the feedback.

Second, how, why and should a robot chassis be grounded? The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

[Kohanbash]

I'm currently designing a robot and want to sort out my electrical system grounding.

The planned circuit is powered by 2 12V SLA batteries in series fed to a main fuse block and split using BECs to onboard 5V and 12V power. Those other voltages are provided by 2 more fuse blocks, for a total of 3. Each block has a ground stud, but each BEC also has a negative wire for the converted voltage, so should each fuse block ground be connected to the battery negative in addition to the BEC negative wire? I don't have a schematic handy yet but will be working on one from the feedback.

Second, how, why and should a robot chassis be grounded? The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

You should have all your grounds attached to a common point. This is often done by having all the fuse block ground pins tied together, and having the battery ground going to one of the fuse blocks.

[jwatte]

Also, if your question is "how come my BECs have ground wires both going in and coming out?" the answer is "because that's what customers expect."
It should be perfectly safe to tie those together. Some BECs may have isolated inputs/outputs, and some circuitry may want to be isolated compared to common ground, but those cases are the exception rather than the rule most of the time.

[knossos]

Second, how, why and should a robot chassis be grounded? The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

Motor controllers are really noisy (electrically).  I would assume that if you attached the Sabertooth 2x25 to the chassis as a heat sink while using the chassis as ground, that noise would then be induced into all your circuitry including your microcontroller, sensors, etc.

[bdeuell]

The planned circuit is powered by 2 12V SLA batteries in series fed to a main fuse block and split using BECs to onboard 5V and 12V power. Those other voltages are provided by 2 more fuse blocks, for a total of 3. Each block has a ground stud, but each BEC also has a negative wire for the converted voltage, so should each fuse block ground be connected to the battery negative in addition to the BEC negative wire? I don't have a schematic handy yet but will be working on one from the feedback.

The two ground wires are likely tied together internally and provided for ease of wiring, in this case it would be advisable to not tie them together at a second point outside of the device as it would result in a ground loop.

why and should a robot chassis be grounded?

There are three reasons I know of to ground a chassis:

1) For safety (typically in high voltage AC systems) where the chassis is connected to fault ground and does not actually carry any current under normal operation but provides a path for the high voltage to reach ground if it finds its way to part of the metal chassis.

2) For convince, this is very common in cars where the chassis is uses as a common ground reducing the amount of wiring needed. This design however is prone to noisy electrical power, can result in ground loops, and can result in galvanic corrosion.

3) For EMI, where the chassis is used as a shield and would need to be a full enclosure.

The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

I have not used the Sabertooth controllers before and I was unable to find mention of any required insulation in their manual. There are many transistors where the package is connected to the drain or collector and thus must be electrically isolated from ground to prevent a short, but I don't suspect that is the case for this controller. Perhaps you can point out where you found this requirement.

[jwatte]

not tie them together at a second point outside of the device as it would result in a ground loop

For all intents and purposes, I don't think ground loops are avoidable in these kinds of systems. If nothing else, then each PCB that has a ground plane with holes for components through it, has a "ground loop" of some sort. Although small :-)

For example, let's say you power a microcontroller from your battery, using a linear regulator.
The microcontroller outputs servo control signals, which are "pwm" and "ground." That ground is the ground of the linear regulator is the ground of the battery.
Now, you power the servo motor using a BEC from the battery. The ground into the BEC is the ground from the battery. But you also need the ground for the control signal.

The options are:
1) run only the signal wire from the microcontroller to the servo, and use the BEC ground as the reference
2) run signal and ground to the servo from the microcontroller, and also run BEC power and ground, and merge the grounds

You have a ground loop, because you have the microcontroller sending the PWM signal, and you have the BEC sending the power signal, and they both share the return ground.
Using option 1) means that you are guaranteed to have the noise from the servo motor AND the switching noise from the BEC on the ground that your PWM is referenced to.
Using option 2) there is a more direct path (or return path) for the PWM signal to the microcontroller. Yes, this is another loop, but it seems to me that it's also a better impedance match and will pick up less of the BEC noise and motor noise.

Now, I'm not an EE -- I may be getting something totally wrong in this analysis -- so I'd love to understand this trade-off better.

[bdeuell]

I agree ground loops may not always be easily avoidable, especially when constructing a robot using multiple off the shelf components. However I do think that it is beneficial to understand the potential impact as well as avoid them where practical.


In the case wesg presented assuming that the grounds are tied internally in the BEC I think there are a couple of practical options.

a) If devices can be supplied solely from this voltage source then they could be simply attached to output of the BEC. The two ground wires would not need to be tied together at a second point in this case.

b) If devices require multiple input voltages then it may be desirable to create a single ground bus. in this case it might be better to not connect the output ground wire and connect only the input ground wire of the BEC. This would avoid the creation of a ground loop. My understanding is that if both wires were tied  back to a single ground point outside of the BEC it could result in a slightly different ground reference voltage at the BEC and would manifest itself as an inaccurate output voltage. Of course if the two ground wires are routed alongside each other and then tied together outside the BEC the effects of a ground loop would be minimized.

Note: in both cases the grounds are common if the assumption that the grounds are tied together inside the BEC is correct.


In the case  Jwatte presented with a servo powered from a BEC
I assume Jwatte is referring to a common hobby servo with a three wire connector.
Again I am assuming the ground from the battery and the ground of the BEC are tied together.

1) run only the signal wire from the microcontroller to the servo, and use the BEC ground as the reference

I agree that this design is more likely to have interference from the servo as well as the BEC affect the signal.

2) run signal and ground to the servo from the microcontroller, and also run BEC power and ground, and merge the grounds

My major concern is that by creating a ground loop with the signal and power grounds, the high current drawn by the motor could travel through the signal ground path which may not be designed to handle the higher current. I would not expect the inaccurate ground reference voltage that is usually the concern of a ground loop to cause a problem for the digital PWM signal. I think this mostly causes problems with analog signals.

I think there is a third option that might be a better design but is probably not as practical.
The ground and power to the servo could be supplied from the BEC. The PWM and ground of the signal could be tied in using an opto-isolator. this would maintain the ground reference for the signal from the microcontroller but would eliminate the ground loop.


I am not an EE either

[jwatte]

This would avoid the creation of a ground loop

I think it still creates a ground loop. Specifically, there are two separate supplys, sharing a single common ground, and thus the common return ground resistance generates a voltage divider into each of the supplies, where they are thus capable of affecting each other.

I agree there is a danger if the impedance along your main power ground return path is not lower than the impedance along the signal ground; this can burn up the signal ground wire. If you use copper wiring throughout, and size the wires appropriately, the impedance along power ground should be lower than signal ground, and thus the majority of the current from power should return there. (There may still be effects on the signal -- the question is how to minimize these.)

Optoisolators between each component is an interesting idea, btw. I had previously not considered that because of complexity, but it's not really THAT complex...

[bdeuell]

I have attached an image of the circuit for reference

There is not much documentation on the BECs in question but given that they are described as "non-isolated" I assume that the input and output grounds are tied internally.

In figure one both grounds on each BEC are tied together this creates two ground loops through the BECs.

In figure two the elimination of the highlighted (red) connections breaks these ground loops. The ground connection of the BEC is provided for reference voltage and will only carry the small current drawn by the BEC itself not the full power output. The voltage divider condition you mention makes a good point; this could be reduced my placing the ground connections of the BECs as close together as possible.

[jwatte]

You still have a common return, which means you still have a ground loop.
Specifically, in the lower-right, where it says "GND," there may be a small resistance. The "24V" voltage and the "5V" voltage both exit through there. Whatever current the 24V sends through that ground resistance will build up an I-R voltage drop, seen by the BEC.

If you have separate grounds, as in the top diagram if you break the red connection and connect that to your separately-powered device instead, then the BEC will generate 5V between its output ground and its voltage rail. This is actually more isolated from the 24V source, assuming the source GND is tied close to the battery negative.
However, in that second case, once you have any signal going into the part powered by the 5V rail, that signal will then see a voltage differential, so why the 5V may be cleaner by using the output ground, the signal may see more interference. Whether that matters depends on the signal :-)

[bdeuell]

I agree that there will be a slight difference in the ground potential at each of the BECs in figure 2 due to the voltage divider effect you described. As I mentioned before this effect could be reduced by moving the ground points of the BECs closer together instead of distributing them along a ground bus as I drew them in the circuit schematic. I am not seeing ground loop you are saying still exists in figure 2, perhaps I am misinterpreting the definition of a ground loop.

I did not draw a schematic where the BECs would be used to power separate devices and a single ground bus would not be necessary. In this case like you say the output ground wires could be connected directly to the devices. However as soon as you add a signal wire between two devices on the separate power supplies it might be better to use a design with single ground bus.

[jwatte]

perhaps I am misinterpreting the definition of a ground loop

Once you add two different loads on the right (one to 24V, one to 5V,) they make a ground loop.
http://en.wikipedia.org/wiki/Ground_loop_(electricity)

The two options I'm aware of:
1) Build everything isolated. A total PITA and costs go up.
2) Over-provision ground leads, and hope for the best.

[johnwarfin]

i am an ee and also hobbyist so can suggest ground loop issues are usually solved with two words: "star ground". the bec grounds are internal short so those two diagrams are basically the same.

a much bigger problem for me has been noise propagated on the supply rail. also easily solved. this time with cap and resistor or inductor in the supply line of sensitive circuits like mcu.

[bdeuell]

the bec grounds are internal short so those two diagrams are basically the same.

I have been told that creating a loop in the ground path, as in figure 1, can result in undesired currents from em waves passing through the loop. These currents can cause different potentials at points on the ground path. As the current drawn by the BEC will be small I was thinking it could be susceptible to these currents and the resultant potential differences on the ground.

Please correct me if all or part of this thought process is faulty.

[jwatte]

Any wire can act as an antenna, and any looped wires can act as an antenna/inductor coil, true.
However, that's not the specific problem with ground loops -- this is true for any wire, no matter whether it's ground or not.
Most of the time the EMI picked up is too small to matter.
If you wire things with supply and return close to each other, it will also largely cancel out.

johnwarfin: How do you "star ground" a design that has battery supply going to a MCU board, that then sends a control signal+ground out to a servo, where the power and ground of that servo is also wired straight to the battery?

[bdeuell]

I had a lengthy discussion with an EE the other day regarding ground loops. The understanding I came away with was that if there is a dead short between the grounds of a device internally connecting them externally would create a loop. This loop could be considered a ground loop, however it is very small and for all practical purposes not an not likely to cause a problem. Particularly in this case where the internal circuitry of the device is not understood/documented it is more important to connect all grounds to insure proper operation. All of the BEC grounds could be tied to a single point or if you do not plan to use the unregulated battery voltage anywhere in your system you could just tie the IN grounds together and the OUT grounds together.

I think ground loops become more of a concern when you start connecting signal cables (that include a ground conductor) between devices, especially when the signals are low current and analog.

[Corpze]

Second, how, why and should a robot chassis be grounded? The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

Where have you read this? I did have some serious grounding problems, at least what i think it is, frying two sabertooth's

[wesg]

Second, how, why and should a robot chassis be grounded? The documentation for the Sabertooth 2X25 says a metal chassis can be used as a heatsink when attached directly but if the chassis is grounded, the controller should be insulated. I'm curious why that is.

Where have you read this? I did have some serious grounding problems, at least what i think it is, frying two sabertooth's

The Sabertooth 2x25 manual specifically mentions if you ground the chassis, but I haven't read enough robot-specific documentation to see that grounding is specifically recommended. I was hoping that we'd get a discussion going about the pros/cons.

[Corpze]

I spoke to D.E's helpdesk yesterday, they don't recomend having the chassi grounded at all, it can cause problems.

[jwatte]

they don't recomend having the chassi grounded at all, it can cause problems.

Everything can cause problems! I'd like to know what problems they are specifically worried about.

For example, a common "problem" with grounded chassis is that, if you happen to touch a conducting wire (or bottom of board) to the chassis, it will be shorted to ground! With an un-grounded chassis, you need to happen to touch two wires together for that to happen, which is less likely.