Attaching AC Earth to DC Negative

[xortan]

Hello everyone,

So I am doing this EMC course and got to a section about cable shielding. It was suggested that the best way is to have a 360 connection to the chassis and to connect the ground of the PCB to the chassis as to provide a better return path for the current. It has been suggested a few times in this course to bond the chassis ground to the PCB ground.

For some of the products I design I will use a rackmount case that will have some AC/DC power supplies in them. These power supplies will provide power to various PCBs inside the case. I always attach the AC Earth to the chassis for safety reasons.

What I am wondering is if it is safe to attach the DC ground to the chassis if that is being used for the AC earth? Will this remove the isolation provided by the AC/DC supplies? Is may be better for EMC, but am I compromising safety by doing this?

Any insight into this would be greatly appreciated. If you want/need more information please don't hesitate to ask!

Thanks!

 

[Averagesupernova]

Generally unless there are isolation requirements it is not a problem to have AC conduit ground bonded to one or more of the power supply grounds. I do have some concerns with how you describe using the chassis to 'assist' in carrying current on the PCB. I have posted in several other threads on this forum about doing this. Here is one: https://www.physicsforums.com/threads/designing-pcb-ground.708346/#post-4490834

 

[berkeman]

No, it's generally fine to connect Earth ground to your PCB grounds. There are times when the requirements are to float your PCB ground, but in those cases you can use an RC connection instead (R~100k and C~0.1uf, 2kV)

EDIT -- ASN beats me again...

 

[xortan]

Ah, so it will remove the isolation. A typical power scheme we will have is:

AC/DC supply creates 48 VDC

48 VDC goes into circuit and gets converted to 12V to drive a motor (we usually isolate this)

48 VDC also creates two 5 VDC rails, one for comms and the other for video (both isolated from each other)

In this scenario I wouldn't want to connect all the grounds together as the isolation will be removed?

If I am concerned about radiated emissions on a cable attached to this box, should I only ground the ground for the power supply of the signal that is on the cable?

 

[Averagesupernova]

Ok, not to offend, but this seems to be over your head a bit. Are you using a switching supply to convert the 48 VDC to the 12 and 5 VDC? How do you maintain isolation otherwise? I am concerned that you were unable to grasp that sharing the AC conduit ground prevents isolation but yet you refer to several 5 volt supplies that are isolated from each other. How can you understand one and not the other?

 

[xortan]

Of course it is over my head a bit, thus why I am posting on this forum :)

Yes I am using a switching power supply to convert 48 VDC to the 12 and 5 VDC.

I get that sharing the AC Earth with the DC negative removes the isolation. What I am not understanding is how I would handle the shielding of the cables attached to this box. I am trying to reduce the effects the common-mode current will have on emissions by providing a return path for the current through the cable shield.

In the EMC course I am taking they say that an ideal cable shield configuration is a 360 degree bonding to metallic chassis at either end and that the chassis connects to PCB ground in the I/O area and at several other areas on the PCB.

This confused me as we need to attach the AC Earth to the chassis for safety and certification purposes. Our PCBs also have a bunch of isolated rails.

Maybe it is not possible to do both and I should attach the cable shield to the PCB plane instead of the chassis?

Edit:

Thinking about attaching the cable shield to the PCB and I don't want to do that. The common-mode noise voltage source would then excite the shield and drive a common-mode current on the outside of the shield, which would radiate emissions

 

[Averagesupernova]

If you read the thread I linked to you will find that my opinion is to avoid allowing ground currents to go where it is not necessary. Now, I would assume that you have a RF device or similar where the shield of your cable connects several devices together that are separated by some distance and plugged into different outlets. So, it is unavoidable to prevent a certain amount of AC mains current (although miniscule) from flowing on the cable shield. Here is where the 360 degree connection you refer to comes in. If the shield is symmetrical the whole way including connectors, it does not matter if some unwanted current flows on the shield. It will cancel and will not induce any currents in the center conductor. However, that is not to say you cannot get noise on the cable but that is a different branch of this discussion. Not all equipment has the shield of a coax cable connected to conduit ground. I am not sure what your situation is so I cannot recommend you do this. As I often say here on PF, can you tell us more?

 

[berkeman]

In the EMC course I am taking they say that an ideal cable shield configuration is a 360 degree bonding to metallic chassis at either end and that the chassis connects to PCB ground in the I/O area and at several other areas on the PCB.

This confused me as we need to attach the AC Earth to the chassis for safety and certification purposes. Our PCBs also have a bunch of isolated rails.

Maybe it is not possible to do both and I should attach the cable shield to the PCB plane instead of the chassis?

Edit:

Thinking about attaching the cable shield to the PCB and I don't want to do that. The common-mode noise voltage source would then excite the shield and drive a common-mode current on the outside of the shield, which would radiate emissions

Can you post a detailed block diagram of your system, the power rails, the cables, and any shielding that you are using. We need more information in order to try to tutor you on how to best connect things for EMC. Also, what are the requirements for "isolation" between your power domains? Is it okay if they share a common ground and just have separate power rails? Or do they need to be "floating" with respect to each other? If so, why?

And sorry if I missed it, but what kind of DC-DC converters are you using to make the power rails? Are they isolated or non-isolated topology converters? What frequencies are they running at?

EDIT -- Looking back over the thread, this may just be a school exercise and not a real system that you are designing? Or do you intend to build and test something?

 

[xortan]

I am not asking for any particular thing I am currently working on. It is not for a school exercise but for a real system. I am a new graduate and have been working in the industry now for a couple years. EMC is not covered in university and a lot of it seems like 'black magic' to me.

I work for a company that designs various robotic systems for pipe inspection, oil rigs, nuclear, etc. Basically design robotic vehicles to go where it's too dangerous for humans to go.

Due to the customers we deal with there has been no requirements for EMC but recently our company has decided we want to make sure our products could pass EMC due to all the other benefits that comes with it (clean power, signal integrity, etc.)

I am not sure how to attach images so I made a dropbox account. There is a vehicle which has a tether attached that carries the power, video, and communications. The tether will plug into a rackmount controller (this is what I usually design) which then connects to a PC to use our control software. The vehicle would be a whole separate EMC issue so I just want to focus on the top-end for now.

System overview: https://www.dropbox.com/s/32wnh1dvwhlw7yw/system_overview.jpg?dl=0

I have another image of what is typically inside of our controller.

From what I have seen around here the isolation needs to have separate ground and power rails. The amount of isolation doesn't really seem to matter, most things here are 1.5 kV of isolation. If we do not isolate the video circuitry power from the communication power (we use RS485) we will see comms noise on the video.

We will create the RS485 by using an USB to RS485 converter. We have seen problems when this does not have isolation. Depending on the PC the customer has it can create a lot of noise so now we just always isolate this. We will either create an USB to RS485 circuit or use an off-the-shelf converter, I usually pick the off-the-shelf stuff when I can so I would say 80% of our controller will use an EasySync or something similar.

As for the DC/DC converters we generally use isolated topology and the switching frequency varies. Depending on the converter they seem to range from 250 kHz - 750 kHz.

I have another dropbox link showing a block diagram of what is typically inside of a rackmount controller.

From that diagram we will have an AC/DC converter which will create 48 VDC and send that down to the vehicle, when it gets to the vehicle it gets stepped down again. The AC/DC converter will be an off-the-shelf power supply which will have isolation.

The 48 VDC will also get send to our video receiver circuit which will step it down to 5 VDC. This will be done using an isolated topology.

The USB to RS485 is powered from the USB port from the computer, also an isolated converter.

All these will have external cabling leaving the controller.

Controller overview: https://www.dropbox.com/s/32wnh1dvwhlw7yw/system_overview.jpg?dl=0

So when I got to the part of the EMC course where they tell me to ground signal common to the chassis, as well as AC Earth to the chassis, I was confused on how to apply that to my situation, maybe I can't.

My concerns for shielding are the tether attaching to this controller, the video cables, the usb cables and the AC cables.

If the pictures don't work or you need more info please let me know.

Thanks for the help!

 

[berkeman]

Very helpful reply, thanks.

It looks like I would need to create a DropBox account in order to access the pictures, so sorry I haven't viewed them. You can use the UPLOAD button in the Reply dialog to upload some of them instead.

Your questions are good. EMC isn't really a "black art" when you learn how to think about it. First, DC-connected RS-485 is not a good choice for comms in a noisy environment. If you are using the transformer-isolated version, then it's probably okay. Otherwise I'd encourage you to look at transformer-isolated alternatives for comms.

Next, one of the most important concepts in electrical isolation and noise isolation is "shared impedance". Let's say that you have two circuit blocks that share some ground impedance. That is, they have connected grounds that have some shared trace length back to the power supply. Then what you get is the transient currents in the ground of one of the systems causes transient voltages in the other system, because they share some impedance back to the power source. That's generally a bad thing. Instead, when you connect/share grounds, the connection should be in the form of a "star ground", where all of the ground connections are short and wide and right back to the power source.

A good example of this is a microcontroller (uC) and a radio chip/module on the same PCB. The uC communicates with the radio chip via some serial interface for communication, but you need to keep the digital noise of the uC out of the radio system in order to achieve the best receive sensitivity. So it would be a design error to run your power and ground connections from the power supply past the uC and to the radio chip. The uC's transient power and ground currents would induce noise voltages in the radio chip that would seriously impact its receive sensitivity. (Don't ask me how I know this...)

Instead, you use a star power distribution and grounding system in this case. You put the final voltage regulators in the middle between the uC and radio chip/module, so that their individual transient currents do not cause transient voltages in each other's power rails. Very important.

Also, keep in mind that even if you have galvanic isolation between circuit blocks via transformer-isolated DC-DC converters, you still have capacitive coupling between those blocks. Depending on the frequencies of interference involved, you may need to use some extra techniques (like shield layers in the transformers) to cut down on that noise coupling.

 

[xortan]

Thanks for the reply. I think I am starting to get it a bit more. I have uploaded the images to this post now. I have attached another image I found online of a single point grounding scheme, it is not from the EMC course I am taking. I think it shows pretty much what I am trying to do.

I can see how this grounding scheme would be better for EMC because there will be no shared impedance as you mentioned. Where the grounding pictures differs from what I typically do is we use an isolated power supply and, depending on the PCB, some of the circuits and logic will be isolated from each other.

If I require this isolation, I wouldn't want to use this type of grounding correct? Or, because there is capacitive coupling between the input and output of the converter anyways, does it even matter?

I am currently designing a PCB and want to apply some EMC concepts to it. This PCB doesn't have a bunch of isolated converters on it so I can implement a star grounding scheme and it makes sense to me. There is only one isolated DC/DC on it so I will just keep the traces on the same layer and not worry about a ground plane for it.

 

[Tom.G]

only one isolated DC/DC on it so I will just keep the traces on the same layer and not worry about a ground plane for it.

The need for a ground plane is determined mostly by frequency of the signals, how sensitive they are to electrical noise, and how electrically noisy the environment is. From video frequencies upward ground planes are the easiest way to cut down interference and signal reflections. By the time you reach maybe 25MHz you generally use either ground planes or wire your signals with coaxial cable.

Here is a link with about 100 pages of overview PCB design, Grounding, and Decoupling, although you probably already know much of it:
http://www.analog.com/library/analogDialogue/archives/43-09/EDch 12 pc issues.pdf