I'm designing a system that provides power/ethernet/serial for up to 5 daisychained units. All of the boxes will be metal and my initial plan is to run power and signal ground through the 14 pin cables that will connect each unit. Each cable's shield will be tied to chassis ground and the plan is to tie chassis and power ground together in the PSU/COMM box.

My question is A) is this a sound idea and B) should I connect an RC connection (1Meg/100pF or something) from my PCB mounting holes to the chassis ground in each box? Pros/Cons?

The boxes will draw between 5 and 10 amps at 48V. We'll have 100 Mbit ethernet and also rs232.

enter image description here

  • 1
    \$\begingroup\$ You show earth grounds on each 'box' do these connect to earth ground or do the grounds return only on the cable? \$\endgroup\$
    – Voltage Spike
    Feb 8, 2017 at 17:49
  • \$\begingroup\$ Sorry, that was just to show that each box was tied to the same potential locally. The grounds only return on the cable (actually, in some cases two boxes may be physically touching, but generally the cables between will be the chassis ground connection) \$\endgroup\$
    – scld
    Feb 8, 2017 at 17:54
  • \$\begingroup\$ You never want return currents on chassis, i don't at least. \$\endgroup\$
    – user76844
    Feb 8, 2017 at 17:57
  • \$\begingroup\$ Agreed, my thought was that the RC connection to chassis ground would keep the power and signal returns in my cable but allow transient/static to bleed out to the chassis \$\endgroup\$
    – scld
    Feb 8, 2017 at 17:58
  • \$\begingroup\$ Transients are not bleeding, you should use capacitors for that. Static is a completely different story. If you mean that you don't want voltage to build up between isolated grounds- you can use either 10M resistor or something like gas discharge tube. If you mean that you need to protect the system from ESD- you should analyze what exactly happens during ESD discharge in different frequencies, and this is very much system dependent stuff. \$\endgroup\$
    – user76844
    Feb 8, 2017 at 18:02

2 Answers 2


There are many options to deal with chassis and GND. Here is what i do:

  1. Power is always supported by two wires (power and return).
  2. Local ground may be connected to power return or isolated.
  3. If local ground is connected to the power return, it can't be connected to any other ground, because in such case current has more paths to return.
  4. Chassis is connected to zero potential (the one that is safe to touch) in only one point in the system.
  5. All digital and analog grounds (in contrast to power ground) are bypassed to chassis with 1nF and 1uF capacitors. If the GND is also considered zero potential, capacitor's voltage rating may be 50V.
  6. Cable shields are connected with low impedance to chassis. With RJ45 it's easy, in DType or other connectors you have to make sure impedance is actually low- don't connect shield to wire to case, but rather find a place to make a wide contact.
  7. Use filters on signals. Single-ended signals with single ended filters, differential- with common mode filters.
  8. Power and GND should always be bypassed to chassis (1nF and 1uF), unless it's AC voltage input (then just 1nF on each line).
  9. Good luck :)
  • \$\begingroup\$ Just stumbled upon this question and answers. Does 1uF cap need to be a non polarized capacitor? \$\endgroup\$
    – user1999
    Oct 31, 2019 at 11:41
  • \$\begingroup\$ Yes. And some high resistance in parallel to not let charge build up on the capacitor. \$\endgroup\$
    – user76844
    Oct 31, 2019 at 19:39

With this type of grounding scheme and the currents involved there will be problems that need to be taken care of, there are two schemes you could use for grounding, series and parallel:

enter image description here

In the series example circuit 3 would have this voltage potential on the ground from the return current of each device and the impedance of the cable (the point after Z3 on circuit 3's ground):

$$ V_{Circuit3} = (I_1+I_2+I_3)Z_1+(I_2+I_3)Z_2+(I_3)Z_3$$

$$ V_{Circuit2} = (I_1+I_2+I_3)Z_1+(I_2+I_3)Z_2$$

and so on...

With amps being used in the described system, and the Z term being in the ohms range the voltage seen by the last device on the ground could potentially be in the tens of volts range. I don't think this will be acceptable. A parallel scheme will probably be better.

Another issue will be isolation and common mode noise on cables between boxes. These issues take more space than I can use to describe here and are design dependent. I would suggest getting Electromagnetic Compatibility Engineering by Henry Ott and reading chapters 2 and 3. This might help also.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.