7
\$\begingroup\$

I was surprised by the number of connections to ground some pc-related connections have. For example, the parallel port has 8:

enter image description here

Also, the ATX power supply connector has 9 ground connections:

enter image description here

Is there any reason for such a large number of ground connections?

\$\endgroup\$
10
\$\begingroup\$

For a power connector: each pin can carry a limited amount of current. All current must return via ground. There are multiple power pins per voltage, hence multiple ground pins are needed to carry all the ground current.

For a data connector: unused pins might as well be connected to ground, and when a ribbon cable is connected, subsequent wires are alternately connected to the top and ground row, hence the ground pins end up connected to wires that sit between signal wires, which add some extra shielding between the signals.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ +! for comment about interleaved ground and signal lines. \$\endgroup\$ – Russell McMahon Jun 7 '15 at 1:04
  • 1
    \$\begingroup\$ It can be very instructive to study how the 40-pin connectors used for the older IDE type hard disk drives changed from being a 50 mil center to center ribbon cable in the earlier days to one with the same connector ends but the cable itself doubled in number of wires with every other one being GND. That design requirement came about to accommodate faster IDE interfaces but needed the alternate sig-gnd-sig-gnd for maintaining signal integrity of the bus. \$\endgroup\$ – Michael Karas Jun 7 '15 at 4:22
2
\$\begingroup\$

For the ATX power supply cable there needs to be enough GND wires to return current from all the other voltage rails back to the power supply. So ideally the number of GND wires would match the size and number of the voltage rail wires.

For the parallel port you need multiple GNDs so that the return current path for the signal lines is reasonably balanced against the number of signal lines. Ideally the number of GND wires would match the number of signal wires and each would work as a twisted pair in the cable.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ In the case of the parallel port, wouldn't a bigger (thicker) GND wire work the same (even if this means that the circuits should also have thicker paths for GND). \$\endgroup\$ – Paul92 Jun 6 '15 at 20:35
  • 1
    \$\begingroup\$ Nope - The concern here is not voltage drop, but rather reflections of the sharp edges on the data waveforms causing errors in sensing the data. As Michael Karas said, the best approach would be a ground for every data line (plus the strobe line). Parallel printer cables are derived from the Centronics cable spec, which used 36 pins (see lammertbies.nl/comm/cable/parallel.html). Going to 25 pins meant that some of the grounds had to be eliminated. \$\endgroup\$ – WhatRoughBeast Jun 6 '15 at 20:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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