Currently I am designing a 4-layer PCB, with the layers as follows:

  • (1st layer, top) signal
  • (2nd layer) ground
  • (3rd layer) power
  • (4th layer, bottom) signal

The clock frequency is either 8 MHz (running at 3.3V) or 16 MHz (running at 5V). There are no active RF components on the board, just surface-mounted SPI-based devices. The total board dimensions are ~4cm x ~3cm.

My current design has no copper pour on the signal layers, but have copper pour on the ground and power layers.

So under these circumstances, would it make any difference whether or not I have a copper pour on the top and bottom signal layers?

  • \$\begingroup\$ Will this board have a data or address bus where flight-time and capacitive loading would be an issue? How complex is this board? Mixed signal types? \$\endgroup\$ – user105652 Mar 31 '18 at 0:43
  • \$\begingroup\$ There is no address bus; the individual SPI devices are activated via digital outputs from a microcontroller. There are data transmitted from the SPI devices, so perhaps that can be considered a data bus. The board is purely digital, no mixed signal types. \$\endgroup\$ – plu Mar 31 '18 at 0:47
  • 1
    \$\begingroup\$ Remember the copper is already there. You pay the board house to etch away what is not needed. Extra ground pour is ok but 2 layers of it means a lot of vias. \$\endgroup\$ – user105652 Mar 31 '18 at 1:31
  • 2
    \$\begingroup\$ @analogsystemsrf And no, I do not need to pass FCC tests, the design is more for research purposes. How would edge speeds and trace lengths influence whether or not copper pours are required? \$\endgroup\$ – plu Mar 31 '18 at 2:58
  • 1
    \$\begingroup\$ If you have very fast edges, the copper pours will tend to become antennas because the traces will couple to the same-layer pours. Read what Neal_UK says below. \$\endgroup\$ – analogsystemsrf Apr 1 '18 at 2:14

All copper on the board should be consciously designed.

I am against copper pours on principle, because they seem to cause more problems than they solve - increased stray capacitance, confusion about where return currents are flowing, unintentional coupling between signals, or creation of an antenna. It can complicate debug and modification as well.

They seem to lead to a lazy mindset, where the designer thinks 'well there's lots of copper on the board, surely the grounding will be OK now!' A copper pour is unnecessary, and insufficient to achieve a good board.

There are good reasons to use a copper pour, thermal conductivity, reducing the amount of etch solution used, improving thickness tolerance, reduction of warping, beefing up of tracks, but 'should' isn't one of them.

  • \$\begingroup\$ All agreed. Just to add one consideration: with extra copper parasitic capacitance may couple between remote and unrelated circuits, like switching power supply and op amp feedback on other side of the board. \$\endgroup\$ – Gregory Kornblum Mar 31 '18 at 6:17
  • \$\begingroup\$ Thanks, in that case, I will keep the top and bottom signal layers of the design without any copper pour. \$\endgroup\$ – plu Apr 1 '18 at 13:57

4Mhz can be done single-sided - Sega did a Z-80 based home computer on single-sided FR2. 16Mhz is radically more especially when the traces are only 5cm or so long

With only 5cm traces, you can probably do the same for your design, (although double sided makes routing easier, and more layers will give you a ground plane if you really need it.

If you put both the power pours close together they act as another power supply de-coupling capacitor.

Surface pours can be useful for heat-sinking if you have some power parts could benefit from that.

Unless you're making this in production quantities the board house only charges for the area of the PCB material and the number of layers, and you don't get any credit for the copper they reclaim. (although they're probably glad to have it back)

  • \$\begingroup\$ Interesting - didn't know about sega doing single-sided boards, but it looks like they did 3 different designs: the SG-1000, the SG-1000 II and the Sega Mark III are all on single-sided boards, and seems didn't need all that many links to be applied, either. Incidentally, they didn't run at 4MHz - the CPU ran at about 3.6MHz, but the graphics processor ran at 10.7MHz and its dedicated RAM was connected via a bus that could switch twice (i.e. to provide RAS and CAS commands) in 3 processor cycles, so was effectively ~7MHz. But then, 7MHz can be done on breadboard (google: grant searle CPM). \$\endgroup\$ – Jules Mar 31 '18 at 18:18

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.