2
\$\begingroup\$

When I jump into a new design, if it's intended for mass production, I try to think how cheap I can make it. Are there any rules of thumb I should follow? Here are a couple items that I THINK are worth aiming for:

  • Minimize component count
  • Minimize number of unique components (ex. reusing a 10k resistor vs. using a 11k)
  • Component availability
  • Connectors vs. No Connectors
  • Minimize number of components that need programming
  • Use smallest PCB size
  • Use the least number of layers
  • Limit PCB complexity (controlled impedance, buried vias, etc.)
  • Keep all components on the top only vs. on both sides
  • Keep trace width and spacing large
  • Limit complex component package types (lead-less vs. leaded)

Not sure if this list reeks of paranoia or if these are good metrics. I need an adult.

\$\endgroup\$

closed as primarily opinion-based by pipe, DoxyLover, laptop2d, Andy aka, Chris Stratton Sep 17 '16 at 0:54

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    \$\begingroup\$ I don't see much, if any, paranoia there (not sure why you would limit package types) and those seem like reasonable bullet points- however, of course, many are contradictory. For example, you can make the PCB extremely small by packing parts together, using tiny trace widths and spaces, and many layers. Testing is a huge consideration and needs to not be an afterthought. \$\endgroup\$ – Spehro Pefhany Sep 16 '16 at 18:04
  • 1
    \$\begingroup\$ Do you think there is a significant cost difference between doing a 2 layer vs. a 4 layer board if they were both the same size? \$\endgroup\$ – Craigfoo Sep 16 '16 at 18:10
  • 1
    \$\begingroup\$ Yes, certainly there is a cost difference- significant in most cases. The process for 4-layer is slower and more expensive (your boards have to be laminated mid-process in a big press). The performance may be better with the 4-layer (ground and power planes), it may be possible to make the board smaller, and some packages (BGA etc.) you actually need more than 2 layers (maybe 6) simply to use the part. \$\endgroup\$ – Spehro Pefhany Sep 16 '16 at 18:18
  • 1
    \$\begingroup\$ Minimize component cost, thats how you do it. Boards are cheap especially with mass production at cents per sq inch with high quantity. If you want to find out how to make things cheap, rip apart some consumer products and see how the Chinese do it, they are the masters of cheap. \$\endgroup\$ – laptop2d Sep 16 '16 at 19:52
  • 2
    \$\begingroup\$ Nothing compares to designing something with reliable performance. Not doing so costs a bomb. Just think about product recalls. VtC \$\endgroup\$ – Andy aka Sep 16 '16 at 21:30
2
\$\begingroup\$

Improve s/n ratio. Noise reduction.

  1. Most important - avoid or mitigate ground loops from becoming a microwave antennae
  2. Or any loop from becoming a microwave antennae
  3. Reduce line/track impedance (especially inductance) first or otherwise then mitigate its effects
  4. Appropriate line termination to debounce oscillations at start and end of square waves.
  5. Test for propagation of noise from power supply to circuitry
  6. Ensure sufficient power injection points at vicinity of ICs to avoid signal flaccidity due to flaccid response of power supply input.
  7. Match or exceed sinking requirements from sourcing currents.
  8. The lower the power used, the better

Testability and debugability

  1. All active components on topside, most impedance mitigation components at the bottom, so that it is easy to desolder and resolder those passive components during debugging.
  2. Well-defined functional pattern based logic design. Such that the state of the application board can be easily deciphered from signal combinations.
  3. So that either a monitor chip is embedded onboard giving status on LED, or there are sufficient appropriate test sockets to probe signal patterns.

Flexibility

  1. Use either configurability or programmability means to allow board to be used differently.
  2. Use programmability to reduce component count.
    • Programmable parts are not that expensive.
    • Programmable parts already come with functional pattern specs you can test against, so that you can base your functional patterns on the programmable part.
    • Why not acquire expertise in embedded programmable ICs?
  3. Design/embed impedances into the circuitry. (I don't know much about this and don't know how viable it is.)
\$\endgroup\$
  • \$\begingroup\$ "a microwave antennae" - sorry for the singular-plural grammatical conflict. \$\endgroup\$ – Cynthia Avishegnath Sep 17 '16 at 0:49
0
\$\begingroup\$

Dave Jones of the EEVblog actually did a few video blog segments on this exact topic:

https://www.youtube.com/watch?v=VXE_dh38HjU

https://www.youtube.com/watch?v=Uemr8xaxcw0

https://www.youtube.com/watch?v=2zGisPMNstI

In each of those videos, Dave offers some very helpful pointers and insight into the DFM (Design for Manufacture) process. You may find them incredibly useful, as I have.

\$\endgroup\$

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