3
\$\begingroup\$

I'm working on a PCB that allows me to attach a TSSOP IO expander to a breadboard more easily for experimenting. I asked a question regarding the configuration of decoupling capacitors for an IC with 2 supply pins.

One of the recommendations was to add ground pours to the top and bottom layers of the PCB to create low impedance ground connections for the decoupling capacitors and to tie the top and bottom layers with several stitching vias.

How do I know how many stitching vias to place and how do I know where to place them?

Here is what the board looks like so far: enter image description here


Edit:

Because I'm new to electrical engineering (I'm barely at hobbyist level), it really helps me to see solutions visually. For folks finding this in the future, here's what my latest state is:

enter image description here

\$\endgroup\$

3 Answers 3

5
\$\begingroup\$

If you flip C2 and C3 around this could be a 1-layer board. Flipping them, would also reduce parasitic inductances due to the required vias.

But, to answer your question, for your (low-frequency) application, stitching vias serve just one purpose, to reduce the impedance for any current traveling on the planes. That directly implies that areas of the plane that have little or no current, due to being far from the current paths, don’t require any vias.

In your case you only have 4 pads and a pin that conduct current from the plane. You just require vias near those pads and perhaps the pin (vias are much less conductive than the pin itself). Perhaps 4 vias on the capacitors, and 2 near the IC Gnd.

Any additional vias would mostly be cosmetic, it would be hard to tease apart the effects due to reduction in impedance due to the vias from the increase in impedance due to the added holes in the planes.

If this was a high-frequency a application (e.g., >500MHz) vias would be required near the PCB edges, to avoid unintended emissions and you would need to take into account the impedances to the plane underneath signal lines.

\$\endgroup\$
7
  • \$\begingroup\$ If I flip the caps, the Vss trace will still be short enough? And the ground pour on top will be low enough inductance that I don’t need the pour on bottom at all? \$\endgroup\$
    – D. Patrick
    Commented Jan 5, 2019 at 22:54
  • \$\begingroup\$ I meant Vdd trace won’t be too long. \$\endgroup\$
    – D. Patrick
    Commented Jan 5, 2019 at 23:02
  • 1
    \$\begingroup\$ @D.Patrick If you flip C2 and C3 VddI2C could actually become shorter. If you are doing a double-layer board (the difference in price might be negligible) put the plane anyway. It does not hurt anything and it will provide better shielding, a negligible effect given the application, but better anyway. \$\endgroup\$ Commented Jan 6, 2019 at 0:15
  • \$\begingroup\$ I added an image to the end of my question that shows my attempt to implement your suggestions. Did I more or less capture what you had in mind? (I also moved a few things around because I wanted to label a some of the pins, but I don't think I changed the circuit substantially). Thanks again for your help!! \$\endgroup\$
    – D. Patrick
    Commented Jan 6, 2019 at 2:29
  • \$\begingroup\$ I hate to be a pest; I very much appreciate your help. I plan to accept your answer, but first I’d like to make sure the image I added to my question accurately represents your recommendation. When you have a second, would you mind terribly taking a quick look? I just want to make sure the question is helpful to other folks new to this like me who might understand more easily with a graphic representation. Thank you again!! \$\endgroup\$
    – D. Patrick
    Commented Jan 6, 2019 at 22:36
2
\$\begingroup\$

Put 4 at each end, and a couple under the body of the chip. I don't know that there is an exact science behind it. I try to get Gnd vias near the Gnd end of the caps too. You could move C3 & C4 up a little to get Gnd vias between them and C2 and C1.

\$\endgroup\$
0
-1
\$\begingroup\$

Just make your top gnd and the bottom vss; this way, you don't need any Vias.

Also, why those weird trace directions at the outer traces of the chip (pin 1 and 2 on each header)?

Also, you don't need a via in j1_12 (aka gnd), as pads go through the board and manufactured boards (boards you don't make yourself) have pads connected through the hole.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for your answer!! What’s weird? I know so little about PCB design that I can’t tell what’s weird and what’s not. :) \$\endgroup\$
    – D. Patrick
    Commented Jan 6, 2019 at 18:29
  • 3
    \$\begingroup\$ An answer is not the place to ask the OP for clarifications ("why those ... trace directions"). Perhaps better to obtain clarifications through comments on the question itself, then post an answer when you have a clear answer to provide. \$\endgroup\$ Commented Jan 6, 2019 at 20:59
  • \$\begingroup\$ The funny traces in the bottom right were an effort to keep the traces from having 90 degree angles and trapping acid during production (or something like that). In the top left, it's because I can't trace pin 1 to j1 between the caps because of the ground pour and I was trying to avoid adding a trace to the bottom of the board so I kind of had to sneak it between the jumpers. Does that make sense? \$\endgroup\$
    – D. Patrick
    Commented Jan 6, 2019 at 22:28

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.