9

I am reading the application note from TI about the LM3409 evaluation board. In the board layout (Figure 3) the bottom layer is a single GND pour.

But top layer has also some copper pours which end up being connected to ground, such as the ones at LED-, C5, D1 and C1.

What I don't understand is: why are they all not connected to each other on the top layer, since they are all the same net ?

Top layer of the demo board PCB

feralgeometry
  • 303
  • 1
  • 6
  • 13
  • 1
    The pour *around* LED- is connected to Vo (follow it down, it connects right to Vo coming out of L1), not to GND. – mbrig Nov 29 '18 at 16:24
  • Can you hilight where you see two ground pours close to each other but not connected? – The Photon Nov 29 '18 at 17:01

3 Answers3

13

why are they all not connected to each other on the top layer, since they are all the same net ?

Because this is a switching converter. In switching converters very high currents can flow, often these or only very short current pulses. If we "just" connect everything to the ground plane directly it is unclear where these currents actually flow. Yes all is connected but due to resistance of the copper (low but it is always there) and inductance (a wire of 1mm has 1nH inductance, also small but still there) these current peaks can still influence the operation of the circuit.

For example the grounding of the chip must be such that no voltage (or as little as possible) is induced across it otherwise the chip will not have a "clean" ground which will prevent it from regulating the current properly.

So most often a "star ground" scheme is used, read more on that here.

I'm not saying the ground scheme used on this PCB is a star ground but it will be a deliberate choice of the designers to make it like it is. Often the IC's datasheet will also include a recommended PCB layout and that might include a grounding scheme as well.

Bimpelrekkie
  • 80,139
  • 2
  • 93
  • 183
4

This looks like a power supply circuit.

The reason for that is a switching supply generates high spikes / transients, and the via (and also copper) acts like a small resistor. To have separate ground on the top layer reduce the impact of high transients into the ground plane by keeping the current where we want it on the top plane, and avoid going to some places we don't want it to flow.

Basically, it allows to better control of where the current is flowing on the board. If it was a plain plane on the top, it would be difficult to predict where the current would flow, it may flow near sensitive IC causing unwanted effects.

Damien
  • 7,827
  • 1
  • 12
  • 29
2

Are you sure the pour at LED- is ground? Since it's a through-hole component I would expect the GND pins to connect to the bottom layer ground pour, and the top layer pour is probably a power net. Therefore connecting the other top pours to each other may not be as practical as you might think.

I expect this was done to prevent ground loops (a strip of pour on the top layer connected by vias on either end to the bottom ground pour). Keeping point-connections to the primary ground plane/pour is a better practice for PCB routing.

DerStrom8
  • 21,042
  • 8
  • 61
  • 95
  • 1
    LED- goes to ground directly, or at least that's what it looks like from the circuit schematic in the same document. – feralgeometry Nov 29 '18 at 12:11
  • "Keeping point-connections to the primary ground plane/pour is a better practice for PCB routing." Why ? – feralgeometry Nov 29 '18 at 12:12
  • To prevent ground loops, like I said – DerStrom8 Nov 29 '18 at 12:17
  • 1
    And I know LED- is connected to ground, I'm talking about the PCB design - I don't think the pour around LED- is GND. – DerStrom8 Nov 29 '18 at 12:18
  • 1
    LED- and GND are indeed connected if you check the circuit diagram in the linked appnote. – JRE Nov 29 '18 at 12:53
  • 2
    @JRE I will say again, I'm not talking about the electrical connection of LED- and GND in the schematic, I am strictly talking about the ***polygon pour*** around LED- on the ***PCB layout***. How do we know that the pour on the top layer surrounding LED- is GND, and not, say, VO? In fact, I think the ***polygon pour*** around LED- in the ***PCB layout*** is actually VO, since it connects to the VO pad of C5 – DerStrom8 Nov 29 '18 at 13:23
  • 1
    This is the only correct answer on the page. The pour is the top right is clearly connected to Vo. The ground pours on the top don't have an easy way to connect to each other, so there's no point forcing one when you can just stitch them to the back. – mbrig Nov 29 '18 at 16:20
  • @mbrig (and the OP) you are actually right about the top right pour being connected to VO. However I find it really confusing because electrically there is no connection, and that GND turret installed in the middle of the pour only makes the confusion worse. – feralgeometry Nov 29 '18 at 17:11
  • @mbrig Also routing-wise there *is* an easy way to connect all the grounds together in the right-hand side of the top layer, if one wished to: by making the top-right pour smaller (ending around the last pins of the LED connector), and then creating in its place a GND pour which connects the GND turret, the top part of C1 and the top part of C5. – feralgeometry Nov 29 '18 at 17:13
  • 1
    @UgoRiboni What do you mean "electrically there is no connection"? Connection from what to what? The ground connections are made through the vias and plated through holes to the bottom layer pour. The VO connection is made through the trace connecting to R1. I suspect the ground turret was placed there for symmetry reasons. Since the entire bottom layer is a ground pour, this turret could have been placed anywhere, so they put it in a place that looked neat and was easy to access (with relation to the VO/LED+ pin). – DerStrom8 Nov 29 '18 at 18:19
  • 1
    @UgoRiboni To your second comment I say that adding a thin snaking ground trace between pours is a bad practice. A thin snaking trace is an inductor, and adding inductance between two ground pours is a very bad idea - this can lead to excessive ground bounce, noise, and different ground levels. Not to mention the creation of a ground loop which I discussed before. Therefore, keeping one solid ground pour on the bottom layer is highly preferable due to its minimal inductance. – DerStrom8 Nov 29 '18 at 18:21
  • @DerStrom8 with a bit of fiddling you could make the pour roughly as thick as it starts around C5 (you'd have the move the LED- PTH if you wanted to keep everything nice and square) but I still agree with you. – mbrig Nov 29 '18 at 18:43
  • @mbrig Oh sure, it could definitely be done, but it still won't be efficient, as you seem to recognize. It will still have a higher inductance than the ground pour, and once again, you'll still have a ground loop – DerStrom8 Nov 29 '18 at 18:53