Group Actions
Engagement
  • Author Author: michaelwylie
  • Date Created: Date Created
  • Views 2677 views
  • Likes 4 likes
  • Comments 31 comments
Related
Recommended

In The Air: Episode 5 - PCB Design

michaelwylie

Update

A Riddle: How many times in three weeks can a 1 year old child catch sick? The correct answer seems to be 3 times ...

 

Previous Posts

In The Air: Epidose 1: Introduction

In The Air: Episode 2 - Preparing for Surface Mount Work

In The Air: Episode 3 - Surface Mount Beginnings

In The Air: Episode 4 - Inductors

 

Introduction

I have finished my Booster Pack design. Figure 1 shows the top and bottom layers. Note, I am an Altium user so the display of the board layers may be unfamiliar to you if you are an Eagle user. If you are familiar with TI's Booster Packs, you'll recognize the characteristic 20 pin headers on opposing sides of the board. In my last post I discussed choosing inductors for filtering/decoupling grounds and power supplies. Accordingly, we will discuss the layout of those inductors and separating grounds for quiet sensor performance in this post.

 

Figure 1: Bottom (left) and top (right) layers for my Booster Pack design.image

 

Connecting Grounds

The general approach I have adopted is shown in Figure 3. The signal ground is like an island surrounded by the power ground, and the only way to get to the signal ground is through an inductor. The connector is on the boundary between the two grounds because the sensor requires both power and signal ground. I've created an annotated version of the board layout of Figure 1 and produced it in Figure 4. I have labelled the sensor interface, signal ground, and power ground. The power ground is directly connected to the ground of the C3200 through the Booster Pack header connections, and the signal ground is connected to the power ground through the aforementioned inductor.

 

image

Figure 3: Basic idea of the Particle Counter Booster Pack.

 

image

Figure 4: Annotated Figure 1 showing the sensor interface, signal ground, and power ground.

 

In Figure 5 I've zoomed in on the board so the signal ground is more visible, and the ground connecting inductor can be seen in the lower left (L2). You can't see the pads from the component because I used Seeedstudio Fusion for my PCBs, and Seeed Studio's design rules don't use a thermal relief pad connection by default. I'm not sure why, but that's the way it loaded. I didn't personally care for this project because I will be hand building the boards, so I left it alone. If I had loaded the rules before designing the board I may have changed it, but I loaded the design rules after. Disclaimer: especially in mass runs, use thermal reliefs on your ground connections to components as it will prevent tombstoning components during reflow.

 

image

Figure 5: Zoomed view of the top layer.

 

The 3D render of the board is shown in Figure 6. You can clearly see the pads for the inductor L2 in the lower left. I didn't bother using models for any components unless they were supplied by default, because I have no concern for height tolerances with this board. At this point in the design you may be wondering, why did he only use a single inductor, and not tie the grounds together at multiple points? The short answer is you'd be creating a ground loop, which can be a terrible problem to troubleshoot. You can look up the analysis for ground loops on PCBs, but if you only connect the grounds at one point you won't have to worry about it. You may have other issues, but a ground loop will not be one of them.

 

image

Figure 6: 3D Render of Particle Counter Booster Pack.

 

Power Supply Decoupling

In my last blog I spoke of decoupling power supplies as well as grounds. I've reproduced the schematic showing these decouplings in Figure 7.

You can see L2 is connecting the grounds, but what about L1; how is that implemented? In the same way. The inductor pads must straddle the gap between the power and signal ground. If you look back at Figure 5, you can clearly see L1 near the upper left straddling the ground boundary.

image

Figure 7: Decoupling inductors

 

Still Confused?

How about a video explaining it then!

 

 

Final Remarks

I ordered my board. It was only $10 USD from Seeed Studio for 10 pieces. No one else could compete with that price, so I had to choose them. I've got my bill of materials ready to go, I'm just waiting to hear back about the possibility of samples from TI. Every IC on this board is made by TI, which was a bit of a challenge for me. Hopefully everything can arrive in time to put it together and write some software.

Parents

  • Nice board.

    especially in mass runs, use thermal reliefs on your ground connections to components as it will prevent tombstoning components during reflow.

    For those of use not used to smd design can you eloborate on the above a little please.

     

    Thanks

    Mark

  • in reply to mcb1

    Mark;

     

    Great point. Solder paste is like magic, in the sense that it can fix your mistakes. So, if you happen to slightly misplace a component, the surface tension of the liquifying solder paste will pull the component into alignment. It doesn't always work, but in most instances it does. The major problem with solder paste comes from this question, "What happens if one pad's solder paste liquifies or "wets" before the other", with regard to a two pad component (resistor, inductor, capacitor). What happens is called tombstoning and looks like this:

     

    image

     

    This image is courtesy of www.pcb007.com. The surface tension of the liquid side basically torques one end of the component into the air and they look like small tombstones if the component goes completely vertical (hence tombstoning). The reason this happened is because one pad heated faster than the other. In most instances I have seen, one pad is directly connected to the entire ground plane and the other is maybe connected to a component pad or two. It takes much more time to heat the pad connected to the ground, so the component end that should be connected to ground lifts into the air. To alleviate this, you try to thermally insulate the ground pad from the ground plane so that the pad heats faster. Here is an example:

    image

     

    This type of connection with the four smaller traces is called a thermal relief, and it allows the pad to heat without waiting for the entire ground plane to warm. There are tradeoffs for using these reliefs though, the most obvious being the increased resistance. The designer has to make a decision for their design. If you're connecting to an analog or signal ground, maybe you want a solid connection with less resistance. This way there will be less chance of ground lifting. If you can afford the space go with a heavier inductor and solid connections. You can see this can get really involved for a seemingly simple choice. For the Wurth inductor I chose, it's pretty small and I think it's very light as well, so there is a chance for tombstoning. But, if a component tombstones, I can just push it back down during the heating process. If I were doing 100,000 boards, I probably don't want to push down multiple components per board.

     

    Hope this answers the question.

Comment
  • in reply to mcb1

    Mark;

     

    Great point. Solder paste is like magic, in the sense that it can fix your mistakes. So, if you happen to slightly misplace a component, the surface tension of the liquifying solder paste will pull the component into alignment. It doesn't always work, but in most instances it does. The major problem with solder paste comes from this question, "What happens if one pad's solder paste liquifies or "wets" before the other", with regard to a two pad component (resistor, inductor, capacitor). What happens is called tombstoning and looks like this:

     

    image

     

    This image is courtesy of www.pcb007.com. The surface tension of the liquid side basically torques one end of the component into the air and they look like small tombstones if the component goes completely vertical (hence tombstoning). The reason this happened is because one pad heated faster than the other. In most instances I have seen, one pad is directly connected to the entire ground plane and the other is maybe connected to a component pad or two. It takes much more time to heat the pad connected to the ground, so the component end that should be connected to ground lifts into the air. To alleviate this, you try to thermally insulate the ground pad from the ground plane so that the pad heats faster. Here is an example:

    image

     

    This type of connection with the four smaller traces is called a thermal relief, and it allows the pad to heat without waiting for the entire ground plane to warm. There are tradeoffs for using these reliefs though, the most obvious being the increased resistance. The designer has to make a decision for their design. If you're connecting to an analog or signal ground, maybe you want a solid connection with less resistance. This way there will be less chance of ground lifting. If you can afford the space go with a heavier inductor and solid connections. You can see this can get really involved for a seemingly simple choice. For the Wurth inductor I chose, it's pretty small and I think it's very light as well, so there is a chance for tombstoning. But, if a component tombstones, I can just push it back down during the heating process. If I were doing 100,000 boards, I probably don't want to push down multiple components per board.

     

    Hope this answers the question.

Children