"Connecting the Dots"

During component placement you were organizing parts into functional groups. During this phase you may have noticed that many of the connections are very short. Often there will be very short netlines, like two pins of a gate shorted together or two surface mount components sitting side by side, and the connection between them can be made on the surface layer with no via. Keep in mind that if a test fixture will be developed for the product, even though the connection can be made between them directly, you may need to add a via anyway to provide access to the short trace from the other side of the board (test fixtures are much cheaper if all of the circuit nodes are accessible from one side of the assembly, usually the bottom). At this point in the design, I usually can't resist doing some of the really short routes on the top and bottom layers, but before going on we should take a step back and have a look at the big picture...

If Component Placement is similar to a Jigsaw Puzzle,
Then Routing Traces is similar to Connect-The-Dots.

You may have spent quite a long time doing very specific component placements for each section, now let's zoom out and have a look at the whole design. By this point in the design process we know that the components will fit in the area provided by mechanical constraints, we can see our functional groups in a general floorplan, and if our CAD system supports it, we can turn on the remaining netlines and see the interconnections between groups. Unless this is a simple two-layer design, we can't really go much further until we define (or refine) the layer structure and plane assignments.

PCB Layer Stack-up

It is difficult to make specific recommendations without knowing the parameters of a design. Many RF and analog boards have relatively simple circuits on two-layer boards, but the feature sizes and material properties are critical. Digital designs typically have more complex circuits with many layers, and power boards may need heavy copper thicknesses, isolation requirements and large copper fills. Each of these will have specific design constraints. It is often beneficial to find a similar design by the same customer for comparison, and ask for details if it is outside the scope of your experience.

Don't let that last paragraph discourage you!
We can make a lot of progress here...

Multi-layer boards usually have an even number of layers, boards with an odd number of layers are very rare.
A diagonal route on a layer will block many other routing channels and will not use the layer efficiently. Layers are added in pairs, where one layer will have mostly horizontal routing, and the other will have mostly vertical routing (connected by vias).
If you are doing a multilayer design, the layer stackup should be symmetrical to avoid board warpage.

Very simple designs may be able to be routed successfully using only the top and bottom layers of the board. If you can complete the routing using only a 2-layer stackup (top and bottom, no internal layers), the board will be relatively inexpensive to fabricate compared to multi-layer designs.

Most modern designs will incorporate power and ground planes. There aren't too many decisions that have to be made for 4-layer designs, except that one routing layer will usually be declared for mostly horizontal traces, and the other layer will be routed with mostly vertical traces. The ground plane will most often placed in the stack to be nearest the side with the most active devices.

The next two layers to be added are usually trace layers to provide more routing capability. On 6-layer boards, it is usually best to put the most critical routing on inner layers, and the outside layers will be used by land patterns, via fanout traces, very short routes to connect adjacent components, and if needed, heavy power traces or small copper area fills for high-current or heat-sinking.

Beyond this brief introduction to PCB layer stack-ups, I would merely be duplicating the work of others who have already explained it far better than I would. For example, Henry Ott has written a very good introduction to PCB Stack-ups which can be found in the Tech Tips section of his website at HottConsultants.com. Look for a six-part article called PCB Stack-ups in Section 10. In case the link becomes unavailable, I have collected all six parts into a PDF file which you can read HERE, but Henry Ott deserves all the credit... READ IT, MEMORIZE IT, LIVE IT!

(and as another compliment to Henry, his new book "Electromagnetic Compatibility Engineering" is getting rave reviews and won the 2009 Publishers Association Prose Award for the field of Engineering and Technology. Good Job, Mr. Ott!)

A Field Guide to Guiding Fields...

If I ever write a book about Signal Integrity, I would like to call it "A Field Guide to Guiding Fields". Because that's what we are really doing; transporting an energy field from one place to another, hopefully along paths that we have carefully constructed, without stray energy causing unintended consequences or being diverted unintentionally.

EMC is an acronym for Electro-Magnetic Compatibility and also Electro-Magnetic Compliance (which is closely related).
Compatibility refers to the influence of a particular electronic circuit ON the environment around it (Emissions) and how well it performs when subjected to influences FROM the environment (Immunity).
Compliance usually refers to how well a particular electronic circuit meets minimum specifications or regulations.

The purpose of this section is to familiarize you with some of the topics related to Signal Integrity and EMC, but the proper application of the fundamentals will require a greater understanding than this text can provide. If you are here because you have a design where signal integrity or EMC characteristics are critical, SEEK HELP!

You may have heard the phrase "electricity follows the path of least resistance", which is not entirely accurate when evaluating modern electronic circuits. Impedance is a better term than resistance, because whether we are looking at 60Hz AC from the wall socket or high-frequency multi-GHz signals, impedance is opposition of current flow from the combined effects of resistance, capacitance and inductance.

If we were using perfect conductors in a vacuum, we could expect our signals to travel from one place to another at the speed of light with no loss. Unfortunately, we haven't devised a practical way to approximate that yet, so we have to study how our design will perform in the real world. We are using copper and other conductive metals which are not perfect conductors, we're separating them with different forms of insulating materials that have various and sometimes inconsistent properties, and we are usually forced to put different types of tranmission lines operating at different frequencies in close proximity to each other.
The signal loss we should expect is a combination of five general factors:

Since copper is not a perfect conductor, it will present a resistance to current flowing through it, and some of the signal strength will be lost in the form of heat
Some signal strength will be lost in the insulating materials surrounding the condustor (including air), since these have less permittivity than a vacuum. The property of materials called the "dielectric constant" is used to make calculations for this.
As current flows through a conductor, an energy field forms around it, and some of this energy will be radiated away (and can even couple into adjacent conductors)
As the characteristics of the transmission line change (like the transition from a connector-to-trace or trace-to-via) impedance discontinuities will cause reflections which will degrade the signal strength
Increasing the frequency will increase the severity of these effects

Signal Integrity is a general term for the methods used to understand and predict the behavior of a circuit in various configurations, and for the steps we take to control these variables to get the circuit performance we need.

Although some measures may have been taken by the design engineer to mitigate unwanted effects (like the addition of filtering and termination components), our focus here will be on the routing of conductors

Here are a few concepts you may need to learn more about, with very brief descriptions:

Layer Stack and Planes

TO BE CONTINUED...

Return Paths

TO BE CONTINUED...

Impedance Control

TO BE CONTINUED...

Differential Pairs

TO BE CONTINUED...

Skin Effect

As current flows through a conductor an electromagnetic field forms around it, and at higher frequencies more of the current will be flowing near the surface of the conductor than near the center. This is called the SKIN EFFECT. As the frequency increases, the skin depth decreases. When the skin depth is very thin, sometimes the roughness of the copper surface has to be taken into account because it can have an effect on the circuit performance. For some circuits, different materials may be needed with smaller copper crystal structure and a smoother surface (sometimes referred to as "less tooth").

What Now?

By this point in the design you have components placed on the board and you are preparing to route traces. If you take a moment to review everything you know about the design parameters (current requirements, voltage clearances, layer assignments, impedance control, etc.) it might seem like there are too many variables to manage at once. We're more than halfway through this tutorial, stressing the importance of many various topics, but we haven't even started routing traces yet!

I hope you aren't overwhelmed with Too Much Information and little actual progress, but important decisions have to be made early in the design cycle that will affect the cost, performance, and reliability of the final product forever. My advice to you is to take advantage of as many of the features of your CAD system as possible, to embed as many of these rules and guidelines into the design as possible, so the computer can do automatic error-checking.

Before you start routing traces, use whatever features your CAD system provides
to help you manage your design constraints.

Most CAD systems have some way of assigning net properties and then using them to make many tasks easier, such as maintaining appropriate distances between features, setting minimum trace widths, and selecting default via sizes. Most systems prevent you from making an error once the rule is in place, but also provide an automated Design-Rule Check (DRC) process to identify potential problems before the final data is generated.