Digital PCB Design Cycle.

Written by Tom on Tuesday 12/12/06


Mission Technologies first project was an ARM development board. This
project was chosen to expose us to a complete PCB design cycle. A PCB design cycle has the following elements.


1. Decide on the Product.

2. Source parts.

3. Complete schematics.

4. Create CAD design rules.

5. Layout PCB.

6. Have PCB manufactured.

7. Get the PCB populated.

8. Debug PCB.

9. Put in a box.


1. Decide on the Product.

For us the decision to make a ARM dev board was easy. Ian had made a couple of ARM dev boards and was prepared to teach us all he knew. However we still needed to decide on functionality - memory size, I/O capabilities and more crucially the power supply requirements (for us we wanted everything - it should allow AC or DC power input from 5v up to 30v using a switched mode high-efficiency design).


2. Source parts.

This can be very difficult for small companies. The distributors are not interested in 10-20 piece orders. Lucky for us we worked for one of the biggest electronics companies in New Zealand.
Using this leverage we were able to obtain most of our parts as samples. We had to confirm we could order (and obtain) the parts before we did the schematics and started layout: not being able to source a crucial part would have killed our plans instantly!


3. Complete schematics.

We used Protel (and also sometimes Orcad) for our schematic capture, and found it performed OK. Various hardware decisions must be made at the point of schematic capture depending on what components must be interfaced. How wide will the busses be? How many chip selects will be used? Addressing schemes?
Connectors take up a lot of time too - deciding what type of connectors to use, what to put on them and trying to guess at future interfacing requirements. For a power connector, the possibility of getting it the wrong way around (an all-too-common mistake) has to be handled.
The most important aspect to stress at this step is the checking and re-checking of the schematics. An error here can kill the product later on and will only be found at step 8. We had at least 2 sets of 'fresh' eyes checking each part of the schematics, and had review meetings to discuss points found in the checking.


4. Create CAD design rules.
These rules include track min and max thickness and spacings. The sharpness of a track corner/bend, via size. The layer stack/thickness. Where to put ground plains. Make sure you talk to the PCB manufacturer to ensure that they can produce to the specification you select.
We all know about thicknesses - and how you mght need to go down to 5/1000 inch (5 thou) for a BGA (ball grid array) layout, and down to 6 or 7 thou for a TSSOP (such as used for flash memory). One big one is the hole size: getting vias between the balls in a BGA layout is absolutely crucial to the layout, and this needs a small hole.
Generally you need to find the separation between the edge of one ball and the next. Then take away the worst case spacing uncertaintly (currently 4 thou for a good PCB manufacturer), take away twice the minimum annular ring width (4 thou), and twice the minimum spacing (4 thou). What's left is the hole size.
The bad news is that this is a critical and defining value for most designs. The allowable hole size for a manufacturer also depends on the number of PCB layers (and their thickness). Around 10 thou is the best you can do with a drill (possibly down to 8 thou if you're lucky). Smaller than this and you need to use a laser to cut the holes which is - guess what - more
expensive!


5. Layout PCB.
This is probably the most time consuming. Laying out the components and tracking between them on multiple layers is a black art and I guess can only be learnt with practice.
Some things to avoid are long runs of parallel tracks and long runs of fast signals on the outer layers of a PCB (since these will create lots of electromagnetic interference - EMI). Overall track length is very important to fast edges. For modern designs you need to calculate a transition electrical length (TEL) and stick lower than this. As a rule of thumb you should keep below 45mm for very fast signals like SDRAM selects, clocks and data bus controls.
For more information on PCB layer, look here.

6. Have PCB manufactured.
Once the huge task of tracking components has been completed it is time to send off the PCB to the manufacture. Usually the PCB information will be in the form of a GERBER file format. This can be fed directly into the manufacturers tools and checked for errors. A couple of project were saved from disaster when the GERBERS where checked and the ground planes was found to be shorted to the +5 rail. You can get a free tool called GVPrevue from http://www.graphicode.com/ that will read and display GERBERS.
It is *always* recommended to view and check the final output using a different CAD/CAM tool to the one in which it is generated. Take it from us, this will occasionally save lots of heartache later.
If you can get the boards electrically tested, then when PCB fabrication process is not perfect there will be defects.


7. Get the PCB populated.
Two choices here by hand or automated. By hand is a lot of hard work and you might do it for the first PCB so it can be debugged. Otherwise paying someone else is the way to go it saves so much pain!
For some devices such as BGAs, it's really difficult to see if the soldering has worked properly. For this it is best to get an X-ray done (which should cost no more than a few 10s of dollars). Looking at the solder joints - or balls under a device - will indicate where there are dry joints or bad patches of soldering.


8. Debug PCB.
Depending on the complexity of the PCB this will take from a day to infinity to debug. Has there ever been a complex PCB produced that was not given mod wires later?? Ian claims to have debugged his first ARM board with a multi-meter watching each address line toggle, for us mere mortals a digital scope is a handy tool. Even better is a logic analyser.

9. Put in a box.
If you have made it through to here you would think it is time to celebrate. But no. Making a product is only 30% of the overall effort to selling a
product. 70% of you effort will go into getting a market share but that is another story.
A box is not only attractive, hides the mod wires, and looks more like a product, but it will protect your one and only working prototype from greasy fingers, electrostatic discharge and coffee spills.