PCB Design Tips

[ARico08]

Hello all,

So I'm currently working towards my EE Degree and my classes are getting more into having me create my own PCB designs and products, however they haven't given us too much information on what kinds of things we should and shouldn't do when creating a PCB so I wanted to pick the collective brains of this forum to see if anyone had any tips.  I just want to know best practices and things to avoid or ways to maximize the performance of any components.  I'm mostly working with logic level stuff right now but I'd like to work with mains power (120VAC for us in the States) soon as well.

Also, what is everyone's favorite PCB creation program? They have us using Cadence at my college which has a lot of features but is anything but user friendly.

Thanks in advance for any help!

[microbug]

Perhaps this might be useful.

KiCad is my tool of choice.

[Wirehead]

I just want to know best practices and things to avoid or ways to maximize the performance of any components. 

Read the datasheets front to back, back to front, and then some more Sometimes, there are small but meaningful statements buried deep in the datasheets (keep ground planes away here, add an additional cap here for improved noise spec there, etc... you get the point )

It's common sense really... You'll make mistakes, you'll learn from them.  With cheap PCB services, it's MUCH more accessible to spin your own boards now, so experiment!

[ARico08]

I saw Dave's video on PCB Design for Manufacturing but must have missed the one about the PSU.  However I was hoping to be able to engage in conversation with others who have experience in this field, so I could learn more about why certain things are done and ways around common issues.  I've only created a few boards and they have all been very simple designs but I have my Ham radio license and want to start trying to build basic transmitters and receivers and I know that RF devices are much more sensitive to PCB board layouts.

[KE5FX]

I saw Dave's video on PCB Design for Manufacturing but must have missed the one about the PSU.  However I was hoping to be able to engage in conversation with others who have experience in this field, so I could learn more about why certain things are done and ways around common issues.  I've only created a few boards and they have all been very simple designs but I have my Ham radio license and want to start trying to build basic transmitters and receivers and I know that RF devices are much more sensitive to PCB board layouts.

Short answer is, you may very well not want to build PCBs at all for this application.  Get this book.  No, really, get it.

[Belgarion]

If you want to mount the board somewhere, remember to add mounting holes..
Too many of my boards do not have mounting holes because I either forgot them or did not think that I would need them, but in the end would have needed them.

Double check your footprints! It's really annoying to realize that the footprint on the board is mirrored.

Double check that you have connected the reset pins according to the datasheet/user manual.

Make sure that connectors and such are in the correct orientation.
I made a mistake with a microsd slot once, where the opening was in the opposite direction from what I intended, which meant that I had to put in the microsd card before soldering some of the components.

[AndyC_772]

Cadence tools are an industry standard, and very well worth taking the time to learn and become proficient in.

Having Allegro on your CV will look good and be useful... free tools, not so much.

[ARico08]

Thanks everyone for the responses.  I'll definitely get my hands on that book, even if I give up on building my own equipment it sounds like a good read.  And the mounting holes is a great point, I mounted my last PCB I created for class using liberal applications of hot glue.
Also good point about Cadence, I know that I can go from idea to PCB much quicker using the free tools but that won't mean much to any potential employers.  I suppose I'll just have to keep at it until their strange organization begins to make sense.

[tggzzz]

Cadence tools are an industry standard, and very well worth taking the time to learn and become proficient in.

Having Allegro on your CV will look good and be useful... free tools, not so much.

But here's the rub: if you have Cadence experience and your employer is an Allegro shop, then your experience will be no more useful than equivalent experience with a free tool.

As an employer, I would look for someone that produced a decent board with whatever tool is available to them. I wouldn't care that you had learned which buttons to push on tool Frobniz2015, but I would care that you had learned when/why things had to be designed, and that you could find out how to do it on whatever tool I happen to be using this year and this job.

Employers/jobs that require knowledge of specific tools (cf types/classes of tools) are to be avoided.

[AndyC_772]

But here's the rub: if you have Cadence experience and your employer is an Allegro shop, then your experience will be no more useful than equivalent experience with a free tool.

Allegro is the Cadence PCB package - though I do get your point regardless.

Personally I'd still rate experience with a professional CAD package as having greater value. YMMV.

[tggzzz]

But here's the rub: if you have Cadence experience and your employer is an Allegro shop, then your experience will be no more useful than equivalent experience with a free tool.

Allegro is the Cadence PCB package - though I do get your point regardless.

Personally I'd still rate experience with a professional CAD package as having greater value. YMMV.

For someone in mid-career, perhaps. But I'd still be more interested in someone that, for example, understood the consequences of vias and slots on signal integrity.

For someone in university, no. University should be about teaching fundamentals that will be useful throughout a career. Learning which buttons to push on this year's tool is not profitable in the long term (or even medium or short term!)

Where it is necessary to learn today's buttons in order that the pupil can demonstrate understanding and mastery of fundamentals, then it is a necessary evil. But it should be kept to a minimum.

[ROBOT]

Double check the pad footprints!  Never trust the library. For example you might say hey this is a 32pin qfn I will use this. Well guess what they have 0.5mm pitch 0.45mm and 0.4mm pitch in qfn 32. It's always best to either make your own or measure.

It your getting boards assembled make sure your >0603 resistors are rotated all the same way! It looks very unprofessional when you get 2 resistors next to each other and the value is upside down compared to the one next to it. Of course you can always change this in the CPL file you send to the assembly house.

When doing high density stuff any components that are direction specific like diodes make sure the pin 1 marker is clearly visible. It sucks when your assembly house puts the LED on backwards!

Not all qfn/dfn packages need the ground plane connected. In fact some do not even make the internal connection from GND to the center pad. (ohm it out if you want to know) but if it doesn't need to be grounded it's normally okay to run traces under the ground plane.

[jolshefsky]

I'd say to make your own schematic and layout checklists. Like literally a checklist (whether it's a spreadsheet or paper is up to you, but make sure there's a way to mark "done".)

For schematics, make sure you include documentation. Add descriptive text wherever you want. Use the tools at your disposal—particularly ERC, and get to zero errors (as a general rule).

For layouts, the first thing I often screw up is to start with an imperfect schematic. Of course I don't realize until I'm done with hours of layout that the pots turn the wrong way or some such silliness. The goal is to minimize these problems. Prototype as much as you can on breadboards so you can prove that it works.

The second mistake is ignoring the physical world. Sometimes I'll put components just barely too close together. Or I make it difficult to assemble. Or I put components where there's not enough clearance in the case (e.g. redesigning a board in an existing device.) The most common is to forget mounting holes. I add text and drawings defining the maximum dimensions of the board; locking the location of mounting holes makes sure I know where things can and can't go. Check your hardware (screws, nuts) and make sure your holes are the right size and with enough clearance; also, should they be connected to ground or not?

Run DRC before you generate Gerber files. Be sure to understand and enter your fab's DRC rules before you start. There's not much more annoying than routing a big old nest only to find out they want 10 mil clearance from copper to pad and you did everything with 8 mils.

Also check your unrouted nets. I'll often use the autorouter when I have a few dozen sub-millimeter traces to lay where it looks like things are touching already.

The final mistake is to make bad Gerber files. I started using makefiles with my projects to create the Gerber zip file to hand off to the fab. That way, if I forget the board outline layer, I'll get an error rather than a rectangular board (as one would from, say, OSHPark). Once you're ready to send, check your Gerber files with a viewer along with your drill file. This is where the checklist comes in really handy ... it's often at the end when I'm all excited to get a board made that I skip a step and end up with boards with errors.