Archive of the former Yahoo!Groups mailing list: Homebrew PCBs

previous by date index next by date
previous in topic topic list next in topic

Subject: Re: Selecting Method For Prototype Boards

From: cs6061@...
Date: 2014-07-21

Being sure your design is error free is always going to be an issue, it's a probability thing.  I have seen simple boards prototyped 4 times before the guy got it acceptable then I have seen them come out of the shoot 100% accurate.    There are two classes of errors, the first one is electrical and how your circuit works, nothing the pcb process can do to fix a bad electronic design, improper component choices or just bad engineering.  So the first thing is to make sure you circuit is correct and been prototyped some how if necessary and working.  The second thing is to carefully select all your components, picking a component that is built from unobtanium or not suited to the task is the first risk you must blunt.  Don't underestimate this risk, a part in stock may not be when you get your boards back.  Find alternate sources up front, best to use parts that have multiple sources.

Now we come to the PCB process.  First is the schematic and the component libraries,  I have always found that the libraries are a big source of error, very easy to get a pin switched around.  So one needs to be proactive and do a formal library check of all newly built components (twice and by a second person if you want to be sure).   The second major source of error is poorly captured schematic.  Its easy to miss wires have wires that look connected to nets but are not, bus numbering errors and the like.   Again double QA the schematic.  I most always name all nets, then print out an text based net list format and spot better 100% check the connectivity.   Once you are sure the schematic is correct and you have identified all your parts only  then is it time to do the PCB layout.

We have an entirely new set of issues with the layout,  first is the mechanical is the board shape correct, are all the mounting holes correctly located and of proper size?  Again it good QA that saves the day.   Again we have a big risk with the pattern library and its coordination with the component library in the schematic.  Is the clock on pin 1 or pin 11.  Easy to make a typo.   Is the pattern footprint correct, does it have proper size pads, holes.  Is the silk art and solder mask correct.  SMT helps a lot here as the number of different pattern types is reduced and more standardized.   Check Check and more checking.   Next do the placement and review it again --you got any component height issues with your housing?   Now you are down to routing -- plenty of places for errors here,  trace width, clearances, SI,  EMC.  Use the tools DRC checks but also use your mark 100 eyeball.  I have found nothing beats printing out the layout 2:1 and using a highlighter to trace it all out comparing to the schematic -- though that is a lot of work and may not be practical on more than a small board double sided board.  I have worked on 8 layer designs with thousands of nets and 70 pages of schematics, in this case the best you can do is a spot check but you are just forced to trust the tools.  In my experience the most risks are with the library and the component selection process, check these good an you have a good chance of making a 100% board the first time.

Check check and more checking -- then prototype if you are still unsure.   But a home built board is not really representative of what you would send to a commercial house to add a lot of value.    The costs on getting 1-3 boards these days is not all that bad unless you are talking multi layer.  Just get a couple of prototypes before sending in your order of 100's.  If nothing changes most houses will not charge a second NRE for the production order.
Craig