Anablog

Jun 24 2008 10:40AM | Permalink | Email this | Comments (2) |
Blog This! using:  Blogger.com | LiveJournal |
Digg This | Slashdot This | add to Del.icio.us

I came across this nice wrap-up of useful system design tips from a guy that sounds like he has really been in the trenches. I found his site after seeing some comments on the newsgroup, sci.elect.design. He has done the impossible, he actually worked with the management of NXP to change their website so it is useable. It seems to work in Opera and from the comments in the newsgroup, it is a great improvement over the old website. Hurray NXP. My comments about bad websites are here and here.

Now back to Joerg’s hints for system designers. This type of advice is sadly rare. The semiconductor companies have a ton of money and they try to have applications departments, but putting a single chip on a demo board is not systems design. The semiconductor companies put all kinds of status on circuit design, which is what their IC designers do so they think that this is the high-holy of design. But as anyone that really got a product designed and built and tested and shipped can tell you, circuit design is just a part of the job, and sometimes a small part. It is getting all the parts working together and the way you want as well as documenting how to put it together that is the hard job.

Documentation is one of the many areas where I wholeheartedly agree with Joerg. I was blessed to work at Ford motor where there were 19 assembly plants making our pickup trucks. When you made a change you had to tell them if you were going to change it at the model year or make a running change. Then you had to tell them if they had to change other parts at the same time. Then you had to tell them if the obsolete parts went into service stock or got scrapped or got run out. And you didn’t do that with Post-it notes. It took lots and lots of documentation to build 1 million trucks a year. And that was just part of it. After the auto job I worked for ESL, a military contractor. Say what you will about 600 dollar hammers,. The boys in camouflage know a thing or two about good documentation. Then I spent 20 years as a consultant where I just assumed I would not be around after the design was done. By doing good documentation I became far more valuable to the companies that hired me. They would often call six months later needing a copy of something I did, and I always had it, on 5-inch floppies maybe, but I had it. The love of documentation is what eventually led me to this writing gig.

So many other things that Joerg says are perceptive. Design for the customer; design for production and one of my big ones —design for serviceability. When it is easy to fix that means it was easy to troubleshoot as you developed the circuit.

He talks about designing for cost and believe me, he is right. I have worked on many projects where the people told me that cost does not matter, but by the end cost always mattered, even for the military contractors making the 600-dollar hammers.

His “design for transport” also makes sense, I have mentioned on this blog how the worst accelerations that an auto experiences are on the railroad cars delivering the new models to the dealer.

One of the neatest parts is where he talks about grounding. Just like Henry Ott, and myself, Joerg does not advocate cutting up ground planes between analog and digital. Henry Ott would repeat this over and over in his lecturers— that the applications people that work for the semiconductor companies advocate cutting up planes, and that these people are wrong. It is one thing to get an A-to D converter working on a little demo board, but when you are designing  a real-world system, chopping up planes causes far more grief than it ever saves. I have fixed a lot of problems putting back that nice slab of copper ground plane, (Henry calls it a reference planes since there is no ground in a satellite or airplane or car). Then as Henry and Joerg point out—you use placement and routing discipline to make sure your analog and digital signals do not interfere. One thing Henry discovered is how less the disturbing field emanates from a buried stripline than on a microstrip on the outside layer. It is something like 5 times the width versus 20 times the width. Even on surface diff-pairs, the disturbing fields can radiate out further if you put them on the top layer.

Joerg goes on to give some tips about clocks and bus terminations and oscillators. One tip I can add is to consider series termination as well as the ac termination that Joerg talks about. The signals look pretty bad but the reflections get absorbed in the series resistor before they hit the output of the driver.

About the only thing I wonder about is on the newsgroups where Joerg says he rarely uses op-amps, instead preferring discrete circuits. I worked at the amplifier group in National Semi and believe me, there is stuff an IC designer can do with a modern process that no amount of discretes can compare to. Then again, a 5-cent voltage follower transistor can do the job of an op-amp buffer on occasion.

Way to go Joerg. This is an example of a real analog artist. He is willing to tell you how he does his job and what tools he uses, unlike a tradesman that tries to keep everything secret. Analog is hard enough without secret societies.