4. MARCO CONTEXTUAL
4.5 MARCO TEÓRICO “Bullying” Acoso Escolar
Parts simply aren’t perfect. I have seen motor bearings wear out prematurely due to capacitive effects and have seen caps overheat and pop their tops. Truly the
best thing to do is keep looking at the datasheet. Parts engineers do their best to characterize the defi ciencies of the part and put it in the datasheet for you.
Thumb Rules
Always ask, is the error source in this component enough to cause the effect I am seeing? If the source of error isn’t large enough to be an issue, forget about it and move on. When fi xing errors, get a better part, shore it up, or design it out. Caps vary with frequency. Inductors have internal resistance. Semiconductors have voltage drops and heat issues. Voltage sources have internal impedance. You can’t study the datasheet too much.ROBUST DESIGN
Most engineers want to over-design, give themselves plenty of headroom, and use parts that are double or triple the spec they need. Usually the manager is there, saying, “It needs to cost less or it won’t sell ” or “Do you really, really need that part? ” To be honest, the answer lies somewhere between these extremes.
Can You Tolerate It?
Let’s start with a completely general off-the-wall statement that you might hear from someone with a sharpened, somewhat devilish hairdo. “A robust design handles two things well: the inside world and the outside world. ” A look of con- sternation scrambles across your forehead. “What in the world does that mean? ” you ask yourself. Let me proceed to enlighten you on this bit of pointy-speak. The inside world is all the parts that make up the design. In any production process, these parts will vary in specifi cation. The question to ask is, will the cir- cuit operate correctly over the tolerance ranges of the parts? If the answer is yes, the design is robust internally. The inside world is good to go. Don’t assume, however, that only electronic components have tolerances. This point is best taught by example. In a design I worked on some time ago, we were using an
optic sensor to detect the motion of a belt. We correctly analyzed the tolerance of the opto, but as we began testing on a pilot run we discovered that the belt we were using varied in opacity. If the optic sensor was at the high end of its toler- ance and the belt was at its most transparent, the signal from the sensor wouldn’t get high enough to guarantee that the logic input would read it correctly.
In a production run, a problem like this would appear as a random failure. This type of failure is known as a tolerance stack-up. It occurs when the stack-up (the additive effect of the variations) of two or more components combine to create a failure. It is more diffi cult to analyze than a single component tolerance issue. Probably the best way to preempt this type of failure is with the help of simu- lators. Take caution, though: Make sure that your simulation accurately repre- sents the design with nominal perfect components before you start running tolerance analysis on it. (See the section on simulators for more suggestions.) The great thing about a simulator, though, is the ability to vary all the compo- nents over their tolerances and see the effects without building a whole bunch of parts. You can then adjust your design and component specs to increase the robustness of the product as far as the inside world is concerned.
Now the outside world is a different animal. A good design can handle the things the outside world throws at it. In the electronic realm all sorts of inter- ference can disrupt your design. I once read an article that described something called a rusty fi le test. After the engineer was done with the part, he would plug it into the wall and plug in a home-built test fi xture next to it. It consisted of a wire from neutral connected to a fi le. The hot wire had a bare end that he would proceed to rub up and down the rusty old fi le, sparks fl ying every- where. 25 If the circuit passed this test without a hitch, he fi gured it was good
to go. This is known as EMI, or electromagnetic interference. It really is a whole topic unto itself, so I have dedicated a chapter to it. Skip ahead to it if you can’t handle the suspense!
Don’t limit your focus on the outside world to electrical interference. There are many cases where other things can cause a problem. Vibration, for example, can cause traces on a PCB to crack and solder joints to become faulty. Increased humidity can swell a cheap PCB, causing mechanical deformation and cracked connections. It can also combine with debris to create electrical shorts on circuits that you don’t want shorted. Temperature can be particularly tough on electrical
25 It was written by Ron Mancini in EDN, but I have to say: Do not try this test at home. There are much safer ways than the procedure described; I mention this test because it cre- ates a vivid picture of the junk out there that is trying to mess up your circuit.
CHAPTER 4 The Real World
188components. You should review the temperature range your circuit will be sub- ject to and compare that to the specs in the datasheet. Don’t forget to include the operating temps of the device you are using in this analysis. For example, power components usually get pretty warm just operating. Toss them into a 70-degree ambient and you could easily push them over the max temperature spec.
How do you go about making your design robust externally? There are several approaches to take:
■ The most important, in my opinion, is doing everything you can in the fundamental design to get it to handle the environment it is in. Often a few changes to the PCB layout itself can make a circuit handle EMI better than putting all the shielding around you can fi t. Larger traces can com- bat mechanical deformation, and a few well-placed holes can help man- age temperatures.
■ Reading, reading, and rereading the datasheet for the component you are using is probably the next most important thing you can do. The more you know about the parts you are using, the better you will recognize things that might upset your design.
■ The third and most extensive effort that will help you is to test, check, test, and retest the design. You need to recreate the environments that it will be subject to and see what happens.
Now, to top it off, you can have a situation where the problem is a combi- nation of the tolerance of the internal design and the environmental effects it is subject to. These situations are nearly impossible to predict and are often simply discovered in the course of business. There is only one thing you can do about that: Figure out what is needed to prevent it, make the change, and document it for future use on similar designs.
I recommend that every engineer and engineering group keep a document of design guidelines 26 where you write down those rules of thumb that you
discover along the way. Don’t just write it down, but read it regularly to keep those things you have learned fresh as you do each new design. This alone can be a powerful tool. Some years back I took over an engineering group. When I fi rst started managing it, it seemed like we were always being called to the pro- duction line for some weird problem or another. We spent more time chasing problems than engineering new products.
26 I like to call themgauntlets. If the design can run the gauntlet of passing guidelines and tests, that is when I deem it good enough.
We began a focus on robust design principles, and one of the fi rst things I implemented was the design guideline documents. Every time we found a new design rule to follow, we wrote it down and referred to it regularly so that it would be implemented with each new design.
Over about a three-year period, those urgent calls to production began to drop off. We went from spending over 50% of our time in production support to spending less than 10%. A couple years after that, we were spending less than 1% of our time dealing with production problems. Considering that we were moving tens of thousands of products out per day, it was a great achievement. Months would go by without a call, where before we got calls every day. When problems did occur you could nearly always trace back to a guideline that we had written down and simply neglected to follow. The hard part became refer- ring back to those documents each time we created a new design. That being the case, I suggest you try not to let your guidelines get too large. The bigger these documents, the less likely you are to read through them. So try to keep them to a few pages, since they will have a tendency to grow a lot.
In an effort to quantify what the outside world can do, many standards have been written. They are some great yawners (meaning they will knock you out in about 5 minutes of reading); however, they can give you some real insight into what your design will be subject to from the outside in. I’m referring to documents like IEEE 62.41, which describes the world of EMI, or UL 991, which describes how to make a control safe. The list goes on and on. Do a little research into what you are working on and see if someone has written something about it. If your boss doesn’t understand the need for time to do this, show him this paragraph:
Boss, it might seem like nothing is getting done when the engineer is sitting there reading, but trust me, this effort can save you millions in production downtime, so give your engineer a chance to succeed and you will not regret it. Engineer, this doesn’t mean that you should just read and never design anything; I would limit this research to about a 10 to 20% ratio of design vs. research; double it if you are doing something you have never done before.
Reading these documents works particularly well if you are tossing and turn- ing all night as you try to fi gure out what is wrong with your design. I would keep them by the side of my bed. That way I could learn some more for a few minutes and also get some sleep. They not only help with the design, they are a great cure for insomnia!