Advertisement

Blog

Murphy Rules!

Our guest Author, the esteemed James Bryant

James Bryant [james@jbryant.eu] has been a European applications manager with Analog Devices since 1982. He holds a degree in physics and philosophy from the University of Leeds. He is also C.Eng., Eur. Eng., MIEE, and an FBIS. In addition to his passion for engineering, James is a radio ham and holds the call sign G4CLF.

Q : Why don’t my amplifiers and converters work correctly in the published circuit?

A : When giving basic training on “Analog ICs, Their Care and Feeding,”1 I am often asked questions of this type. My questioner seems amazed when I ask for more detail before attempting to reply; seemingly “the published circuit” should provide all the information I need.

Experience shows that when the circuit in the data sheet or application note misbehaves, it is rarely the IC that is at fault. The problem is almost always due to some failure to create the circuit as given in the documentation, and that that failure is often due to Murphy’s Law. The simple form of Murphy’s Law states, “What can go wrong, will go wrong,”2 but I prefer to restate it as, “The Laws of Physics always work, even when you’re not paying attention.”

The published circuit makes a number of assumptions: the supplies will be noise-free, well-decoupled at HF and LF, low impedance, and the correct voltage; the source and load impedances will be reasonable for the intended application; the passive components will meet their specifications; the environment will not introduce noise into the circuit; and the test gear will work correctly.

Even though the engineers presumably want to solve the problem, it is sometimes difficult to persuade them to provide the details needed for diagnosis. It is necessary to know the source and load impedances likely to be encountered; if the power supply is less than ideal; if the working environment has magnetic, electrostatic, electromagnetic, or even acoustic noise; if the components used have parasitic reactances, large thermoelectric, piezoelectric, or photoelectric sensitivities, or are dissipating more power than they can handle; and what test gear is being used and if it has been checked for correct operation. In fact, in many cases when I ask for these details, I either hear no more of the problem, or am told “Oh, we solved it!” Often this means that looking for the details had revealed the cause of the problem.

I have discussed supply decoupling, environmental noise, and the unexpected properties of simple components in a number of previous RAQ columns3 , and would advise anyone with unexpected circuit performance issues to review their system in the light of the advice given in those columns.

If a system is not working correctly and you have checked that the circuit is exactly the same as the one in the circuit diagram, it is well to perform simple function tests on as many of its components as you can. Often this can be done in situ with a digital multimeter (DMM) and an oscilloscope. Check that the voltages on every node are as expected and that the system is not oscillating (unless it is supposed to be oscillating, in which case check the frequency, amplitude, and waveform). Touch all of the components to ensure that nothing has become too hot. Also check for dry-soldered (high impedance) joints.

When there’s a problem, check everything. As former U.S. president Ronald Reagan said in good Russian (he was also an actor, and knew how to learn his lines): “Доверяй, но проверяй.”4 This includes test gear; it’s easy to assume that it’s working, but it might not be. I wasted a couple of hours trying to discover why my alternator was only generating 500 mA instead of >20 A. Then I discovered that one of the leads of my current meter had a resistance of over 25 Ω due to a dry joint and the entire problem was due to the test gear, not the alternator itself.

Only when you’ve proved that Murphy is not responsible should you actually start analyzing the circuit.

1 Details of my courses on analog basics can be found at Courses in Basic Analog Electronics

2 Never forget Mrs. Murphy’s Law: “Murphy is an optimist!”

3 Links to RAQs:

Radio Frequency Interference (or how to get a free lunch at a very good restaurant)

Resistor Noise can be Deafening, and Hard to Reduce

Op-Amp Noise can be Deafening Too

Resistors (and old ladies) have hidden depths

Stress Is Bad For You – and For Analog ICs

Lock Down That Noise – Don’t Let It Escape

Glass Diodes May See the Light – and Hum

Taming A/D Converter Power Supplies

Make Do!

Do Something with that Unused Pin!

In and Out – or – Why Have a Capacitor?

Nobody is Perfect – Not Even an Engineer!

4 “Doveryai, no proveryai.” An old Russian proverb. In English, “Trust, but verify.”

1 comment on “Murphy Rules!

  1. eafpres
    May 6, 2015

    Hi James–there are so many great points you make it is impossible to really choose a favorite!  But, I did like “touch everything”.  In this high tech digital world the simple stuff gets forgotten.

    Your narrative regarding ensuring the “published circuit” has been followed is more elegant than my usual refrain when I suspect problems are operator-related due to egregious ignorance:  “RFM”  (politely:  Read the Freakin' Manual)

    I worked for many years in RF/wireless products.  We were in production of an antenna assembly which packaged together a cellular antenna (at that time, 800 MHz) and a GPS antenna assembly that integrated a 25 dB LNA.  A customer began reporting that some units not only did not work, but appeared to be jamming other GPS receivers in the vicinity.

    Long story short, although we were sure it was a customer issue, and if not, an issue with the GPS module supplier, it was neithre.  It turned out that the way we had mechanically integrated the GPS antenna/LNA involved changes to the shield configuration, which every now and then caused the amplifier to oscillate, and the entire assembly to radiate right in-band.  Murphy 1, us 0.  Lesson learned.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.