ESR of Regulator Output Capacitor

About baldness...

Regarding my recent EDN article "Not all EMC Engineers are Bald", several readers wrote asking how I determined the average career lifetime for digital engineers. You should know that my figure is supported by few, if any, facts, but here they are nevertheless. I developed this information in the process of researching the market for technical books and magazine articles.

  1. EDN magazine surveys indicate the total available market of practicing digital engineers at any one point is somewhere in the range of 70,000 to 100,000 (this information is about 3 years old).
  2. Colleges and universities worldwide graduate 10,000 - 14,000 digital engineers each year. Do the math.
  3. The IEEE does surveys every year that have the whole story. I'm remembering the last one I saw in Spectrum from a l--o--n--g time ago had a 7 year expected career lifetime and I think it's getting shorter.
  4. Few digital engineers are bald :-) In my experience working with (literally) hundreds of companies what happens to digital folks eventually is that they move to management, marketing, or go get an MBA. A few flame out to become artists or ski instructors (no offense meant, but that seems a waste of raw brainpower).

The major factor responsible for our collectively short career span is the lack of training. With the rapid evolution of technology digital engineers cannot keep current without massive amounts of on-the-job education. Unfortunately, the tight schedule in most development departments leaves no time for training. Compounding our difficulties is the executive-level perception that as long as universities pump out enough qualified designers there isn't a problem. Well, maybe not for management, but if you want to keep doing the job you love to do you'd better keep educating yourself. Nobody else is going to do it for you.

ESR of regulator output capacitor

Rodrigo Rodriguez at Siemans:

How can the ESR of a bulk capacitor (tantalum or electrolytic capacitor) affect a linear voltage regulator? The capacitor I am thinking of is used at the output of a linear voltage regulator. This capacitor is used to stabilize the output voltage. Voltage regulator suppliers suggest, for example, that ESR of the capacitor must be less than 10 ohms to assure a stable voltage at the output.

Thank you very much,
Rodrigo Rodriguez

Thanks for your interest in High-Speed Digital Design.

The interaction between the ESR of the load capacitor and the performance of your linear regulator is a matter of analyzing the stability of the feedback network used to control the regulator. As with many linear systems, the regulator uses a high degree of feedback in its control loop. The stability of the resulting circuit hinges crucially upon the type of load applied to the output of the linear amplifier. By restricting the ESR of the capacitor, the manufacturer ensures that the phase response of your load remains within the range over which the linear amplifier won't oscillate.

You can test the importance of the ESR by connecting a few ohms in series with your bypass capacitor. If there are no other loads present (no other bypass caps or anything) the linear regulator may oscillate in a spectacular fashion. Sometimes the oscillation persists only briefly before the regulator burns out. This is one of those cool demonstrations I wish I could do in my public seminars (but it's not safe!).

Regulator instability was a big problem in the days of old two-layer board design. In a two-layer board you don't have the luxury of using solid power and ground planes, so the capacitors all hook up to the regulator with long traces. The impedance that the regulator sees in such a board is determined almost entirely by the closest capacitor. The specs on that first capacitor therefore become crucial.

In a modern multilayer board with (possibly several) power and ground planes the impedance that the regulator sees is more a function of the entire array of bypass capacitors. Depending on the exact properties of the internal feedback loop within the regulator, you may in some cases get away without have a bulk decoupling capacitor at all. Still, I suggest you follow the manufacturers recommendations as that should produce a design that is always stable.

Best Regards,
Dr. Howard Johnson