Electronic Design
Bob’s Mailbox

Bob’s Mailbox

Hi Bob,
My friend was asking me about overload recovery in op amps, and all I could tell him was to avoid those older op amps that lock up on overload. ( Yet these are not necessarily useless. The scheme listed below will let one of them work okay. /rap). I looked in Herpy (What or who is Herpy? Never heard of her. /rap) and I looked at a lot of datasheets, and I never saw any reference to how quickly op amps can recover from the output stage hitting the rail (or the input stage hitting the rail) and what happens when they do. So, I figured I would ask the op-amp guru.

Scott Dorsey

Hi Scott,
I wrote a very good story about this about 30 years ago. It relied mainly on feedback zener bounds. I wonder if I can find a copy. It is very rare for an op amp to be characterized in any way for over-drive recovery. I think ComLinear had one or two, long out of production.

Assuming you have a decent big power supply, like ±15 V dc, you can put in zener diode clamps to stop the output around 11 V, yet keep full accuracy to ±9 V. What functions are you trying to do? Fast linear amplifiers? Integrators? There is generally no simple solution with fast performance to please everybody. How fast do you need recovery? I can’t make everybody happy.

Here’s a hack. Call the summing point a, another point b, another point c, another point d. Put in a 9.1-V zener diode from VOUT to d and another zener opposing from c to d. Put a 2N3904’s VBE from c to b and another from b to a. Put another 2N3904’s VBE from a to b and another from b to c. And when I say “VBE,” I mean base and collector tied together, versus the emitter.

Put 2k from c to ground and 2k from b to ground. When the output is swinging ±9 V, there will be no leakage. When it hits 12 V, the clamp will begin to work. There are many variations. How’s that make you smile? That’s what I told you back in 1979. I can’t solve every problem, but I can solve almost any one problem.

Beast regrds,

Hi Bob,
I had to come up with a simulation of a metal oxide varistor (MOV) for some voltage surge requirements. I design high-power sonar amplifiers, and ship power systems have surge voltages that our amplifiers must withstand.

I have been using MOVs for about 40 years, but I never tried to model their voltage and current relationship. A new +50% voltage surge for 2 seconds spec made me calculate the extra dissipation. I used a ruler to measure and calculate the slope of the voltage-versus-current graph from Panasonic:


The fourth page has the V-versus-I graphs. I was amazed at how large the exponent was. I looked at the other MOV curves and decided that the ERZV20D271 had the flattest voltage curve. From 1 mA to 1 A, the voltage increases from 308 V to 360 V. That curve can be described as I = (V/360)48.5. The voltage ratio is raised to the 48.5 power! I was only interested in the 1-mA to 1-A region.

Continue on next page

(That is a little wild! That is a voltage ratio of 1.17:1. A silicon transistor can easily go from 520 mV of VBE (at 1 mA) to 700mV (at 1 A) , and then when it warms up, to 660 mV. But that is only a voltage ratio of 1.27:1. However, silicon zener diodes can easily have a voltage ratio better than 1.07, for a 1000:1 ratio of currents. So while it’s not truly exponential, zeners can have a very sharp knee. When does a knee become an exponential curve? Heck, I dunno. /rap)

I have never seen an exponential power that large. The largest I knew of before was the failure rate of incandescent lamps increasing as the 12th power of the voltage. I should have suspected something unusual from a graph that extends 11 orders of magnitude on the horizontal axis, but only two on the vertical axis. Do you know of any other “big exponent” phenomena?

(Well, every day, silicon transistors do amazing things, but their exponent isn’t as high as 48.5. Hey hey! Go to Antarctica. At –55°C, a silicon transistor has improved gm, only 132 mV per three decades. So the instantaneous delta VBE will be 132 mV, and after you allow a little time for warmup, the ratio of VBEs will be very close to 1.17:1. So, a cold transistor has just about the same curve as the MOV you saw. The exponential will be surprisingly close to 48. /rap)

By the way, our MOV easily handled the surge voltage.

Mike Fry

Hi Mike,
Very good. I may be using some MOVs soon. And sorry for the delay. I couldn’t reply to your math until I had my slide rule and some paper and pencil.

Beast regrds,

I hate to bug you, but I am writing this in desperation and as a last resort. I am a relatively new EE student and have purchased Troubleshooting Analog Circuits to go along with the troubleshooting class I’m taking. I love the book, but as you know, it no longer comes with the CD. I have searched everywhere for an older edition of the book, but just can’t track down a copy with the CD.

(Butterworths only printed about 1000 copies of the book with the CD. When they finally sold out, they decided to print no more with the CD. I don’t have a copy of the CD. I never looked at the CD. /rap)

Do you know of any way to get a copy of the CD, or at least access to the sample circuits that it contains?

(I think the sample circuits on that CD are just exactly the circuits in the book. I never looked at them. I know how to analyze any of those circuits without any computer help or Spice. So, I would never think of analyzing them with any computer analysis.

Note that some of the circuits in the book use features of diodes or transistors that do not come supported by any device makers. Look at the 1N4148/1N914s that turn on late. Nobody has a model for that. Most engineers have no idea that that could happen. In fact, if you have a bag of 1N914s, some may remember how to turn on, and some may not. Computer analysis is going to be less than useless for analyzing this. ook at the 4N28 driving a 2N3904.

Continue on next page

Look at the 4N28 driving a 2N3904. Can computer models of these match my observed response? Note that the range of 4N28s goes from high gain to LOW– and High speed to LOW–. You’ll never see any 4N28’s that look like each other, so, how can you ask a computer model to look consistent? /rap)

I guess I should also mention that I am using Multisim 10.0 and 11.0, so I’m not even certain that the CD is even compatible. If not, is that the reason the CD is no longer included? In any event, I would really appreciate any assistance you can offer.

Michael Baseman

Hi Michael,
The CD is no longer available, as near as I can tell. It has been unavailable for the last six or eight years. You’re better off without it. There is nothing it can do that you can’t do better yourself.

You want help? Very simple. Here is my advice: take any circuit you are interested in, and type it or enter it into your favorite Spice engine. If that is what you want, that is how easy it is to get it.

Let me observe that just in case Spice does not like to do good analysis, well, I never said it did. I never ran any of the electronic simulation studies. That was left as an exercise for the reader. I’ve always warned you that Spice often tells any kind of untruths about circuits. So whatever Spice tells you, take it with a grain of salt.

Beast regrds.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.