(Editor's note : this is part of an on-going series of “dialogues” between the authors; there are links to the previous installments at the end, immediately above the “About the Authors” section.)
Dr. T (Tamara Schmitz) : Hey, Dave! Watcha doing out here by the fountain?
Dave (Dave Ritter) : Cooling off in the shade–we're heading for 104 degrees by this afternoon!
Dr. T : It's actually cooler inside.
Dave : Yeah, but then I couldn't check out this little toy I got for my nephew. Neat little solar-powered boat. See, it even “putt putts” when you turn it on.
[Dave puts little tug boat replica into fountain pool and flips switch. Boat slowly crosses pool, gets caught in a shower of falling water, and promptly sinks to the bottom, nestled on top of glistening coins.]
Dave : Hmmm. I guess motor boating isn't my sport.
Dr. T : Maybe you should stick to electronics, at least it's drier. Which reminds me, isn't the term “motor boating” used in electronics somewhere?
Dave : Yes, it is. Haven't heard it for a while, but old audio amps used to “motorboat” when you didn't bypass things well enough and got inadvertent positive feedback.
Dr. T : So they called it “motor boating” because it “sunk” the amp? That seems strange.
Dave : Well, actually, it made the amp sound like a little two-cycle engine. You know, “putt putt putt”, like my little boat did before it sank to the bottom.
Dr. T : Oh, I get it. The very-low-frequency oscillations made the amp “pop” every time it slewed from rail to rail.
Dave : And sometimes, the rate would change as you adjusted the supply or volume control. It sounded like you were rev'ing the engine. So they called it motor boating. I still tend to use the term anytime a system goes into high-amplitude, low-frequency oscillation.
Dr. T : Does that happen very often?
Dave : Of course not. A well-designed circuit will not “motor boat”. I haven't had an issue with that since, well, let's see. . . since yesterday afternoon.
Dr. T : Rare indeed! What happened?
Dave : My signal-processing chain wouldn't start up. It would start to start, the supplies would ramp up, the bandgap reference would settle, and bias currents would start to flow.
Dr. T : Sounds good. What went wrong?
Dave : Hmmm. . . .I can't tell you exactly because I'm in design now and everything is very hush-hush, top secret, highly classified.
Dr. T : Of course, Top Secret Agent Dave. But the problem must happen in discrete circuits too.
Dave : Actually, don't tell anybody, but the chip I'm doing is based on an apps circuit for equalizing long cables. I could tell you about that.
Dr. T : Sounds great. Let's pretend I needed to equalize a really long cable. Maybe it's for a video application.
Dave : Very shrewd guess, Dr. T. We have a board that uses our very-low-noise ISL28290 rail-to-rail op amp in a multi-stage equalizer. It's great for this app, very low noise and plenty of bandwidth. I'll use that to illustrate the problem. The simulations looked good at 5 microseconds as the final stage started to ramp up and hit the rail for a few hundred nanoseconds. Then suddenly it shot to the negative rail, hung out there for a microsecond or so and then popped back to the positive rail. Once it started doing that, it never stopped. So I had to tell my boss it was motor boating.
Dr. T : And he thought you were vacationing at the marina?
Dave : No, he just wanted to know how fast. The chip was motor boating at about 25 MHz, which is more like speed boat territory, I suppose. Our discrete circuit would be a bit slower.
Dr. T : Okay, so it was a really fast motor boat, but still the result of some inadvertent positive feedback.
Dave : Sure was.
Dr. T : Can you sketch out the signal chain?
Dave : [Kneeling in sand] It's something like this (Figure 1 ). There's a differential front end that buffers the input signal and establishes the common mode DC level.
Figure 1: Differential input stage
(Click on image to enlarge)
Dr. T : So the signal was on two wires, each out of phase with the other.
Dave : Yep. The input comes in on a twisted pair. U1 is a dual amp that buffers the input signal and U2 senses the common mode level and corrects the DC at the inputs. After the front end, it enters a series of “pseudo differential” equalizers, (Figure 2 ).
Figure 2: Adding the pseudo differential equalizer stages
(Click on image to enlarge)
Dr. T : Hmmm, pseudo differential. What do you mean by that?
Dave : Basically, a true differential stage gives differential gain while suppressing common-mode gain. A pseudo differential stage can give the same differential gain, but does not suppress the common-mode gain. Instead, in the pseudo differential case, the common-mode gain is typically 1.
Dr. T : You are going to tell me why a designer might prefer to use a pseudo differential stage instead of a true differential stage, aren't you?
Dave : All in good time.
Dr. T : Let's get back to your system description. The intended signal was differential and the equalizers were designed to boost that.
Dave : Yes.
Dr. T : But there was a common mode signal (the same in both wires) that got carried along, but not boosted.
Dave : Wow! Pretty smart for a girl.
Dr. T : (response inaudible) And somewhere, probably at the output, you converted to a single-ended signal, just one wire with a voltage relative to ground.
Dave : Exactly. But we skipped over something. We used the pseudo differential stages because they are the lowest noise and most immune to power-supply ripple, but we also need to reject the common mode variations from the input as soon as possible.
Dr. T : Well, the pseudo diff stages won't do that, they just carry the common mode signal along. So we want to reject the common mode at the front end, and I'll bet that happens right here. (Figure 3 , blue arrows)
Figure 3: Entire circuit with inadvertent feedback
(Click on image to enlarge)
Dave : Right again.
Dr. T : So front-end amp U2 feeds back to reject (balance out) the common-mode input, and it would need a reference voltage to set its output common-mode level.
Dave : Sure does! In the old days the output reference would be ground, zero volts. But these days many chips need to operate from a single supply, so the signal floats between the rails. The “effective ground” here needs to be half supply or about 2.5 volts. Let's call it VREF .
Dr. T : What about the output common-mode reference? Your differential-to-common-mode stage also needs a reference, too, doesn't it?
Dave : Yes, and that's where I got into trouble, because I used the same voltage.
Dr. T : Let me think about this. Even though they need to be the same voltage, they are being applied at different ends of the signal chain. What's the gain of that whole chain?
Dave : Ummm, it's about 60 dB differentially. Sometimes more.
Dr. T : And you used the same node to reference both?
Dave : Well, I did yesterday.
Dr. T : In Figure 3 you show that the last stage is actually driving current into the VREF through two resistors. And those resistors are. . . .?
Dave : About 1 kΩ
Dr. T : So you were pumping milliamps into the output of the reference amp?
Dave : That's a couple of volts divided by 1000 Ω, yes.
Dr. T: What was the effective resistance of the VREF source?
Dave : 10 or 20 Ω.
Dr. T : So you've got tens of millivolts of feedback to the input through that resistance and you had a gain of over 1,000×.
Dave : Sounds bad when you say it that way.
Dr. T : And it motor boated.
Dave : at 24 MHz.
Dr. T : So you fixed it by buffering the VREF feeding the output stage, breaking the inadvertent feedback loop? (Figure 4 )
Figure 4: Corrected circuit with buffered VREF at output
(Click on image to enlarge)
Dave : Would like a job in design, Dr. T?
Dr. T : Sorry, got enough to do already!
Previous “dialogues” in this series:
- A bypass-capacitor dialogue peels back the layers, Part 1
- A bypass-capacitor dialogue peels back the layers, Part 2: The theory of ground relativity
- A bypass-capacitor dialogue peels back the layers, Part 3: Continuing the discussion on layout considerations
About the authors
Dave Ritter grew up outside of Philadelphia in a house that was constantly being embellished with various antennas and random wiring. By the age of 12, his parents refused to enter the basement anymore, for fear of lethal electric shock. He attended Drexel University back when programming required intimate knowledge of keypunch machines. His checkered career wandered through NASA where he developed video-effects machines and real-time disk drives. Finally seeing the light, he entered the semiconductor industry in the early 90's. Dave has about 20 patents, some of which are actually useful. He has found a home at Intersil Corporation as a principal applications engineer. Eternally youthful and bright of spirit, Dave feels privileged to commit his ideas to paper for the entertainment and education of his soon to be massive readership.
Tamara Schmitz grew up in the Midwest, finding her way west with an acceptance letter to Stanford University. After collecting three EE degrees (BS, MS, and PhD), she taught analog circuits and test-development engineering as an assistant professor at San Jose State University. With 8 years of part-time experience in applications engineering, she joined industry full-time at Intersil Corporation as a principal applications engineer. In twenty years, she hopes to be as eternally youthful as Dave .