Engineers often follow simple rules of thumb that lead invariably to functional designs. However, there can be overriding conditions that render rules of thumb or even so-called best practices incomplete and sometimes misleading.
Take shielded cables. Without a detailed analysis, most engineers regard using shielded cable as inherently “good.” It heads off noise problems, after all. But experience shows it can also introduce undesirable effects. And when circuits don’t play by the rules of thumb, even experienced engineers can be stumped.
Here are three examples from when I worked as a part-time electrical engineer at Sundstrand Aerospace—as an engineering professor the rest of the time, I was usually the one called in to figure out what went wrong and how to fix it.
Professor, we have an unstable DC signal
I was at my Sundstrand desk when I received a frantic call from my engineering colleague Tim. He was behind schedule and the test stand he was working on was exhibiting erratic behavior. Consequently, our manager, Doug, was there by his side to “help.” For any engineer, that’s usually a very uncomfortable situation.
I walked downstairs to the test stand and asked Tim to explain the problem.
“We are ringing out an integrated test article (ITA) module which interfaces the unit under test (UUT) with the test stand. We are performing a wrap-around self-test. We are directing a DC voltage into an analog-to-digital converter (ADC). The output of the ADC is fed into a digital-to-analog converter (DAC). The scaling is designed so the DAC output voltage should match the applied DC test signal.”
I asked Tim to see his ITA board schematic. He was using a noninverting op amp circuit to deliver the test signal. I puzzled for a few moments and then asked, “Tim are you using shielded cable on the op amp’s output?” Tim replied, “Sure, we did not want to have noise problems!”
Bingo.
“Tim, you’re connecting a capacitive load across the op amp output and it’s oscillating. The DC shifts are produced by the asymmetrical clipping of the op amp’s output voltage.”
Doug was visibly upset and asked, “Tim, didn’t you put a scope across the op amp’s output?!” I tried to cover for Tim and responded, “This is a pretty obscure problem. I only recognized it because I had the same problem a few months ago.” With a sigh and a huff, Doug left immediately.
I sketched the below image for Tim to explain the reason why the capacitive load caused oscillation.
The added feedback capacitor, CF, in this circuit allows cancellation of the pole and zero contributed by CL. To put it simply, the zero from CF coincides with the pole from CL, and the pole from CF with the zero from CL. Therefore, the overall transfer function and phase response are exactly as if there were no capacitance at all. In order to assure cancellation of both pole/zero combinations, the above equations must be solved accurately. Also note the conditions; they are met easily if the load resistance is relatively large.
Tim pleaded, “So, what do we do?” I suggested he find a 20Ω resistor for Rx and use the 30pF disc ceramic capacitor I retrieved from my shirt pocket. I announced, “Here’s our CF.” (I had been in the lab earlier.) Tim installed the components and the problem—like Elvis or Frasier—left the building. I suggested a detailed analysis should be performed later when managers were not hovering about.
I surmised, “Tim you just experienced Murphy’s Law on Experience: The trouble with using experience as a guide is that the final exam often comes first and then the lesson.” Tim smiled as I left.
Lesson 1: Using shielded cable minimizes common mode noise but has a capacitive characteristic. If an op amp must drive it, incorporate frequency compensation. Alternatively, use a buffer with adequate capability to drive the capacitive load presented.
Erratic frequency counters
Mark, an experienced engineer, asked me if I would join him in the lab. I agreed and we trotted down to the project development laboratory. He was working on a test stand module called an oscillator control panel. A 1MHz clock generator was connected to the panel which contained dividers to produce 400Hz (the generator point of regulation (POR) voltage frequency) and 1200Hz (the permanent magnetic generator, or PMG) frequency. The 400Hz and 1200Hz frequency counter displays were bouncing terribly.
I suggested we put a scope across the 1MHz input to the oscillator control panel. Mark announced, “I’m already on it.” The 1MHz input clock was ringing significantly (see below image). We looked at the input circuit: it used a CMOS digital buffer and the circuit designer had connected a 10kΩ pull-down resistor from its input terminal to ground.
Mark exclaimed, “Yes, and it has a characteristic impedance of 50Ω! We have a transmission line problem!” Mark powered down the panel and replaced the 10kΩ resistor with a 51Ω unit. Problem solved. With no ringing, the two frequency counters displayed 400Hz and 1200Hz properly.
Mark muttered, “These digital designers don’t have a clue about transmission line effects. They are a bunch of tossers.” (Mark is British and “tossers” is slang for stupid or despicable people.) I could not help myself and smiled at his colorful assessment before walking back to my desk.
Lesson 2: While a 10-kΩ pull-down resistor is often used for CMOS digital inputs, the resistance should instead be equal to the characteristic impedance of the coaxial cable connected to it. Otherwise, significant ringing can occur.
Bogus acceleration readings from a vibration panel
I received a call from Ben, the Sundstrand Plant 8 supervisor, just after I had finished teaching my morning engineering circuit analysis class. He asked if I could come to Plant 8, the Sundstrand remote test lab, right after lunch. I agreed. Although remote, Plant 8 is still within easy walking distance from the main plant. The Plant 8 facility test cells often employ explosive fuels like hydrazine for Space Shuttle auxiliary units (APUs) or JP4 jet fuel for engine start systems (ESSs). For safety reasons, the operational tests are run in test cells with thick concrete walls. Instrumentation and controls are in the hallway just outside the test cells.
I entered Plant 8 and went to Ben’s office. Ben was there with a young test equipment engineer, Frank. Ben welcomed me and said, “Hi Professor! We are having erratic readings on a vibration panel that Frank designed.” I replied, “Hi Frank. Let’s go look at that pesky panel!”
We walked down the hall to the test cell. I looked through the cell window and asked, “So, what kind of testing is being done?” Ben replied, “We are running vibration and shock tests on a 737 generator control unit (GCU) using a Sentek shaker.” Beaming slightly, Frank added, “We’re doing random, sine and shock.”
I inquired, “What type of accelerometer is being used?” Proudly, Frank reported a Wilcox model 766 self-amplifying accelerometer (that includes a seismic mass, piezoelectric transducer and an integral charge-to-voltage converter). It is mounted to the GCU. The vibration panel provides a shutdown signal if the GCU flies apart. One of the accelerometer pins ties its case and the case is connected to ground. I thought a bit more, and the shining light of truth appeared.
The Wilcox model 766 (see below image (a)) is in a stainless-steel case. It includes a seismic mass, piezoelectric sensor, integral electronics and a 3-pin connector (b). The basic electrical schematic of the accelerometer is provided in (c). There is an internal shield that is electrically isolated from the case. The integral 3-pin connector is detailed in (d) with its electrical connections provided in (e). The essence of the vibration panel interface to the accelerometer is given in (f). The vibration panel provides an analog meter of rms g’s as indicated in (g). The meter is for reference only.
Frank cut the vibration panel shield ground connection. All became quiet and the analog meter held at the expected value of zero. Frank and Ben were relieved. Although this was a very simple problem, I felt as though I was the Lone Ranger—give them each a silver bullet and ride my trusty steed Silver off into the sunset. I never heard another word after I left. Everything else evidently worked well. Our “Tonto” was shielded cable that gave us new reasons to process its subtle effects.
Lesson 3: Shielded cable combats common-mode noise, but the shield should only be connected to ground at one point. Ideally, the connection to ground should occur at the signal source. More than one ground connection will result in a ground loop.
---
Using shielded cable combats possible noise contamination, but its subtle effects must be recognized. It possesses a capacitive characteristic that must be neutralized or made negligible. Coaxial cable acts like a transmission line and needs to be terminated with its characteristic impedance to avoid ringing and standing waves. If a shielded cable is used, it should only be tied to ground at one point to avoid the possibility of ground loops.