Instrumentation

After this chapter you can name every link from the motion to the number on the screen and say how each can lie; choose between IEPE, charge and piezoresistive sensors for a stated test; read a datasheet and say which specification limits this test; prepare a surface, mount a sensor, torque it and state its usable frequency limit; cable and power it for the run length; place, mount and back up the control accelerometer; read a calibration certificate and run the field check; and recognize a lying sensor in the time history and the spectrum.

An article came off the shaker after its random test in the second axis with a question hanging over it. The low-level signature runs before and after the full-level run (short runs at a small fraction of the test level, made the same way both times so that their responses can be laid over each other to show whether the article changed) disagreed on one response channel. A peak near the top of the band had moved down by about a sixth of its frequency, and it had grown taller and wider, and the floor of the spectrum above it had risen. A resonance that drops after a full-level run is what a crack looks like, or a joint that has lost its preload, and that is how the analyst reported it. The program stopped the series. The article went back to its owner's shop, came apart over two days and went through inspection one part at a time. Nothing was cracked, and every fastener broke away at its torque.

The answer had been on the fixture the whole time. When the technician went to take the response accelerometers off, the one on that channel came away in his fingers. The adhesive under it had cracked across, and the sensor had ridden the second half of the run on a broken glue line and rattled against it at the peaks. The resonance that moved was the sensor's own, on a mount that had gone soft and pulled it down into the band, and the raised floor was the rattle. Two other channels saw the same mode of the article, and on both of them it sat exactly where it had sat before the run. Nobody had laid the three side by side before the article was sent away.

One comparison would have saved the teardown. A resonance of the structure moves on every channel that sees that mode; a resonance that moves on one channel alone is the sensor until the sensor has been cleared. This chapter is about the instruments that make every number in a test, and about knowing when one of them is lying.

It is also the first chapter of Volume 2. Volume 1 gave the machine: the shaker and its performance envelope, the amplifier and its headroom, the facility, the fixture and its survey, and the machine's side of the plan, handed to the test on one signed page. This volume gives the rest: the measurement, the control loop, the tests, the plan, the data, the run and the report. The two together are what it takes to start testing. What this volume takes from Volume 1 it repeats in a line where a chapter needs it, the sine relations and the armature resonance from its shaker chapter and the fixture's accelerometer mounts and its survey from its fixtures chapter most of all. Multi-axis testing, and the multiple-input multiple-output control that runs it, is Volume 3's subject, for the reader who already runs tests. Each volume stands alone.

The volume starts with the measurement because every later chapter is read through it. A sine test, a random test and a shock test are each a statement about motion at a point, and the statement is only as good as the chain that made it. This chapter takes that chain apart from the motion to the screen (1.1); then the sensor types and what each is for (1.2); the datasheet, and which line on it limits the test in hand (1.3); mounting (1.4); cables, conditioning and grounding (1.5); the control accelerometer and the backup it is owed (1.6); calibration and the field check (1.7); and the signatures of a sensor that lies, each with its five-minute check (1.8). Chapter 2 takes the chain from the conditioner onward: sampling, ranging, the channel table, and the end-to-end check as a procedure.