1.2 Sensor Types and What Each Is For

After this section you can: choose IEPE, charge or piezoresistive for a stated test and say when a force gauge, strain gauge or laser instrument is the right channel.

Three kinds of accelerometer do almost all the work on a shaker, and each exists because the other two cannot do something. They differ in where the electronics are, in the lowest frequency they can read, and in the heat and the shock they will stand.

The everyday sensor is the IEPE accelerometer. The letters stand for integrated electronics piezoelectric: a piezoelectric element, a crystal or ceramic that makes a charge when it is squeezed, with a small amplifier built into the case beside it. The amplifier is powered by a constant current fed down the same two-conductor coaxial cable that brings the signal back, so one cable and one connector do both jobs. Sensor manufacturers sell the design under trade names of their own (ICP, Isotron, DeltaTron and CCLD among them), and a sensor sold under any of them is an IEPE sensor: it takes a constant current on its signal cable. One sensor manufacturer's published introduction to these sensors (PCB Piezotronics) gives the supply as 2 to 20 mA from a source of 18 to 30 V, and the output as a signal riding on a steady bias of 8 to 12 V, with a full-scale swing of about 5 V either side of the bias on a typical sensor. Several things follow from the design, and each matters on the floor.

The range is fixed. The amplifier's gain was set when the sensor was made, so the sensitivity is fixed, and the full-scale output divided by the sensitivity is the range. A sensor of 100 mV/g reaches 5 V at 50 g; one of 10 mV/g reaches it at 500 g. That is arithmetic, and it is the first check on any IEPE sensor: the range comes from the sensitivity, not from a setting.

The low-frequency limit is fixed too. The built-in amplifier lets a steady charge bleed away through its own time constant, the discharge time constant, set when the sensor was made, and below a frequency set by that time constant the sensor reads low. The chain's own coupling can set a higher limit than the sensor's, and the shortest time constant anywhere in the chain is the one that governs. 1.5 has the arithmetic, and it decides whether an IEPE sensor can hold a long-stroke sine at the bottom of its band.

The temperature limit is the electronics', not the crystal's. It varies by model, and the datasheet for the sensor in hand is the only number that counts.

And the bias is the first diagnostic. Every IEPE input reads the steady voltage the sensor sits at, before any motion. A bias inside its window says the sensor is powered and its amplifier is alive. A bias near the supply voltage says the circuit is open: a cable unplugged, broken or never connected. A bias near zero says it is shorted. The controller watches it (4.2), and 1.8 reads it as the first line of the field guide. What the bias cannot say is whether the sensitivity is right, whether the sensor is mounted, or whether it is on the right axis.

The IEPE sensor is cheap to run, it drives a long cable well within limits (1.5), and its output is a low-impedance voltage that is hard to disturb. It is the right sensor for most of the channels on most shaker tests, which is why it is in every drawer.

The charge accelerometer is the same piezoelectric element with no electronics in the case. Its output is charge, in picocoulombs per g, and an external charge amplifier turns it into a voltage. It is still used for two reasons. With nothing in the case to cook, it can go far hotter than an IEPE sensor, and for the hottest chamber tests it is often the only choice; the limit is again the datasheet's. And because the gain lives in the charge amplifier, the range can be set for the test rather than having been set at manufacture. The cost is the cable. The line between the sensor and the charge amplifier carries a tiny charge at very high impedance, so it has to be a low-noise cable, made to keep the triboelectric effect (the charge a cable makes when its layers rub as it flexes) out of the signal; it has to be short and held still; and its connectors have to be clean and dry, because any leakage across them bleeds the charge away and takes the low frequencies with it. Where the run to the rack is long, an in-line charge converter near the sensor turns the charge into an IEPE-type output for the rest of the way, and the converter, which has electronics, stays outside the chamber.

The piezoresistive accelerometer works differently. A tiny mass on a flexure carries a bridge of resistors that change with strain, so the sensor reads acceleration down to a steady value, zero hertz. It needs a bridge excitation voltage and an input coupled to accept the steady part of the signal, and its sensitivity is lower than a piezoelectric sensor's. It earns its place in shock. MIL-STD-810H Method 516.8, the shock method, in paragraph 4.4, requires a piezoresistive accelerometer measurement where response below 2 Hz is wanted, and the reason is the velocity change. A shock pulse's velocity change and displacement live in its lowest frequencies, and a piezoelectric chain, which cannot hold a steady level, gets them wrong on a long pulse. A severe shock can also leave a piezoelectric element with an offset in its output that is not motion, called zero shift (1.8). That is mainly a failing of piezoelectric elements, and a piezoresistive sensor is the usual answer to it; some are built with internal damping so their own resonance does not ring under a sharp pulse. Its steady response gives a free field check, too: turned over, it should read 2 g of change, plus one g one way up and minus one g the other.

A fourth kind, the capacitive accelerometer made by micromachining, also reads to zero hertz, with low noise at low frequencies and a modest upper limit. On a shaker it is a specialist's sensor, for long-stroke tests at a few hertz where its strengths are the test. It is not a peer of the other three for general work.

How the element is held inside the case matters as well. In a compression design the element is stacked under a preload against the base, so anything that squeezes the stack, motion or not, makes output. If the surface under the sensor bends, the base bends with it and the sensor reads the strain as acceleration; this is base-strain sensitivity, and MIL-STD-810H Method 514.8, the vibration method, says in paragraph 4.2.3 that it should be evaluated. A compression stack also responds to a temperature change in its base, with a slow drift that is not motion. In a shear design the element is loaded sideways, isolated from the base, and both effects are much smaller. On a fixture plate that bends in band, or in a chamber whose temperature is moving, the author reaches for a shear design, and the datasheet says which one a sensor is.

A triaxial accelerometer carries three elements in one case, one for each axis, with the axes marked on the case. It is three channels, with three sensitivities on its certificate and three entries in the channel table, and the most common fault with one is an axis assigned to the wrong channel. Every sensor responds a little to motion across its measuring axis, and the specification for this is the transverse sensitivity, given as a percentage of the main sensitivity. Method 514.8 asks for 5 percent or less in paragraph 4.2.3, and Method 516.8 asks for 7 percent or less for shock in paragraph 4.4. The triaxial sensor is the usual cross-axis monitor at the article's mounting face, and where it is mounted on a single stud its axes are set by how far the stud turned it. Its orientation is recorded against a feature of the article and photographed close enough to read, as every sensor on the setup is.

Accelerometers are not the only channels a test is handed. Three others turn up often enough that the reader will set one up sooner or later.

A force transducer, a piezoelectric element that reads force, goes between the fixture and the article to measure what the interface carries. It is the channel force limiting runs on: the controller limits the interface force at the article's resonances to what its service mounting could deliver, as 4.1 describes in a paragraph, and the method is published in NASA-HDBK-7004, NASA's handbook on force-limited vibration testing. Force gauges carry the article's load and its preload, so they are designed into the fixture, calibrated as installed, and planned long before the test.

A strain gauge measures stretch at a point, which is often closer to what breaks than acceleration is. It goes on the article or the fixture where an analyst cares about stress, usually as a monitor channel, and it needs its own bridge completion and excitation, as the piezoresistive sensor does.

Non-contact instruments add no mass at all. A laser vibrometer reads the velocity of a spot on the surface from across the room, which suits an article too light to carry a sensor and a surface nobody may touch. A displacement transducer, either the linear variable differential transformer (LVDT) on the armature (the shaker's moving element, which carries the table) or a proximity probe aimed at it, reads stroke directly, which is how a lab checks the shaker's displacement against its rating on a long-stroke test. A laser displacement gauge does the same from a stand-off without touching anything: it triangulates the distance to a spot on the table, the fixture or the shaker body and reads the displacement directly. It costs a fraction of a laser vibrometer and its bandwidth ends at a few hundred hertz, which is exactly the band where displacement lives, so it earns its place on the shelf for two jobs: confirming a predicted shock displacement on the calibration load before the article sees the pulse (Chapter 8), and reading the body's own motion at the low end of a long-stroke test, which no accelerometer on the table can see.

Two more non-contact methods turn up in the better-equipped labs and are named here so the reader recognizes them. High-speed video of the article under test, read frame by frame, shows what moved and when, and with a speckle pattern on the surface it becomes digital image correlation (DIC): the software tracks the pattern between frames and returns displacement and strain over the whole surface, not at one point. Scanning laser vibrometers do the same for velocity, point by point. These are analysis tools with their own setup, lighting, calibration and processing, and a program that needs them brings the people who run them. This book keeps to the traditional chain, the accelerometer and its relatives, because that is what every test is controlled and judged on; where a full-field method would answer a question the accelerometers cannot, the plan names the method and the analyst owns it. MIL-STD-810H names strain gauges, laser instruments and force gauges beside accelerometers among the devices a test may use, in Method 514.8 paragraph 4.2.3 and Method 516.8 paragraph 4.4, the same paragraphs that set the accelerometer requirements.

Pyroshock, the very high-frequency, very high-level shock from explosive devices and the machines that imitate them, has its own instrumentation problems, and Chapter 8 takes them.