Even during routine operation of an instrumentation amplifier, a fault can develop between the sensor and the amp. This fault could result from misuse, the environment, or low-quality assembly, among other reasons. Discrete approaches exist to detect these faults, but these approaches can affect system performance.
A variety of failures can occur between the sensor and the instrumentation amplifier (Fig. 1). At point A, potential failures include an open between the supply and the bridge or a deteriorating connection that’s causing a resistance between the bridge and the supply. The same failures are possible at point B between the bridge and ground. Points C and D may suffer faults such as an open between the bridge and amplifier; a deteriorating connection that causes a resistance between the bridge and the amplifier; a short to the supply; or a short to ground. Also, a short can occur between points C and D.
All of these faults will no doubt affect the performance of the system. For example, in normal use, if the bridge isn’t being strained, the amplifier’s inputs will be at V+/2. If the inputs were shorted at points C and D, the inputs will still be at V+/2.
However, when the bridge is being strained, no voltage differential will be present at the amplifier’s inputs. Each input will stay at V+/2. This fault can be sensed by injecting a small current at point C and measuring the voltage drop. If there’s no short, the amplifier will measure the voltage drop across the resistors in the bridge. If a short occurs at points C and D, the voltage drop will be very small.
Three different methods can be used to detect each of these faults:
• Measure the voltage at the input pin of the instrumentation amplifier (not the differential voltage at the instrumentation amplifier inputs).
• Inject a current and measure the resulting voltage at the input pin of the instrumentation amplifier.
• Inject a current and measure the voltage differential at the instrumentation amplifier input pins.
Table 1 lists the potential faults and the method that can be used to detect each type of failure. For example, if the connection at point B were open, both inputs of the instrumentation amplifier will be pulled to V+. By measuring one of the instrumentation-amplifier inputs, this failure can be detected by the fact that the input isn’t at the expected common-mode voltage of V+/2.
As noted earlier, it’s possible to detect these faults by using discrete circuitry on the input of the instrumentation amplifier. They can’t be detected by the instrumentation amplifier alone, because the amplifier typically sets up to measure the differential input voltage, not the individual voltage at each pin.
Several issues arise when using discrete circuitry:
• It will add parasitic leakages and capacitances, which will affect the instrumentation amplifier’s performance.
• A method is required to communicate with the fault circuitry and switch it in and out of the system as needed.
• If the instrumentation amplifier is used to detect the fault and the common mode of the fault goes outside of the instrumentation amplifier’s input common-mode voltage range, the reading will not be correct. As a result, the input must be brought back into the input common-mode voltage range of the amplifier.
Continue on next page
These detection problems can become rather thorny, but there are new solutions to help combat them. For instance, National Semiconductor’s LMP8358 instrumentation amplifier includes built-in fault-detection circuitry (Fig. 2). This circuitry doesn’t affect the performance of the part and includes a method of detecting a voltage when it’s outside of the input common-mode voltage range.
The LMP8358 consists of three parts: an instrumentation amplifier; a control block that allows the user to set the internal register (which sets the state of the amplifier); and a fault-detection block. The fault-detection block includes a current source for injecting current, a multiplexer between the external and internal input pins of the instrumentation amplifier, and a divide-by-50 block that can be inserted in the signal path.
The internal register controls the fault-detection features, sets the gain and compensation of the amplifier, and turns on the power-save and zero-calibration states. The internal register and an external master communicate via a serial peripheral interface (SPI) bus.
How Do The Fault-Detection Features Work?
If there’s a break at point B (Fig. 1, again), points C and D will be at V+. This lies outside of the LMP8358’s input common-mode voltage range (CMVR), which is specified as –0.1 V < CMVR < 3.6 V if using a +5-V supply.
If the LMP8358 internal register is configured with setting 5 of the multiplexer (Fig. 2, again), the input on the external +IN pin will be divided by 50, which will put the input back into the input CMVR of the internal instrumentation amplifier. The negative input of the internal instrumentation amplifier will be connected to the negative supply. The instrumentation amplifier is set to a gain of 50 and the output will show the voltage on the external +IN pin (Fig. 3).
Another example is an open at point C (Fig. 1, again). The input to the instrumentation amplifier will be floating, which would cause the output to float. This fault is found by inserting the divide-by-50 block into the signal path, setting the gain to 50, and then injecting a current on the input pin (Fig. 4).
As with the previous example, the divide-by-50 block will ensure that the input is within the instrumentation amplifier’s input common-mode voltage range. The current will cause the input to pull up to V+. The output will also be at V+.
Checking The System
Fault-detection circuitry also helps determine the state of the system. An obvious time to check the system is when it’s powered on. A routine can be written in the firmware that runs through the fault setups to detect any possible faults. The routine then compares the output of the LMP8358 to a table and subsequently indicates whether the system is ready to use or there’s a problem. If a problem exists, this routine makes the repair easier to handle because it indicates what type of error is present.
A table can be used to indicate if the system is working correctly or if there is a fault (Table 2). The first row in Table 2, labeled “No Fault,” shows the LMP8358 output using the eight test settings with no fault in the system. The first two settings are normal use settings, while the following six settings are fault-detection settings.
In the values shown, there’s a small offset from the bridge of about 4 mV. The 0000h setting multiplies this by 10 for an output of about 40 mV, while the 0003h setting multiplies it by 100. The rest of the first row shows typical values under the six fault tests.
The bolded values in the boxes in rows 2 to 14 show the values that will be seen at the LMP8358 output for that particular fault. For example, if there’s a fault of the +IN pin shorted to V+ (Row 7), the output will be at V+ when the register of the LMP8358 is set to 0082h.
At startup, a routine can run through these eight tests. If the output values are close to the values in the first row, there’s no fault. If one of the values in bold in rows 2 to 14 are seen at the output, then a fault is present. End users also can use this test through a self-test option.
The LMP8358’s on-chip fault-detection features simplify failure mode and effects analysis (FMEA) during product design. Moreover, with fault detection, system designers can offer more value to the end user, because they can ensure that equipment is operating correctly with no faults in the system. If a fault does happen to occur, the repair technician has important information to help correct the fault, which will shorten the amount of time needed to repair the equipment.