Question 2 of 7: Signal Conditioning and Transmission

Question 2: Signal Conditioning and Transmission (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

a) The first stage sees the raw sensor signal before any gain has been applied, so its own noise and offset add directly to that small signal at full weight. Any noise or offset contributed by a later stage is effectively divided by the gain of every stage ahead of it when referred back to the input, so it contributes far less to the total input-referred error. The front-end amplifier's noise and offset therefore set the practical noise/offset floor of the whole instrument, however good the later stages are.

b) Loop current in a series 4-20 mA loop is the same at every point around the loop regardless of wire resistance or voltage drop along the run, so voltage picked up along the cable (IR drop, ground-loop offsets, induced noise) does not change the current that the far-end receiver measures. A differential voltage signal, by contrast, is degraded by any resistance or induced voltage in the wire path itself. The current loop also has a "live zero" (4 mA, not 0 mA, represents the true zero reading), so a broken wire or dead transmitter (0 mA) is immediately distinguishable from a legitimate zero reading — a fault-detection feature a voltage signal does not have.

c) Key specifications include input bias current and input offset voltage/drift (critical when the source impedance is high), open-loop gain and gain-bandwidth product (sets usable bandwidth at the required closed-loop gain), common mode rejection ratio (CMRR), slew rate (limits large-signal bandwidth), input impedance, and input-referred noise voltage/current density. Which of these dominates the choice depends on the sensor's own source impedance, the signal bandwidth required, and the precision needed.

d) Noise ultimately arises from fundamental physical processes — thermal (Johnson) noise in every resistance at any temperature above absolute zero, and shot noise from the discreteness of charge carriers — that are present in any real conductor or semiconductor device. These processes can be reduced (narrower bandwidth, lower temperature, averaging over more samples) but never driven to exactly zero, so some residual noise floor is always present in a physical measurement.

e) Drift is compensated by periodically re-zeroing/re-calibrating the instrument against a stable external reference, or by designing the measurement to be ratiometric — measuring the signal of interest against a reference channel that experiences the same drift mechanisms simultaneously, so that a shared drift term cancels in the ratio. Where the dominant drift mechanism is temperature, adding an independent temperature sensor and applying a computed correction (electronic compensation) removes most of the drift without needing a fresh calibration each time.

f) The three op-amp configuration gives very high input impedance at both $V_1$ and $V_2$ (they drive non-inverting inputs directly, drawing negligible current, so the source is not loaded), and it sets the overall gain with a single resistor $R_g$ without disturbing the closely-matched resistor network that gives the output stage its common-mode rejection. Because the input buffers isolate the sensor from the difference-amplifier stage, the configuration achieves high CMRR together with high input impedance and easily adjustable gain — a combination a single difference amplifier alone cannot provide, since a plain difference amp both loads the source and ties its gain and CMRR-critical resistor ratios together.

g) A $3\tfrac{1}{2}$ digit voltmeter displays three full digits (each 0-9) plus a "half" leading digit that can only ever show a 0 or a 1. This gives it a maximum reading of $\pm1999$ counts (i.e. $\pm1.999\times$ the nominal range decade) rather than the $\pm999$ counts a strict 3-digit meter would give, extending both the resolution and the usable overrange without the cost of a full fourth digit.

h) Every real op-amp input draws a small bias current that must have a continuous DC path to the common/ground reference; if that path does not exist, the bias current has nowhere to flow, the input node charges up uncontrolled, and the amplifier drifts into saturation. A high-impedance sensor such as a capacitive probe provides no DC path of its own, so a bias/bleed resistor must be added deliberately to give the bias current somewhere to go and keep the input biased within its working range.

i) The twisted pair carries the signal and its return as a balanced differential pair, which cancels magnetically-induced noise because both conductors enclose essentially the same loop area and pick up equal induced voltages that subtract out at the differential input. The surrounding shield intercepts electrostatically (capacitively) coupled interference and should be grounded at one end only (typically the source/sensor end) — grounding both ends creates a second ground path through the shield that can carry a ground-loop current and reintroduce noise.

j) An anti-aliasing filter is a low-pass filter placed ahead of an analog-to-digital converter, with a cutoff at or below the Nyquist frequency (half the sampling rate), whose job is to attenuate any input energy above that frequency before sampling occurs. Without it, out-of-band content folds back (aliases) into the sampled band and is permanently indistinguishable from a genuine lower-frequency signal once digitized.