ECE246: Sensors For Robotics — Unit 1: Measurements in Robotics

1. Introduction: The Perception Layer of Robotics

Autonomous and semi-autonomous robotic systems exist to interact meaningfully with the physical world. While actuators serve as the robotic "muscles" and microprocessors as the "brain," sensors provide the essential sensory organs. Without reliable sensing, a robot is functionally blind, deaf, and uncoordinated—incapable of closing feedback loops, adapting to environmental variations, or operating safely around human workers.

Robotic measurement bridges continuous physical reality (force, torque, photons, temperature, magnetic flux, acoustic pressure) and discrete cybernetic computation. A robust measurement architecture guarantees:

• Kinematic State Estimation: Real-time feedback of joint angles, angular velocities, and end-effector cartesian poses.
• Dynamic Interaction Control: Direct force/torque sensing for delicate assembly, compliance control, and haptic teleoperation.
• Environmental Mapping & Localization: SLAM (Simultaneous Localization and Mapping) using cameras, LiDAR, and ultrasonic rangefinders.
• Safety Interlocking: Human presence detection, collision avoidance, and thermal monitoring of high-power joint actuators.

CODECopy Code
+-----------------------------------------------------------------------------------+
|                        CLOSED-LOOP ROBOTIC SYSTEM                                 |
|                                                                                   |
|   +------------+    Command     +-----------+   Driving Torque    +-----------+   |
|   | Controller | -------------> | Actuators | ------------------> | Physical  |   |
|   | (DSP / MCU)|                | (Motors)  |                     | Robot Link|   |
|   +------------+                +-----------+                     +-----------+   |
|         ^                                                               |         |
|         |                     PERCEPTION FEEDBACK                       |         |
|         |                                                               v         |
|   +---------------------------------------------------------------------------+   |
|   |  Measurand ---> Primary Sensor ---> Conditioning ---> ADC ---> Bus Comms  |   |
|   +---------------------------------------------------------------------------+   |
+-----------------------------------------------------------------------------------+

2. Fundamental Definitions: Sensors, Transducers, and Instruments

Although frequently conflated in informal technical discussions, sensor, transducer, and instrument have distinct engineering definitions:

2.1 The Sensor

A sensor is the primary physical element that responds directly to a specific physical measurand ($x$) and converts it into an intermediate physical or electrical response ($y$).

• Example: A piezoresistive foil strain gauge bonded to an elastic cantilever beam changes electrical resistance ($\Delta R$) in response to mechanical strain ($\epsilon$).

2.2 The Transducer

A transducer is an integrated energy-conversion device that converts one form of energy into another—typically converting a non-electrical variable into an analog electrical signal (voltage, current, or charge). A complete transducer incorporates the primary sensing element and secondary conditioning circuitry.

• Example: A pressure transducer combines a silicon diaphragm (sensor), a Wheatstone bridge, and an on-board instrumentation amplifier to produce a standard $0 - 10\,\text{V}$ output proportional to gauge pressure.

2.3 Primary vs. Secondary Conversion

In complex robotic measurement chains, energy transformation occurs in distinct stages:

1. Primary Sensing: Direct interaction with the measurand (e.g., a Bourdon tube expands elastically under pneumatic pressure, converting pressure into mechanical linear displacement).
2. Secondary Conversion: Converting the intermediate mechanical variable into an electrical signal (e.g., an LVDT coupled to the tip of the Bourdon tube translates displacement into a differential AC voltage).

3. Energy Domain Classification: Active vs. Passive Transducers

Transducers are classified according to whether they generate their own electrical output or require external excitation power:

CODECopy Code
                            TRANSDUCER TAXONOMY
                                     |
             +-----------------------+-----------------------+
             |                                               |
     ACTIVE TRANSDUCERS                              PASSIVE TRANSDUCERS
  (Self-Generating Energy)                       (Modulate External Power)
             |                                               |
  - Piezoelectric Quartz                          - Strain Gauges (Piezoresistive)
  - Thermocouples (Seebeck)                       - RTDs (Resistance Temperature)
  - Photovoltaic Cells                            - LVDTs & RVDTs (Inductive)
  - Tachometer Generators                         - Capacitive Proximity Sensors
  - Pyroelectric IR Detectors                     - Hall Effect Sensors (IC Type)

3.1 Active Transducers (Self-Generating)

Active transducers operate on fundamental electro-physical principles where mechanical, thermal, or radiation energy is directly transformed into electrical voltage or charge. They require zero external power to generate signal energy.

1. Piezoelectric Transducers: Non-centrosymmetric crystalline lattices (e.g., quartz, PZT) generate surface electrical charge ($q = d \cdot F$) when subjected to mechanical stress. They exhibit infinite DC electrical impedance and cannot measure static steady-state force.
2. Thermoelectric Transducers (Thermocouples): Dissimilar metals joined at two junctions generate a Seebeck voltage ($\Delta V = \alpha_{AB} \Delta T$) proportional to the temperature differential between the hot sensing junction and cold reference junction.
3. Electromagnetic Induction Sensors: A moving conductor in a fixed magnetic field generates an electromotive force ($e = -N \frac{d\Phi}{dt} = B \cdot l \cdot v$), utilized in linear velocity transducers and analog DC tachometers.

3.2 Passive Transducers (Modulating)

Passive transducers do not generate electrical energy independently. Instead, they modulate an externally applied electrical parameter—such as resistance ($R$), capacitance ($C$), inductance ($L$), or magnetic reluctance—in response to the input measurand.

1. Piezoresistive Strain Gauges: Require a stable external DC excitation voltage ($V_{\text{ex}} \approx 5 - 10\,\text{V}$) applied across a Wheatstone bridge to convert fractional resistance changes ($\Delta R / R = G_f \cdot \epsilon$) into differential output voltage.
2. Linear Variable Differential Transformers (LVDT): Require high-frequency AC sinusoidal carrier excitation ($1 - 10\,\text{kHz}$) applied to the primary winding. The position of a magnetic core modulates the differential mutual inductance coupled to twin secondary coils.
3. Capacitive Displacement Sensors: Require high-frequency AC excitation to detect changes in capacitance ($C = \epsilon_0 \epsilon_r A / d$) caused by variations in plate separation distance ($d$) or dielectric constant.

4. Sensor Static Characteristics & Performance Metrics

When selecting sensors for robotic joints, end-effectors, or mobile bases, engineers analyze static characteristics—performance metrics evaluated under slow or steady-state input conditions.

CODECopy Code
  Output (y)
      ^                     / Ideal Linear Response (y = S*x + b)
      |                    /
      |      +------------/---------------------+
      |      |           /       Upper Boundary |
      |      |          /. . . .                |
      |      |         /  :    :                |
      |      |        /   :    :  Hysteresis    |
      |      |       /    :    :   Loop         |
      |      |      /     :    :                |
      |      |     /. . . :    :                |
      |      +----/-------:----+----------------+
      |          /        :    : Lower Boundary
      |         /         :    :
      |        /          :    :
      +-------+-----------+----+-------------------------> Input Measurand (x)
             Dead       Threshold
             Band

4.1 Sensitivity ($S$)

Sensitivity is the ratio of incremental change in output response ($\Delta y$) to the corresponding incremental change in input measurand ($\Delta x$):
$S = \lim_{\Delta x \to 0} \frac{\Delta y}{\Delta x} = \frac{dy}{dx}$

• For an ideally linear sensor, sensitivity is constant throughout the operational range ($y = S \cdot x + b$).
• For non-linear sensors (e.g., thermistors obeying the Steinhart-Hart equation), sensitivity is an operating-point dependent derivative ($S(x)$).

4.2 Range and Span

• Range: The lower and upper limits of the input measurand over which the sensor meets its specified calibration accuracy ($[x_{\min}, x_{\max}]$). For example, an IMU gyroscope rated for $-2000^\circ/\text{s} \le \omega \le +2000^\circ/\text{s}$.
• Span (Full-Scale Input, FSI): The algebraic difference between maximum and minimum operational limits:
  $\text{Span} = x_{\max} - x_{\min}$

4.3 Linearity and Linearity Error

No physical sensor is perfectly linear. Non-linearity is quantified as the maximum deviation ($\Delta y_{\max}$) of the actual calibration curve from a specified ideal reference line, expressed as a percentage of Full-Scale Output (% FSO):
$\text{Linearity Error (\%)} = \left( \frac{\Delta y_{\max}}{\text{FSO}} \right) \times 100\%$
Common reference line formulations include:

1. End-Point Linearity: Straight line connecting measured calibration points at $x_{\min}$ and $x_{\max}$.
2. Best-Fit Straight Line (BFSL): Linear regression line determined using the method of least squares across all calibration data points, minimizing residual variance.
3. Independent Linearity: Best-fit line positioned to minimize maximum absolute deviation.

4.4 Hysteresis

Hysteresis describes the phenomenon where a sensor produces different output readings for identical input values, depending on the direction of approach (increasing vs. decreasing measurand).
$\text{Hysteresis Error (\% FSO)} = \left( \frac{\max |y_{\text{upscale}} - y_{\text{downscale}}|}{\text{FSO}} \right) \times 100\%$

• Physical Root Causes: Magnetic domain pinning in ferromagnetic cores, elastic plastic deformation in metal spring flexures, and mechanical gear lash in geared couplings.

4.5 Resolution and Threshold

• Resolution: The smallest incremental change in input measurand ($\Delta x_{\min}$) capable of producing a discernible, measurable change in output signal. For a digital sensor with an $n$-bit ADC over span $X$, theoretical resolution is $q = X / 2^n$.
• Threshold: The minimum input magnitude required to evoke any detectable non-zero output starting from zero.

4.6 Repeatability and Reproducibility

• Repeatability: The degree of agreement among successive output measurements under identical operating conditions, identical operator, same environment, and short time intervals.
• Reproducibility: Agreement between measurements taken under varying conditions (different test benches, different operators, across extended time horizons).

5. Sensor Dynamic Characteristics

When robots execute high-speed maneuvers or contact dynamic surfaces, sensor inputs fluctuate rapidly. The time-domain and frequency-domain behavior is governed by ordinary differential equations.

5.1 Zero-Order Sensors

Output tracks input instantaneously with no dynamic lag or energy storage:
$y(t) = K \cdot x(t)$

• Examples: Precision conductive plastic potentiometers, resistive strain gauge bridges (operating below mechanical resonant frequencies).

5.2 First-Order Sensors

Incorporate a single energy storage element (thermal mass, fluid capacitance, or electrical capacitance):
$\tau \frac{dy(t)}{dt} + y(t) = K \cdot x(t)$
Where:

• $\tau$: Time constant of the sensor ($\text{seconds}$).
• $K$: Static steady-state sensitivity.

Step Response of a First-Order System:

When subjected to a step input of magnitude $X_0$ at $t = 0$:
$y(t) = K \cdot X_0 \left( 1 - e^{-t / \tau} \right)$

• At $t = \tau$: $y(\tau) \approx 63.2\%$ of final steady-state value.
• At $t = 3\tau$: $y(3\tau) \approx 95.0\%$.
• At $t = 5\tau$: $y(5\tau) \approx 99.3\%$ (settled).
• Cutoff Frequency ($-3\,\text{dB}$ Bandwidth):
  $f_c = \frac{1}{2\pi \tau} \quad (\text{Hz})$

5.3 Second-Order Sensors

Incorporate two coupled energy storage elements (inertial mass and compliance/spring):
$\frac{1}{\omega_n^2} \frac{d^2 y(t)}{dt^2} + \frac{2\zeta}{\omega_n} \frac{dy(t)}{dt} + y(t) = K \cdot x(t)$
Where:

• $\omega_n$: Natural undamped angular frequency ($\text{rad/s}$).
• $\zeta$: Damping ratio (dimensionless).
• Examples: Piezoelectric accelerometers, force/torque load cells, pressure sensors.

Damping Regimes:

1. Underdamped ($\zeta < 1$): Rapid rise time, but exhibits overshoot, ringing, and resonant peak at $\omega_r = \omega_n \sqrt{1 - 2\zeta^2}$.
2. Critically Damped ($\zeta = 1$): Fastest transient response without oscillatory overshoot.
3. Overdamped ($\zeta > 1$): Sluggish, non-oscillatory rise to steady-state.

• Optimum Instrumentation Damping: In robotic sensing, $\zeta \approx 0.707$ ($1/\sqrt{2}$) is preferred, providing the flattest amplitude response curve and linear phase shift over the widest operational bandwidth.

6. Robotic Measurement Errors: Taxonomy & Statistical Propagation

Any real robotic measurement $x_m$ deviates from the accepted true value $x_t$. The absolute error is defined as:
$e = x_m - x_t$

6.1 Systematic Errors (Determinate)

Systematic errors are constant or predictably repeatable offsets in one direction across repeated measurements.

• Root Causes: Calibration drift, uncompensated temperature offsets, zero-point errors, incorrect amplifier gain settings, and constant gravitational preloading on force sensors.
• Remedy: Systematic errors can be identified and mathematically eliminated through rigorous calibration, sensor zeroing (taring), and dynamic environmental compensation algorithms.

6.2 Random Errors (Indeterminate)

Random errors are unpredictable, stochastic fluctuations caused by inherent physical noise: thermal Johnson-Nyquist noise in resistors, electromagnetic interference, and atmospheric turbulence in optical paths.

• Statistical Representation: Random errors follow a Gaussian (normal) distribution characterized by mean $\mu$ and standard deviation $\sigma$:
  $p(x) = \frac{1}{\sigma \sqrt{2\pi}} \exp\left( -\frac{(x - \mu)^2}{2\sigma^2} \right)$
• Remedy: Cannot be eliminated by static calibration. Suppressed by sample averaging, low-pass filtering, or Kalman filtering. For $N$ independent measurements, the standard error of the mean shrinks by $\sqrt{N}$:
  $\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{N}}$

6.3 Gross Errors (Blunders)

Severe anomalies arising from catastrophic hardware faults, broken cables, disconnected grounds, bit slips in serial buses, or algorithmic bugs.

• Remedy: Detected and rejected using statistical thresholding (outlier rejection, chi-squared innovation tests in Kalman filters, and median filtering).

6.4 Error Propagation Law

When a calculated robotic state $Z$ depends on multiple independently measured variables ($X_1, X_2, \dots, X_n$) via a functional relationship $Z = f(X_1, X_2, \dots, X_n)$, the combined standard uncertainty $\sigma_Z$ is given by Taylor expansion:
$\sigma_Z = \sqrt{ \sum_{i=1}^n \left( \frac{\partial f}{\partial X_i} \right)^2 \sigma_{X_i}^2 }$

• Example: For a robot end-effector position calculated from link length $L$ and joint angle $\theta$ as $x = L \cos\theta$:
  $\sigma_x^2 = (\cos\theta)^2 \sigma_L^2 + (-L \sin\theta)^2 \sigma_\theta^2$

7. Smart Sensors: Architecture and Embedded Intelligence

A traditional sensor outputs raw analog millivolts vulnerable to external noise. Modern robotics relies on Smart Sensors—integrated measurement nodes that combine sensing, conditioning, computation, and digital fieldbus communication on a single substrate or module.

CODECopy Code
+-----------------------------------------------------------------------------------------+
|                                SMART SENSOR NODE ARCHITECTURE                           |
|                                                                                         |
|  +---------+   Analog    +--------------+   Conditioned   +--------+                    |
|  | Sensing | ----------> | Signal Cond. | --------------> |  ADC   |                    |
|  | Element | (mV/Charge) | (Amp/Filter) |   (0 - 3.3V)    | (Core) |                    |
|  +---------+             +--------------+                 +--------+                    |
|       ^                         ^                             |                         |
|       | Excitation Control      | Gain/Offset Trimming        | Digital Words           |
|       +-------------------------+--------------------+        v                         |
|                                                      |   +---------------+              |
|                                                      +-- | Microcontroller|             |
|                                                          | (DSP / Cortex)|              |
|                                                          +---------------+              |
|                                                              |   ^                      |
|                                   +--------------------------+   | Calibration LUT,     |
|                                   v                              v TEDS (IEEE 1451)     |
|                            +--------------+               +---------------+             |
|                            | Transceiver  |               |  EEPROM /     |             |
|                            | (CAN/SPI/I2C)|               |  Flash Mem    |             |
|                            +--------------+               +---------------+             |
|                                   |                                                     |
+-----------------------------------|-----------------------------------------------------+
                                    v Digital Bus to Robot Controller

7.1 Architectural Building Blocks:

1. Primary Sensing Core: MEMS structure (accelerometer, gyroscope, pressure cell) or silicon photodiode array.
2. Analog Front-End (AFE): Programmable-gain amplifiers (PGA), anti-aliasing low-pass filters, and constant-current/voltage bridge excitation sources.
3. Analog-to-Digital Converter (ADC): Successive Approximation (SAR) or Delta-Sigma ($\Delta\Sigma$) converter translating continuous voltages into $12$- to $24$-bit digital words.
4. Embedded Microcontroller / DSP: Executes embedded digital signal processing:
   • Non-Linearity Correction: Evaluates polynomial calibration curves ($y = c_0 + c_1 x + c_2 x^2 + c_3 x^3$).
   • Thermal Compensation: Reads internal temperature diode and offsets thermal drift in real time.
   • Digital Filtering: Running FIR/IIR low-pass filtering, outlier trimming.
   • Self-Diagnostics: Continuous integrity monitoring (broken wire detection, supply rail monitoring, saturation alerts).
5. Transducer Electronic Data Sheet (TEDS - IEEE 1451): Non-volatile EEPROM storing unique calibration coefficients, serial number, unit of measurement, and calibration date. The robot controller queries TEDS over the bus, achieving true plug-and-play installation.
6. Communication Transceiver: Industrial communication protocol interfaces:
   • Chip-Level: SPI (up to $50\,\text{MHz}$), $\text{I}^2\text{C}$ (Standard $100\,\text{kHz}$, Fast $400\,\text{kHz}$).
   • Robot Bus Level: CANopen / CAN FD ($1 - 5\,\text{Mbps}$), EtherCAT ($100\,\text{Mbps}$ deterministic real-time), RS-485 Modbus RTU.

8. Precision Calibration Techniques

Calibration is the formal experimental process of establishing the exact transfer function between a known, traceable reference standard input ($x_{\text{ref}}$) and the sensor's indicated output ($y_m$).

CODECopy Code
Calibration Workflow:
[Primary Standard] ---> [Apply Known Stimulus] ---> [Record Sensor Response]
                                                            |
[Compute Error Vector] <--- [Fit Regression Curve] <--------+
          |
[Generate Correction LUT / Coeffs] ---> [Flash to Sensor EEPROM (TEDS)]

8.1 Static Multi-Point Calibration

• The sensor is subjected to a series of known, stable stimuli across its entire operational span (e.g., $0\%, 20\%, 40\%, 60\%, 80\%, 100\%$ and back to $0\%$ to measure hysteresis).
• Standard least-squares linear regression evaluates optimal slope (gain $a$) and zero intercept (offset $b$):
  $y = a x + b$
  $a = \frac{N \sum x y - \sum x \sum y}{N \sum x^2 - (\sum x)^2}, \quad b = \frac{\sum y - a \sum x}{N}$

8.2 Two-Point Calibration (Offset and Span)

A streamlined routine used in production test benches:

1. Apply zero stimulus ($x = 0$): Measured output $y_0 = b$.
2. Apply known full-scale stimulus ($x = x_{\text{FS}}$): Measured output $y_{\text{FS}}$.
3. Compute calibrated gain:
   $a = \frac{y_{\text{FS}} - y_0}{x_{\text{FS}}}$

8.3 Tare / Zero-Point Calibration

In robotic end-effectors, tooling attachments (grippers, welders, rotary drills) exert static gravitational load on force/torque sensors.

• Taring samples sensor output while the manipulator is held motionless in a specific configuration, storing the baseline offset vector $\mathbf{y}_{\text{tare}}$.
• Subsequent dynamic measurements subtract this vector: $\mathbf{y}_{\text{net}} = \mathbf{y}_{\text{measured}} - \mathbf{y}_{\text{tare}}$.

8.4 In-Situ & Automated Self-Calibration

• In-Situ Calibration: Performed while the sensor remains installed on the robot arm, preventing mechanical mounting strain and thermal coupling variations that invalidate benchtop calibrations.
• Cross-Calibration / Dual Sensor Consistency: Comparing readings between redundant sensors (e.g., dual joint encoders or IMU + wheel odometry) to detect sensor drift dynamically during operation.

9. Analog Signal Conditioning & Instrumentation Circuitry

Raw sensor signals are often weak (microvolts to millivolts) and high-impedance, requiring conditioning before digitization.

9.1 The Wheatstone Bridge

Used to measure fractional resistance variations in resistive sensors (strain gauges, piezoresistive pressure diaphragms, magnetoresistors).

CODECopy Code
              V_in (Excitation Voltage)
                        |
            +-----------+-----------+
            |                       |
           [R1]                    [R2]
            |                       |
            +-----(+) V_out (-)----+
            |                       |
           [R3]                    [R4]
            |                       |
            +-----------+-----------+
                        |
                       GND

For a bridge excited by $V_{\text{in}}$:
$V_{\text{out}} = V_{\text{in}} \left( \frac{R_3}{R_1 + R_3} - \frac{R_4}{R_2 + R_4} \right)$

• Balanced Condition: $V_{\text{out}} = 0 \iff R_1 R_4 = R_2 R_3$.
• Quarter Bridge (Single Active Gauge $R_4 = R + \Delta R$):
  $V_{\text{out}} \approx \frac{V_{\text{in}}}{4} \left( \frac{\Delta R}{R} \right) = \frac{V_{\text{in}}}{4} G_f \epsilon$
• Full Bridge (4 active strain gauges in push-pull):
  $V_{\text{out}} = V_{\text{in}} \left( \frac{\Delta R}{R} \right) = V_{\text{in}} G_f \epsilon$
  Full bridges quadruple sensitivity, eliminate lead-wire resistance errors, and cancel thermal expansion strain automatically.

9.2 The Instrumentation Amplifier (In-Amp)

A classic three-op-amp topology providing ultra-high input impedance, high differential gain, and exceptional Common-Mode Rejection Ratio (CMRR).

CODECopy Code
         V1 o-------|+\
                    |  >----+
                 +--|-/     |
                 |          |
                [R1]       [R2]
                 |          |
                [Rg]        +---[R3]---+
                 |          |          |
                [R1]       [R2]        +---\-
                 |          |               >---o V_out
                 +--|+\     |          +---/ +
                    |  >----+          |
         V2 o-------|-/        +---[R3]---+
                               |          |
                              [R3]       GND
                               |
                              GND

The differential output voltage is:
$V_{\text{out}} = \left( 1 + \frac{2 R_1}{R_g} \right) \left( \frac{R_3}{R_2} \right) (V_2 - V_1)$

• Setting $R_2 = R_3$:
  $\text{Differential Gain } A_d = 1 + \frac{2 R_1}{R_g}$
• Changing a single external resistor $R_g$ programs the overall gain from $1$ to $10,000$.
• CMRR: Rejects 50/60 Hz mains hum and motor drive common-mode switching spikes induced equally on both differential lines ($> 100\,\text{dB}$).

10. Noise Sources, Interference Mitigation & Filtering

10.1 Dominant Electrical Noise Sources in Robotics

1. Thermal (Johnson-Nyquist) Noise: Stochastic voltage across any resistive component caused by thermal agitation of electrons:
   $v_n = \sqrt{4 k_B T R \Delta f}$
   Where $k_B$ is Boltzmann's constant ($1.38 \times 10^{-23}\,\text{J/K}$), $T$ is absolute temperature ($\text{K}$), $R$ is resistance ($\Omega$), and $\Delta f$ is measurement bandwidth ($\text{Hz}$).
2. Shot Noise: Current fluctuations due to the discrete quantum nature of electron charge ($q_e = 1.602 \times 10^{-19}\,\text{C}$):
   $i_n = \sqrt{2 q_e I_{DC} \Delta f}$
3. Electromagnetic Interference (EMI) & Ground Loops: PWM motor switching in robotic drives (switching $400\,\text{V}$ at $20\,\text{kHz}$) induces massive $dv/dt$ and $di/dt$ capacitive/inductive coupling into adjacent sensor signal wires.

10.2 Noise Mitigation Best Practices

• Shielded Twisted-Pair (STP) Cabling: Cancels magnetically induced noise loops.
• Single-Point Earth Grounding (Star Topology): Eliminates ground loops where circulating ground currents induce spurious millivolt drops.
• Galvanic Isolation: Optocouplers, digital capacitive isolators, or DC-DC isolated power supplies sever ground loops between high-power motor drivers and low-voltage MCU ADCs.

10.3 Analog and Digital Filters

1. Analog Anti-Aliasing Filter: By the Nyquist-Shannon Sampling Theorem, the sampling frequency $f_s$ must satisfy $f_s \ge 2 f_{\max}$. Any signal component above $f_s / 2$ folds back into the passband as an indistinguishable false low-frequency alias. A continuous low-pass filter (Butterworth or Bessel) precedes the ADC to attenuate frequencies above $f_s / 2$.
2. Digital Moving Average Filter: Replaces raw sample $x[k]$ with the arithmetic mean of the last $N$ samples:
   $y[k] = \frac{1}{N} \sum_{i=0}^{N-1} x[k-i]$
   Eliminates white Gaussian noise with low computational overhead; however, it introduces an inherent group delay of $D = \frac{N-1}{2}$ samples.
3. Digital Median Filter: Sorts $N$ samples in a sliding window and selects the central value. Highly effective at rejecting non-Gaussian impulsive spike noise (e.g., optical encoder glitches or ultrasonic specular reflections) without blurring sharp step transitions.

11. Solved Numerical Problems

Problem 1: ADC Quantization and Resolution

A 16-bit successive-approximation ADC has an input voltage span of $0\,\text{V}$ to $3.3\,\text{V}$.

1. Calculate the voltage weight of the Least Significant Bit (LSB / quantization step $q$).
2. If this ADC digitizes a load cell with sensitivity $S = 2.0\,\text{mV/N}$ connected via an amplifier with gain $A_v = 50$, determine the minimum detectable force resolution ($\Delta F$).

Solution:

1. Quantization step $q$:
   $q = \frac{V_{\text{ref}}}{2^n} = \frac{3.3\,\text{V}}{2^{16}} = \frac{3.3}{65,536} \approx 50.354\,\mu\text{V}$
2. The effective system sensitivity at the ADC input is:
   $S_{\text{eff}} = S_{\text{sensor}} \times A_v = 2.0\,\text{mV/N} \times 50 = 100.0\,\text{mV/N} = 0.1\,\text{V/N}$
3. Minimum detectable force resolution:
   $\Delta F = \frac{q}{S_{\text{eff}}} = \frac{50.354 \times 10^{-6}\,\text{V}}{0.1\,\text{V/N}} \approx 5.035 \times 10^{-4}\,\text{N} = 0.504\,\text{mN}$

Problem 2: Thermal Noise Calculation

A resistive temperature detector (RTD) has an internal resistance of $R = 1000\,\Omega$ operating at $T = 300\,\text{K}$. An instrumentation front-end samples this sensor over an effective noise bandwidth of $\Delta f = 10\,\text{kHz}$. Determine the RMS Johnson-Nyquist thermal noise voltage.

Solution:
$v_n = \sqrt{4 k_B T R \Delta f}$
Substitute known constants:

• $k_B = 1.38 \times 10^{-23}\,\text{J/K}$
• $T = 300\,\text{K}$
• $R = 1000\,\Omega$
• $\Delta f = 10,000\,\text{Hz}$

$v_n = \sqrt{4 \times (1.38 \times 10^{-23}) \times 300 \times 1000 \times 10^4}$
$v_n = \sqrt{1.656 \times 10^{-13}} \approx 4.069 \times 10^{-7}\,\text{V} = 0.407\,\mu\text{V}_{\text{RMS}}$
Insight: Because thermal noise is only $0.4\,\mu\text{V}$, it is well below the ADC LSB ($50.4\,\mu\text{V}$), confirming thermal noise will not dominate the measurement chain.

12. Common Engineering Pitfalls & Exam Summary

Common Pitfalls:

1. Confusing Resolution with Accuracy: A 24-bit ADC yields high numerical resolution, but if systematic calibration offsets or thermal drift are large, the measurement remains inaccurate. High resolution does not imply high accuracy.
2. Ignoring Sensor Loading Effects: Connecting a low-impedance ADC ($10\,\text{k}\Omega$) directly to a high-impedance voltage divider sensor siphons off current, causing significant signal droop. Always buffer with high-impedance op-amp voltage followers ($R_{\text{in}} > 10^{12}\,\Omega$).
3. Excessive Filtering in Feedback Loops: Increasing the window size $N$ of a moving average filter creates smooth sensor graphs, but introduces phase lag that can cause severe instability and oscillation in high-speed robot position and force control loops.
4. Aliasing Without Anti-Alias Filters: Software filtering cannot fix aliased signals. If an analog signal contains $100\,\text{Hz}$ noise sampled at $120\,\text{Hz}$, the noise manifests irrevocably as a $20\,\text{Hz}$ false disturbance.

Quick Revision Summary:

• Sensor: Detects physical measurand; Transducer: Converts one energy domain to another.
• Active Transducers: Self-generating (Piezoelectric, Thermocouple, Photovoltaic); zero external power needed.
• Passive Transducers: Modulating (Strain Gauge, RTD, LVDT, Capacitive); require external excitation.
• Sensitivity: $S = dy/dx$; Linearity Error: Maximum deviation from reference line as % FSO.
• Systematic Errors: Constant/repeatable; eliminated by calibration.
• Random Errors: Stochastic/Gaussian; reduced by sample averaging ($\sigma_{\bar{x}} = \sigma / \sqrt{N}$).
• First-Order Time Constant $\tau$: Reaches $63.2\%$ in $\tau$, $95\%$ in $3\tau$, $99.3\%$ in $5\tau$. Cutoff $f_c = 1/(2\pi\tau)$.
• Smart Sensor: Integrates Sensing Core + AFE + ADC + MCU + TEDS Memory + Digital Fieldbus.