Appendix CDesign-Review Checklists
The questions a reviewer should ask of a biopotential front-end, collected from the pitfall boxes of the chapters. Each item links to where it is explained.
C.1Specification
- Every amplitude stated as peak-to-peak or rms, and every noise figure with its bandwidth (Chapters 2 and 6).
- The largest signal, the electrode offset and the artefacts all fit the input range at every gain setting (Chapter 11).
- The specification names the electrodes it assumes, wet or dry, and the noise budget includes their thermal noise (Chapters 3 and 29).
- The CMRR requirement is stated with the standards’ electrode imbalance, not for the amplifier alone (Chapter 4).
- The system budget includes the radio, not only the front-end (Chapter 27).
C.2Analog front-end
- Noise. Input-referred noise integrated over the signal band, including the amplifier’s current noise times the electrode impedance, the input-protection resistors, the servo loop’s noise gain and the flicker noise left after chopping (Chapters 6, 9, 10).
- Wideband noise. The LNA’s noise above the signal band is filtered before any sampling, or it folds into the band (Chapter 8).
- CMRR. Effective CMRR with the imbalance network at 50 and 60 Hz, including the input impedance, the RF filter’s mismatch and the amplifier’s intrinsic CMRR across frequency (Chapters 4 and 7).
- Offset. ±300 mV of electrode offset tolerated at every gain, with the servo loop’s range and recovery time checked (Chapter 10).
- Input impedance. At the chopping frequency, with the parasitic capacitance of pads, ESD and routing, and with the boosting loop’s accuracy over corners (Chapters 9 and 10).
- Pseudo-resistors. High-pass corner over temperature (27 to 85 °C moves it by about two decades) and recovery after a large step (Chapter 8).
- Chopper ripple and spikes. Residual ripple at the output and the ripple’s alias after sampling (Chapter 9).
- Protection. ESD on every pad, series resistance small enough for the noise budget, defibrillation protection off chip if required (Chapter 12).
- RF filtering. Input RC filters matched, so that they do not turn common mode into differential mode (Chapter 12).
- Right-leg drive. Output current limited for patient safety in normal and single-fault conditions; phase margin over the full range of body capacitance (Chapter 12).
- Shields. Driven, not grounded; a grounded shield costs tens of megohms of common-mode impedance (Chapter 4).
- Lead-off detection. DC currents small enough not to disturb the signal, and their effect on the input offset checked (Chapter 12).
C.3Converters
- kT/C. Sampling capacitors sized for noise at the converter’s resolution, differential noise included (Chapters 14 and 15).
- Matching. Unit capacitors sized for the DNL target at 3σ, or redundancy and calibration planned (Chapter 17).
- Comparator. Input-referred noise below the LSB budget, offset handled, kickback and metastability checked (Chapter 16).
- Switches. On-resistance across the input range (bootstrapped if needed), charge injection, and leakage in hold at the highest temperature (Chapter 15).
- Reference. Charge per conversion supplied without disturbing the reference; decoupling sized (Chapter 17).
- ΔΣ loop. Stability at the largest input including offsets, integrator swing, idle tones with DC inputs (Chapters 19 and 20).
- Decimator. Word lengths, droop and attenuation at the chopper frequency (Chapters 20 and 23).
- Clocks. Jitter adequate for the input frequency; chopper, converter and excitation clocks derived from one source, with no harmonic collisions (Chapters 13 and 28).
C.4Simulation
- Simulator settings do not dominate the result: \(g_\mathrm{min}\) and the absolute tolerance lowered for teraohm and femtoampere work (Chapter 30).
- Time-varying circuits (choppers, switched capacitors, comparators) simulated with transient or periodic-steady-state noise, not AC noise (Chapter 30).
- Corners (process, ±10 % supply, temperature) and Monte Carlo with the PDK’s mismatch libraries for every specification (Chapter 30).
- Spectral tests use coherent sampling or a window whose leakage stays below the noise floor (Chapter 22).
C.5Layout
- Input pairs and capacitor arrays common-centroid with dummies (Chapter 17).
- Guard rings around the input pads and between analog and digital areas; high-impedance nodes shielded (Chapter 12).
- Choppers and comparators laid out symmetrically; clock lines kept away from the inputs.
- Separate supplies and grounds for analog, converter and digital sections (Chapter 27).
- DRC, LVS and post-layout simulation clean at all corners (Chapter 30).
C.6Before tape-out
- Every block’s reliability checked: no thin-oxide device sees more than its rated voltage, including the bootstrapped switch and the 3.3 V servo (Chapter 15).
- Test access: raw converter outputs, test points or scan for the digital logic, and a bench plan (Chapter 22).
- Seal ring, ESD on every pad, padframe from the template (Chapter 30).
- Design files, simulation scripts and results under version control and an open licence.