Dry-Electrode ECG Front End With 50 Hz Interference
A single-lead ECG front end uses dry electrodes with a 5 kΩ electrode–skin contact mismatch between the two active inputs. The instrumentation amplifier has a differential gain of 1000, an input impedance of 10 MΩ, a CMRR of 80 dB at 50 Hz and an input bias current of 5 nA. Capacitive coupling from the mains raises the patient's body to a 1.5 V, 50 Hz common-mode potential. The QRS amplitude at the electrodes is about 1.0 mV, and the ADC after the amplifier has a ±5 V input range. A 50 Hz notch filter is available in the analogue path, and a colleague proposes switching it on as the sole fix. Motion tests show the baseline wandering by ±300 µV, and the electrode contact impedance changes noticeably when the patient shifts. The supply is medical-grade and the measured enclosure leakage current is 40 µA.
1. Quantify each path by which the 50 Hz common-mode voltage reaches the output and rank them.
Read the worked answer
Three paths matter. The electrode mismatch converts common mode to differential mode directly: 1.5 V × 5 kΩ/10 MΩ = 750 µV. The finite CMRR contributes 1.5 V/10⁴ = 150 µV, since 80 dB is a factor of 10⁴. The input bias current flowing through the 5 kΩ mismatch adds a DC offset of 5 nA × 5 kΩ = 25 µV, which matters for saturation but not for 50 Hz. The first two are at the same frequency and add, giving roughly 900 µV, which is 90 % of the 1.0 mV QRS. After a gain of 1000 the wanted signal is 1.0 V and the interference is 0.9 V: the amplifier does not saturate against a ±5 V range, but the signal-to-interference ratio is essentially zero. The dominant path is the mismatch term, so any fix that does not raise input impedance or reduce the mismatch leaves the loop in trouble.
2. Evaluate the notch-filter proposal. What would you do instead?
Read the worked answer
A 50 Hz notch would attenuate both dominant terms, but only the steady component. Contact impedance changes with movement, so the mismatch term is amplitude-modulated by motion at sub-hertz rates; the resulting sidebands sit either side of 50 Hz and pass straight through a narrow notch. A notch also rings, distorts the low-frequency ST segment, and leaves residual interference because mains frequency drifts and notch components have finite depth. Treat it as a supplementary measure, not the fix. Instead, raise the input impedance: a buffer with 1 GΩ input reduces the mismatch term from 750 µV to about 7.5 µV, a hundredfold. Add a driven-right-leg circuit to actively null the common-mode voltage at the body, which reduces both terms proportionally. Then improve CMRR from 80 dB to 100 dB, which brings the CMRR path to 15 µV. The residual would then be roughly 22 µV, about 2 % of the QRS, which is a workable design.
3. State a specification for the input stage and check the safety figures given.
Read the worked answer
Specification: input impedance at least 1 GΩ in parallel with less than 50 pF, to keep the mismatch term well below 20 µV; CMRR at least 100 dB at 50 Hz, measured with a 5 kΩ imbalance in one input lead, not with matched sources; input bias current below 1 nA so that a 5 kΩ mismatch contributes under 5 µV and the DC offset stays small; input-referred noise below 1 µV RMS over 0.5–40 Hz so the 1 mV QRS is not buried. Safety: the measured 40 µA enclosure leakage is within the 100 µA patient-auxiliary limit for normal condition, but any driven-right-leg or buffer stage must include a current-limiting resistor and must not increase leakage if a single component fails. Verify the isolation barrier rating and that the 50 Hz notch, if retained, is switched on only after the input stage is corrected, since a notch can hide a failing electrode connection.