Traceline — What 60 Hz Hum Taught Me About Signal Integrity

A decade of designing bedside monitors taught me that the cleanest waveform is the one you never have to filter.

I inherited the CardioTrack 840 firmware three weeks before its 510(k) resubmission. The ECG trace developed a 2.8 mV ripple on every third run. Three cascaded IIR notch filters, each tuned to 60 Hz with a Q of 30, worked in simulation and failed under fluorescent ballasts.

The False Security of the Notch Filter

A Q-30 notch at 60 Hz attenuates exactly one frequency in theory. In practice, every analog front-end drifts. A 0.1% resistor mismatch tilts common-mode rejection by 6 dB, and the filter skirts begin carving into the QRS complex itself.

Every filter you apply to a signal is an admission that you failed to protect it at the source.

Ground Loops in Bed 7

Bed 7 exposed the fault. A 1998 magnetic ballast shared conduit with nurse-call wiring, and the leads picked up 58.7 Hz at 3.6 mV peak-to-peak. Shielded twisted pair, a bonded bed frame, and no notch stages dropped the ripple below the ADC’s least significant bit.