English is currently available.
Biomedical Engineering DISCUSSION
For a university lab I am building a single-lead ECG amplifier to feed a microcontroller ADC with a 3.3 V range. My first attempt with an ordinary op-amp differential stage gives mostly 50 Hz hum and a baseline that wanders off to the supply rail.
What signal levels and frequencies should I design for, how should the gain be split between stages, and what has to be done so that a circuit like this is safe to connect to a person?
The numbers to design around: the ECG at the skin is about 1 mV peak for the R wave, with features down to a few tens of microvolts. The useful band is roughly 0.05 to 150 Hz for diagnostic-quality recording and about 0.5 to 40 Hz for simple rhythm monitoring. On top of the signal sits the electrode half-cell offset, a DC voltage that can reach a few hundred millivolts and drifts with movement, and a mains-frequency common-mode voltage on the whole body that can be volts.
So the task is to pick 1 mV out of a common-mode signal a thousand times larger, with a DC offset hundreds of times larger.
That is why an instrumentation amplifier is used. A single op-amp differential stage has a low, unequal input impedance, and its common-mode rejection depends on resistor matching; 1 percent resistors give only roughly 35 to 40 dB. An instrumentation amplifier has very high input impedance on both inputs and typically 80 to 100 dB or more of rejection at mains frequency.
Electrode impedance imbalance still converts common-mode into differential signal, roughly V_cm × ΔZ / Z_in. With 1 V of common mode, 10 kΩ of imbalance and 1 GΩ input impedance, that is 10 µV, which is acceptable. With a 1 MΩ input it would be 10 mV, ten times the ECG.
Split the gain so the offset cannot saturate the first stage. With up to ±300 mV of electrode offset, a first-stage gain of 5 gives ±1.5 V, which still fits inside the supply. Then remove the DC with a high-pass filter (0.5 Hz is a common choice for monitoring) and put the remaining gain after it. A second-stage gain of 200 brings the total to 1000, so 1 mV becomes 1 V, centred on a mid-supply reference for the ADC.
Follow with a low-pass at 40 to 150 Hz and sample at 250 to 500 Hz. The wandering baseline you saw is the electrode offset being multiplied by the full gain.
For the hum, a driven right-leg circuit helps a great deal: the common-mode voltage is sensed at the instrumentation amplifier's gain-setting network, inverted, and fed back to a third electrode through a resistor of a few hundred kilohms that limits the current. It lowers the common-mode voltage on the body instead of relying on rejection alone. Twist or shield the electrode leads, keep them short, and use proper gel electrodes on clean skin to keep the impedance low and balanced.
A notch filter at mains frequency is a last resort, because it distorts the waveform near that frequency.
On safety, the rule for a student build is simple: the circuit and everything connected to it must run from batteries only while electrodes are on a person. That means no mains-powered oscilloscope, no bench supply, and no laptop on its charger connected through USB. Any of those can provide a path from mains to the electrodes if something fails. Send data over a wireless link or through a properly rated isolator, and keep series resistors in the electrode leads.
Test with a signal generator and attenuator or an ECG simulator first. A circuit like this is for learning; it is not a medical device and its output must not be used for diagnosis.
Back to Biomedical Engineering community
