Back to timeline
school Sep 2025

A discrete emitter-follower buffer

An NPN emitter-follower built around a BC547 (the canonical small-signal transistor) for buffering a 500 mV / 1 kHz sine between a 5.6 kΩ source and a 680 Ω load. The project is half "design the bias point", half "measure how reality diverges from the textbook."

The design choices

The numbers

Calculated input impedance: ~52 kΩ, predicting a source-side voltage divider of 52k / (5.6k + 52k) ≈ 0.903. So the buffer should hand the base roughly 90% of the source amplitude, and from there an emitter-follower almost-unity-gains the signal to the load.

Measured: input 492 mV, output 410 mV → ratio 0.83. Below target. Tuning RE down to 1.52 kΩ (which raises IE to ~23 mA, giving better re) brought it up to 441 mV / 490 mV = 0.90, exactly the predicted ratio. Lower-frequency knee from measurement: f−3dB ≈ 13 Hz.

Oscilloscope trace of the theoretically-valued circuit, showing input and output sinusoids with the output noticeably smaller than the input.
Designed values: input 492 mV, output 410 mV. Ratio 0.83, short of the 0.90 target.
Oscilloscope trace after lowering R_E to 1.52 kΩ. Input and output sinusoids are now much closer in amplitude.
RE retuned to 1.52 kΩ. 441 mV / 490 mV = 0.90, exactly the predicted ratio.
Measured frequency response of the buffer: flat above ~20 Hz, rolling off below, with the −3 dB point near 13 Hz.
Low-frequency response. The −3 dB knee at ~13 Hz, set by the 22 µF coupling caps.

What I would do differently

Higher divider stiffness if the source could tolerate it (5–20× IB instead of 3×) for more β-independence. Bigger coupling caps if low-frequency response matters. And if precision is the goal, do not build an emitter follower at all. Use an op-amp with a buffered output. The discrete transistor was the assignment, not the optimum.

Writeup

Back to timeline