BJT Bipolar Junction Transistor

For the same NPN transistor, it must remain in the forward-active region when used as an amplifier, but is often desired to enter saturation when used as a low-voltage switch. Applying $I_C=\beta I_B$ to both cases without checking the operating region will yield impossible current values. This article starts with carrier transport in the thin base, then uses voltage-divider bias and small-signal models to turn "amplification" into a calculable process.

Not Equivalent to Two Back-to-Back Diodes

A BJT has an Emitter (E), Base (B), and Collector (C), with common structures being NPN or PNP. Although there are two junctions, BE and BC, the transistor action depends on a sufficiently thin base region between adjacent junctions and its doping structure. Connecting two independent diodes back-to-back cannot reproduce the same carrier transport across the base.

Taking the NPN forward-active region as an example, the forward-biased BE junction causes the emitter to inject electrons into the base. The base is thin and properly designed so that most injected electrons reach the BC depletion region before recombining, and are collected by the electric field into the collector; the base current includes contributions from base recombination and injection from the base into the emitter. The energy required for the collector current comes from the external power supply, not generated out of thin air by the small base current.

Under applicable approximations, $I_C\approx I_S\exp (V_{BE}/V_T)$, and $I_B\approx I_C/\beta$, so we can also write $I_C\approx \beta I_B$. "Small current controlling large current" is a convenient port description, but $\beta$ is not a stable material constant; voltage, temperature, operating current, and manufacturing variations all affect it. The emitter current satisfies $I_E=I_C+I_B$.

The arrow on the NPN symbol points outward, indicating that conventional forward emitter current flows out; for PNP, the arrow points inward, and voltage and conventional current directions are reversed accordingly. The arrow does not represent a path of "electrons passing sequentially through N→P→N," and you cannot guess the B/C/E pins solely by the package outline.

Operating Regions and Load Constraints

Cutoff is not "all currents exactly zero," and saturation is not "$V_{CE}$ always 0.1 V." For example, with the emitter grounded and the collector connected to $V_{CC}$ via $R_C$, the external relationship is $V_{CE}=V_{CC}-I_C{R}_C$. If the current required by $\beta I_B$ makes this expression much lower than the actual on-state voltage drop, the device cannot maintain the forward-active approximation.

$V_{BC}=V_B-V_C$ is the entry point for directly checking conditions. When designing an amplifier, do not judge the active region solely by $V_{CE}$ being positive; if $V_C$ is lower than $V_B$, the BC junction is already forward-biased. When designing a switch, you must combine the $V_{CE(sat)}$ given in the datasheet for specific $I_C$ and $I_B$, rather than multiplying the minimum $h_{FE}$ from the active region by an arbitrary coefficient to ensure saturation. Nexperia: BJT Application Handbook

Complete Biasing Example: Accounting for Base Loading First

The following circuit uses $R_1=68\text{ k}\Omega$ and $R_2=22\text{ k}\Omega$ to voltage-divide the base, $R_E=1\text{ k}\Omega$ connected to the emitter, $R_C=3.3\text{ k}\Omega$ connected to the collector, and $V_{CC}=12\text{ V}$. Temporarily assume a DC teaching model with $\beta =100$ and $V_{BE}=0.7\text{ V}$, and check the operating region later. These assumptions are used only to demonstrate the biasing algorithm.

Replace the base voltage divider with a Thevenin source: $V_{th}=12\times 22/(68+22)\approx 2.933\text{ V}$, $R_{th}=68\text{k}\parallel 22\text{k}\approx 16.622\text{ k}\Omega$. The base current will cause the actual $V_B$ to be lower than the no-load divider value; you cannot directly fix $V_B$ at $2.933\text{ V}$.

The base loop is $V_{th}=I_B{R}_{th}+V_{BE}+I_E{R}_E$, and using $I_E=(\beta +1)I_B$, we get:

$I_B=(V_{th}-V_{BE})/[R_{th}+(\beta +1)R_E]$.

Substituting yields $I_B\approx 18.99\text{ muA}$, $I_C\approx 1.899\text{ mA}$, $V_E\approx 1.918\text{ V}$, $V_B\approx 2.618\text{ V}$, $V_C\approx 5.734\text{ V}$, and $V_{CE}\approx 3.816\text{ V}$. Since $V_C>V_B$, the BC junction is reverse-biased, and the forward-active assumption is self-consistent.

How the Emitter Resistor Provides Feedback

If temperature or device variations cause the collector current to tend to increase, the emitter current and $V_E$ increase accordingly; when the base voltage is constrained by the bias network, $V_{BE}$ decreases, suppressing the initial current increase. It reduces sensitivity, but does not eliminate base loading, $V_{BE}$ temperature drift, or $\beta$ variations.

With other model parameters unchanged, if $\beta$ changes from 50 to 200, the complete formula gives $I_C$ changing from approximately $1.651\text{ mA}$ to $2.052\text{ mA}$. Compared to simply fixing $I_B$ and letting $I_C$ change proportionally with $\beta$, the feedback significantly reduces the variation. Whether it is stable enough still depends on output swing, power consumption, and precision requirements.

The smaller $R_{th}$ is relative to $(\beta +1)R_E$, the "stiffer" the divider, the less impact base loading has, but static power consumption is higher. You cannot simply replace error budgeting with the rule "divider current is fixed at ten times the base current."

Does doubling β double collector current?

At RE=1 kΩ, compare β=50 and 200 and watch IB change through feedback. Reducing RE can move the candidate past VC=VB. Candidate values then explain why the assumption fails; they are not actual saturation current.

$$I_B=\frac{V_{th}-V_{BE}}{R_{th}+(\beta +1)R_E}$$

Hybrid-π and Gain: Linearizing Around the Q-Point

Taking the derivative of the exponential relationship of $I_C$ in the forward-active region yields $g_m=I_C/V_T$, representing how much $I_C$ changes for a small change in $V_{BE}$; $r_{\pi }=\beta /g_m$ describes the small-signal input resistance of B–E. The Early effect causes output current to also vary with $V_{CE}$, commonly described by the approximation $r_o\approx V_A/I_C$, where $V_A$ is the Early voltage parameter, not the base voltage.

In the diagram, $r_{\pi }$ must be across B–E, and the controlled current source $g_m{v}_{\pi }$ and $r_o$ across C–E. It is not placing a resistor between B–C; connecting the wrong terminals will change the entire amplification relationship. MIT: BJT Incremental Model

Continuing with $I_C\approx 1.899\text{ mA}$ and $T=300\text{ K}$, $g_m\approx 73.45\text{ mS}$, $r_{\pi }\approx 1.36\text{ k}\Omega$. If the emitter can be considered an AC ground at the frequency of interest, ignoring $r_o$ and connecting a $10\text{ k}\Omega$ load to the output, the effective collector load $R_C\parallel R_L\approx 2.481\text{ k}\Omega$, and the gain from base voltage to collector voltage is approximately $-g_m(R_C\parallel R_L)\approx -182$.

This large gain does not mean you can input hundreds of millivolts without distortion. The derivation uses local linear changes where $|v_{\pi }|$ is much smaller than $V_T$; the output is also limited by static headroom and cutoff/saturation. The source impedance and bias network will further attenuate the signal actually reaching the base.

Why Gain Drops When There Is No Bypass Capacitor

If $R_E$ remains in the AC path, part of the base voltage change becomes an emitter follower change, making the actual $v_{\pi }$ smaller. Ignoring $r_o$:

$A_v=-g_m(R_C\parallel R_L)/[1+(g_m+1/r_{\pi })R_E]$.

In this example, it is approximately $-2.42$, much smaller than the $-182$ in the fully bypassed case, but it brings stronger feedback and a larger input range. The input impedance looking into the base is approximately $r_{\pi }+(\beta +1)R_E$, which must be placed in parallel with the bias resistors. The statement "BJT input impedance is always a few kΩ" ignores the circuit feedback structure.

Placing a capacitor $C_E$ in parallel with $R_E$ can reduce AC feedback at certain frequencies while retaining DC bias feedback. However, the degree of bypassing varies with frequency, and its pole is determined by the resistance actually seen; you cannot guarantee the same gain across the entire frequency band based solely on $R_E{C}_E$.

The common-collector emitter follower takes the output at the emitter, with a voltage gain typically close to but less than 1, providing current drive and impedance transformation; its output voltage is still limited by $V_{BE}$, supply headroom, and load, and cannot be treated as an ideal lossless voltage replicator.

Switching and Rated Conditions

As a low-side NPN switch, the collector connects to the load, the emitter grounds, and the base is supplied by the driver through a resistor. You need to know the load current and allowed voltage drop first, then select a base current that meets the saturation specifications at that point.

For example, the onsemi 2N3904 datasheet gives a maximum $V_{CE(sat)}$ of $0.2\text{ V}$ under conditions of $I_C=10\text{ mA}$ and $I_B=1\text{ mA}$; this condition corresponds to a forced $\beta =I_C/I_B=10$. This does not mean the device's active region $\beta$ is only 10, but rather indicates the drive conditions used to guarantee this saturation. onsemi: 2N3904 Datasheet

Assume a teaching driver source guarantees a high level of at least $3.0\text{ V}$ under the corresponding load. Using the upper bound of $V_{BE(sat)}$ given at the above test point, $0.85\text{ V}$, to provide at least $1\text{ mA}$, the actual $R_B$ should not exceed $2.15\text{ k}\Omega$. Afterwards, you must also consider resistor tolerance and the driver's maximum current supply; the common "3.3 V GPIO" label does not guarantee $3.3\text{ V}$ output under load. This calculation cannot be directly extended to guaranteed performance under all temperatures and arbitrary collector currents.

Deeper saturation usually increases stored charge, which needs to be extracted or recombined during turn-off, potentially increasing storage time. Increasing the base current does not always simultaneously achieve the lowest loss and fastest turn-off; actual driving also involves current removal, reverse base current capability, and B–E reverse ratings.

Darlington and Device Comparison

A Darlington pair connects the emitter of the first transistor to the base of the second, and connects both collectors together. When both transistors satisfy their respective active approximations, the total current gain is ${\beta }_1{\beta }_2+{\beta }_1+{\beta }_2$, which approximates the product only when $\beta$ is large. Two BE junctions are in series from input to final emitter, requiring higher voltage; the saturation voltage drop and storage behavior are also different, and you cannot simply replace a standard BJT with high gain.

Comparison between BJT and MOSFET should be based on specific operating points. BJTs have a continuous base current, while MOSFETs have small DC gate current but require charge/discharge during switching; BJTs can have very high transconductance efficiency, while MOSFET conduction loss is often analyzed based on $R_{DS(on)}$ at specific $V_{GS}$ and temperature. Noise, speed, power, and reliability cannot be ranked solely by device category.

from math import isclose
vcc, r1, r2, rc, re, vbe = 12, 68000, 22000, 3300, 1000, .7
vth, rth = vcc*r2/(r1+r2), r1*r2/(r1+r2)
def bias(beta):
    ib = (vth-vbe)/(rth+(beta+1)*re)
    ic, ve = beta*ib, (beta+1)*ib*re
    vb, vc = ve+vbe, vcc-ic*rc
    assert vc > vb
    assert isclose((vcc-vb)/r1, vb/r2+ib)
    return ib, ic, ve, vc
ib, ic, ve, vc = bias(100)
vt = 8.617333262e-5*300
gm, rpi = ic/vt, 100/(ic/vt)
load = 1/(1/rc+1/10000)
av = -gm*load/(1+(gm+1/rpi)*re)
assert -2.5 < av < -2.4
assert bias(50)[1] < ic < bias(200)[1]
print(f"Ic={ic*1000:.4f} mA，VE={ve:.4f} V，VC={vc:.4f} V，VCE={vc-ve:.4f} V")
print(f"gm={gm*1000:.3f} mS，Unbypassed Gain={av:.3f}; Bias KCL and Active Region Check Passed")

These results come from a teaching model with fixed $\beta$, $V_{BE}$, and low-frequency linearization, and do not replace device corner and temperature analysis. The next article MOSFET continues to discuss threshold, on-resistance, and gate charge, avoiding the misunderstanding that "voltage control" means no drive current is needed.