Why Is BJT More Noisy Than a MOSFET or FET? Explained

Why is BJT more noisy than a comparable MOSFET? Because a bipolar transistor conducts using both electrons and holes, and each of those carriers drags shot noise, thermal fluctuation, and flicker energy across the junction with it. Field-effect parts move current with majority carriers only, so they start quieter. That single physical asymmetry is most of the answer. Everything below unpacks the six real noise mechanisms hiding inside the device, tells you which one bites hardest at which frequency, and hands over the exact bench playbook I run when a bipolar front end refuses to shut up.

Last updated: August 13, 2026. Rewrote as a definition-first explainer, added a BJT-vs-FET noise table, a bench workflow, and an expanded FAQ.

Why Is BJT More Noisy Than a MOSFET?

Short version. A BJT is current-controlled. Actual charge flows into the base, and that traffic is built from countable electrons and holes hopping across a junction. Countable carriers mean shot noise. A MOSFET gate sips almost nothing. Its channel obeys a voltage across an insulator, so no steady gate current rattles the input. Simple as that.

That gap shows up in the datasheet. A quiet bipolar part like the 2N4403 or the BC550C sits near 0.7 to 2 nV/√Hz for equivalent input voltage noise. A low-noise JFET can hit the same floor while its input current noise stays 10 to 100 times lower. So two candidates can tie on voltage noise and still behave wildly differently once source impedance climbs above 5 kΩ.

The rule I keep pinned above the bench is simple. Below roughly 500 Ω of source resistance, the bipolar part usually wins. Above 20 kΩ, the FET pulls ahead because its current noise barely counts. In my experience the mismatch between transistor and source ruins more preamps than any single component swap ever does, and when I inherit a hissy front end the source resistance is the first quantity I actually measure. Match the device to what feeds it. Not to a slogan.

Noise trait	BJT	MOSFET / JFET
Carriers in conduction	Electrons and holes	Majority carriers only
DC input (gate/base) current	Microamps	Picoamps to nanoamps
Input current noise	Higher (shot noise on base)	Very low
Best source impedance	Low (under ~1 kΩ)	High (tens of kΩ+)
1/f corner	Often lower	Often higher (surface effects)

The Six Noise Sources Inside a BJT

Six mechanisms stack up inside the part. Some are shared with every resistor on Earth. Some are pure bipolar.

Thermal noise (Johnson-Nyquist noise) comes from electrons jostling around at any temperature above absolute zero. It lives in the base resistance and rises with the square root of bandwidth. Cool the part or shrink the bandwidth and it drops.
Shot noise (shot noise) comes from charge crossing a junction one carrier at a time. Its arrival times are random. It scales with the square root of the DC current, and it is flat with frequency.
Base current noise rides on that small base current. The base current is not smooth. It is a stream of discrete carriers, so it carries shot noise straight into the input.
Emitter-base junction noise is the same story at the input diode. Every junction acts like a diode, and every diode shows shot noise.
Generation-recombination noise happens when carriers appear and vanish in the base region. Each event is a tiny random kick to the current.
Flicker noise (1/f noise) climbs as frequency falls. Surface traps and lattice defects cause it. It dominates below the 1/f corner, often somewhere from a few Hz to a few kHz.

Two of these, base current shot noise and generation-recombination, exist only because a BJT relies on minority carriers. A FET does not carry them the same way. That is the deep reason the bipolar variant starts noisier. If you have ever wrecked a small-signal transistor by feeding it backwards, some of the same junction physics explains the collapse, and I unpack it in what happens when you reverse polarity on a transistor.

Order of magnitude matters. For a typical low-noise bipolar biased at 1 mA, the base-current shot term sits around 1 pA/√Hz. Push that pA against a 100 kΩ source and you get 100 nV/√Hz of extra hiss riding on the signal. The same source paired with a JFET at 1 fA/√Hz contributes essentially zero. Thirty decibels of gap, from one design decision. The IEEE noise-analysis literature has been documenting that exact tradeoff since the 1960s.

Why Minority Carriers Make It Worse

A BJT is a minority-carrier device. Base-region conduction leans on charge that is dwarfed by the surrounding dopant crowd. A smaller population fluctuates harder in percentage terms. Same reason a survey of 10 respondents swings wilder than a survey of 10,000.

So every recombination event and every fresh injection stamps a measurable wobble on the signal. That wobble peaks below a few hundred hertz, exactly where the 1/f tail lives. Wire a bipolar preamp to a scope with the gain cranked and you’ll hear the whole story. Quiet hiss in the audio midband. A gentle whistle near the corner. Then a slow rumble grows as you drift toward DC and the minority-carrier fluctuations start bullying everything else the stage is doing. Call that rumble the bipolar tax. You pay it in exchange for raw current gain.

How I Cut BJT Noise on the Bench

Here is the workflow I actually run when a bipolar front end is too noisy. Do it in order. Order matters more than the shopping list. Every time I have skipped ahead to swap in a fancy transistor before fixing the bias and the source, I have paid for it with an afternoon of chasing a floor that would not budge. Do not skip step one. It is free.

Fix the source impedance first. Work out the source resistance, then pick the part whose optimum source impedance is close. A BJT wants a low-ohm source. Get this wrong and no amount of tweaking saves you.
Set the collector current on purpose. Voltage noise falls as you raise collector current. Current noise rises. There is a sweet spot, usually a few hundred µA to a few mA for small-signal work. I sweep it and watch the noise floor on a spectrum analyzer.
Keep the base current low where you can. Less base current means less base shot noise. A higher-beta part helps, so a device with strong current gain earns its place. If you want the physics behind that gain, I wrote up why the base current runs weaker than the collector current separately.
Bandlimit hard. Thermal and shot noise both grow with bandwidth. If your signal lives in 20 kHz, do not let the amplifier see 2 MHz. Filter it down.
Pick a genuine low-noise part. Some BJTs are designed for it, with clean surfaces and doping tuned to suppress recombination. Check the nV/√Hz and pA/√Hz numbers on the datasheet, not the marketing.
Parallel devices for the last dB. Running several matched transistors in parallel drops input voltage noise by roughly the square root of the count. Four in parallel buys you about 6 dB. Sixteen buys 12 dB. It costs current and board real estate, so I save it for last.

One more thing. This all applies to the front-end device only. Once the first amplifier stage has meaningful gain, later stages barely matter for noise, thanks to Friis’ equation. So spend your energy, and your best transistor, on stage one. If instead you sit on the FET side of the fence, the tradeoffs flip and speed becomes the fight, which I cover in speeding up a MOSFET. Op-amp stability is another rabbit hole worth reading if you plan to wrap that transistor inside feedback, so bookmark how to check op-amp unity-gain stability.

Frequently Asked Questions

Why is BJT more noisy at low frequencies?

Because 1/f noise and generation-recombination noise both rise as frequency drops. In a BJT these come from minority-carrier events in the base. That is why the hiss turns to a rumble near DC.

Why is BJT more noisy than a JFET at high source impedance?

The BJT has real base current, and that current carries shot noise. Across a large source resistance that current noise turns into a big voltage. A JFET draws almost no gate current, so it stays clean when the source is tens of kΩ or more.

Is a MOSFET always quieter than a BJT?

No. At low source impedance a good BJT often beats a MOSFET on voltage noise, and MOSFETs tend to have a worse 1/f corner from surface traps. The winner depends on your source resistance and your frequency band.

Can I remove BJT noise completely?

No. Thermal noise and shot noise are set by physics, so you can only reduce them, never zero them out. Cooling, bandlimiting, and a smart bias get you most of the way. The rest is choosing the right part.

Does higher collector current lower the noise?

It lowers voltage noise but raises current noise, so there is a trade-off. For most small-signal front ends a few hundred µA to a few mA is the useful range. Sweep it and watch the spectrum.

Which single change helps most?

Matching the transistor to the source impedance. It is free, it takes five minutes, and it fixes more noise problems than any exotic part swap. Do it before you buy anything.

Bottom Line

Why is BJT more noisy? Because bipolar conduction leans on minority carriers and a genuine base current, and both throw shot noise plus a stubborn 1/f tail that a FET mostly avoids. You cannot delete that hiss, but you can tame it. Match the source impedance. Bias a smart collector current somewhere between 100 µA and 3 mA. Bandlimit hard. Pick a genuine low-noise part for the first amplifier stage. Nail those four and a bipolar front end can be startlingly quiet, well under 1 nV/√Hz above 1 kHz.