Circuit Design

Bandgap Reference

/band-gap ref-er-ents/
A precision voltage reference circuit that produces a temperature-stable output of approximately 1.25 V by summing a CTAT (complementary to absolute temperature) base-emitter voltage with a PTAT (proportional to absolute temperature) thermal voltage difference. The 1.25 V output corresponds to the extrapolated bandgap energy of silicon at 0 K. First-order temperature coefficients of 10 to 50 ppm/°C are standard; curvature-corrected topologies achieve below 5 ppm/°C across -40 to +125°C, making them essential for biasing RF ADCs, DACs, LNAs, and PLL charge pumps.
Output: ~1.25 V
TC: 5-50 ppm/°C
Process: CMOS / BiCMOS

Understanding Bandgap References

Every RF integrated circuit needs stable bias voltages that remain constant across temperature, supply variation, and process corners. The bandgap reference solves this by exploiting two complementary temperature behaviors of bipolar junctions. A single transistor's VBE drops at roughly -2 mV/°C (CTAT). The difference in VBE between two transistors at different current densities (ΔVBE = VT × ln(N)) rises at +0.085 mV/°C per unit of ln(N) (PTAT). By scaling the PTAT voltage with a resistor ratio K and summing it with VBE, the linear temperature terms cancel.

In modern RF transceiver SoCs, the bandgap reference is typically shared among multiple subsystems. The PLL charge pump draws a reference current to set loop bandwidth. The ADC and DAC rely on the reference for full-scale accuracy. LNA and PA bias DACs use it to maintain consistent gain. Any drift in the reference propagates to all these blocks, so post-layout simulation including package parasitics and substrate noise coupling is essential for production-grade designs.

Core Equations

Bandgap output voltage:
VREF = VBE + K × ΔVBE
VREF = VBE + K × (kT/q) × ln(N)

At the zero-TC point:
∂VREF/∂T = 0
K × (k/q) × ln(N) = |∂VBE/∂T|
K × ln(N) ≈ 23.2 (for ∂VBE/∂T = -2 mV/°C)

Typical values:
N = 8: K = ln(8) × correction ≈ 11.2
VREF ≈ 0.6 V + 11.2 × 0.054 V ≈ 1.21 V

Reference Type Comparison

TypeOutput (V)TC (ppm/°C)PSRR (dB)CurrentIntegration
Basic BGR1.2520-50-405-50 μAOn-chip CMOS
Curvature-corrected1.252-10-6020-100 μAOn-chip BiCMOS
Sub-1V BGR0.5-0.910-30-501-10 μAOn-chip deep CMOS
Buried Zener7.00.05-1N/A1-5 mADiscrete only
FET-basedVaries50-200-300.5-5 μAOn-chip CMOS
Common Questions

Frequently Asked Questions

How does a bandgap voltage reference work?

It sums two voltages with opposite temperature coefficients. V_BE drops at -2 mV/°C (CTAT), while ΔV_BE between two transistors at different current densities rises proportionally to temperature (PTAT). A resistor ratio K scales the PTAT term so the linear temperature dependence cancels, yielding a stable output near 1.25 V. Curvature-corrected designs add a second-order compensation term to reduce residual parabolic drift below 5 ppm/°C.

Why are bandgap references critical in RF systems?

RF data converters need reference stability better than half an LSB over temperature. A 14-bit DAC with 1 V full scale requires better than 61 μV stability (50 ppm). PLL charge pump current, set by the reference, directly affects phase noise. LNA bias drift changes gain and noise figure. In phased arrays with hundreds of channels, reference drift causes beam pointing errors as each channel's phase shifter bias shifts differently.

How does a bandgap reference compare to a Zener reference?

Bandgap references operate at 1.25 V, consume microamps, integrate on-chip in CMOS/BiCMOS, and achieve 5-50 ppm/°C. Zener references use reverse breakdown at 5-7 V, achieve sub-1 ppm/°C in buried designs (LTZ1000), but require higher voltage, dissipate milliwatts, and remain discrete-only. For RF ICs operating at 1.2-3.3 V supply, bandgap references are the standard choice.

RF IC Design Support

Request a Quote

Need precision bias circuit design, reference characterization, or RF transceiver integration support? Contact our engineering team.

Get in Touch