Why would I choose an active mixer over a passive one?

Active mixers provide conversion gain (typically +8 to +15 dB) instead of conversion loss (typically -6 to -8 dB for passive). This means you can often eliminate a post-mixer IF amplifier stage, saving board space and cost. Active mixers also require much less LO drive power: -10 to 0 dBm versus +7 to +17 dBm for passive double-balanced mixers. The trade-offs are worse IIP3 (typically -5 to +5 dBm vs. +15 to +30 dBm for passive) and higher noise figure (10 to 15 dB vs. 6 to 8 dB for passive).

How does a Gilbert cell achieve frequency conversion?

The Gilbert cell uses a differential pair of transistors as the transconductance (RF) stage, which converts the RF input voltage to a current. This current feeds into a quad of cross-coupled switching transistors driven by the LO. The switching transistors alternately steer the RF current to opposite sides of the load, effectively multiplying the RF signal by a square wave at the LO frequency. The resulting product contains sum and difference frequencies. A filter selects the desired IF output.

What limits the frequency range of a Gilbert cell mixer?

The switching transistors must transition between on and off states within a fraction of the LO period. As LO frequency increases, the finite switching time causes the transistors to spend more time in the linear (non-switching) region, degrading conversion gain, increasing noise, and worsening port isolation. In CMOS, Gilbert cells work well to about 10 GHz in 65 nm and 30 GHz in 28 nm. SiGe HBT processes push the limit to 60 GHz and beyond due to their higher fT.

Active Components

Active Mixer

Passive diode-ring mixers have been the workhorse of RF receivers for decades, but they come with a built-in penalty: 6 to 8 dB of conversion loss and a demand for +7 to +17 dBm of LO drive, which is expensive to generate at high frequencies. An active mixer replaces the diode ring with transistor switching pairs (the Gilbert cell topology) and trades those disadvantages for conversion gain of +8 to +15 dB and LO drive requirements as low as −10 dBm. The price is worse linearity, higher noise figure, and DC power consumption, but for battery-powered receivers and highly integrated RFICs where board space and LO power are at a premium, the active mixer wins.

Category: Active Components

Key Topology: Gilbert Cell

Conversion Gain: +8 to +15 dB (typical)

When Gain Matters More Than Linearity

The Gilbert cell, invented by Barrie Gilbert at Analog Devices in 1968, remains the dominant active mixer topology in CMOS, SiGe, and GaAs integrated circuits. Its core is a differential transconductance stage (two transistors converting RF voltage to current) feeding a quad of cross-coupled switching transistors driven by the LO. The switching quad steers the RF current alternately to opposite load resistors, creating a multiplication of RF and LO signals that produces the sum and difference frequency outputs.

Active vs. Passive Mixer: Specification Comparison

Parameter	Active (Gilbert Cell)	Passive (Diode DBM)	Passive (FET Ring)
Conversion gain/loss	+8 to +15 dB	−6 to −8 dB	−5 to −7 dB
NF (DSB)	10 to 15 dB	6 to 8 dB	7 to 9 dB
IIP3	−5 to +5 dBm	+15 to +30 dBm	+20 to +35 dBm
LO drive	−10 to 0 dBm	+7 to +17 dBm	+7 to +13 dBm
DC power	5 to 50 mW	None	None
LO-RF isolation	30 to 50 dB	30 to 40 dB	25 to 35 dB
Integration	Excellent (on-chip)	Discrete module	Good (CMOS switch)

Improving Gilbert Cell Linearity

Source degeneration: Adding 5 to 20 Ω resistors in the source/emitter legs of the transconductance pair trades conversion gain for IIP3. Each 3 dB of gain reduction yields approximately 3 dB of IIP3 improvement.
Current bleeding: Injecting a DC current directly into the switching quad (bypassing the transconductance pair) increases the LO switching speed and improves conversion gain and noise figure without degrading IIP3.
Dynamic current injection: An auxiliary circuit injects signal-dependent current to cancel third-order distortion products. This can improve IIP3 by 8 to 12 dB at the cost of circuit complexity.

Common Questions

Frequently Asked Questions

Why choose an active mixer over passive?

Active mixers give +8 to +15 dB conversion gain (eliminating a post-mixer IF amp), need only −10 to 0 dBm LO drive (versus +7 to +17 dBm), and integrate well on-chip. The trade-offs: worse IIP3 (−5 to +5 dBm vs. +15 to +30 dBm) and higher NF (10 to 15 dB vs. 6 to 8 dB).

How does a Gilbert cell work?

A differential pair converts the RF voltage to current. Cross-coupled switching transistors driven by the LO steer that current alternately to opposite loads, multiplying RF by a square wave at the LO frequency. The product contains sum and difference frequencies; a filter selects the desired IF.

What limits the frequency range?

Switching transistors must transition within a fraction of the LO period. In 65 nm CMOS, Gilbert cells work to ~10 GHz. In 28 nm CMOS, ~30 GHz. SiGe HBT pushes the limit past 60 GHz due to higher f_T.

Related Terms

← Active Bias Active Multiplier →

RFIC Design

Explore Our Gilbert Cell Design Kit

Schematic templates, testbench setups, and layout guidelines for Gilbert cell mixers in 65 nm CMOS and 130 nm SiGe BiCMOS.

View Design Kits