A millimeter-wave low-noise amplifier (LNA) is a critical first-stage component in receivers where weak signals must be amplified without adding excessive noise. For system designers, selecting an appropriate millimeter-wave low-noise amplifier involves balancing noise figure, gain, bandwidth, power handling, manufacturing constraints, and operating environment. Those trade-offs affect applications ranging from radio astronomy and spaceborne sensing to radar, security imaging, and high-capacity wireless links.
Millimeter-wave frequencies span 30 to 300 GHz, where circuit parasitics, transmission-line losses, and transistor characteristics become especially important. Technology choice and circuit design therefore need to be considered together rather than as independent decisions.
Choosing a semiconductor technology
Leading millimeter-wave MMIC LNAs commonly use indium phosphide (InP) HEMTs or gallium arsenide (GaAs) metamorphic HEMTs. Both use a heterostructure that forms a two-dimensional electron gas, supporting high electron velocity and operation at high frequencies. Gate lengths of 100 nm or less are common in the devices discussed in this field.
InP HEMT: prioritize sensitivity
InP HEMTs have a strong record in very-low-noise amplifiers and can provide high gain and low noise at frequencies extending into the millimeter-wave range. They are particularly relevant to scientific instruments and other receivers where sensitivity is a dominant requirement, including systems operated at cryogenic temperatures.
These advantages come with practical limitations. InP is comparatively expensive and less suited to high-volume manufacturing. Its relatively low breakdown voltage also restricts its suitability for applications requiring greater power handling.
GaAs mHEMT: balance performance and production needs
GaAs mHEMT technology uses a metamorphic buffer between the GaAs substrate and the active transistor layers. This approach enables high-indium channel materials while retaining a GaAs substrate. Compared with InP, GaAs offers manufacturing and wafer-size advantages, along with improved power-handling capability. It can therefore be attractive when process stability, repeatability, cost, or production volume matters alongside low noise.
The choice is not simply a ranking of one technology over another. It reflects the receiver’s priorities: the lowest achievable noise may favor InP, while production considerations and power handling can make GaAs a practical alternative.
Key design decisions in a millimeter-wave low-noise amplifier
Designers may follow a bottom-up process, starting with the available technology and transistor, or a top-down process that begins with chip-level requirements such as gain, noise figure, power consumption, and input and output matching. In practice, both approaches are iterative: a later circuit decision can require earlier assumptions to be revisited.
Device geometry and bias
Transistor gate width affects the impedances available for matching and the device’s output-power capability. At millimeter-wave frequencies, a total gate periphery around 50–100 µm is commonly used for low-noise operation in a common-source stage. Implementing that periphery with a small number of fingers helps limit parasitic reactance; too many fingers can make the device difficult to use at high frequencies.
Bias also sets a noise-versus-gain trade-off. The bias point selected for minimum noise is generally different from the one that maximizes gain. A low-noise design often operates the transistor at a relatively low fraction of its maximum current, accepting a gain compromise to reduce noise. The precise operating point depends on the transistor and the circuit requirements.
Source degeneration and stability
Source degeneration adds inductance between a transistor’s source and ground. In a common-source stage, this can reduce gain, improve stability, and help align the impedances preferred for noise performance and signal matching. It can also support wider bandwidth. As operating frequency rises, the inductance may be implemented with a short transmission line or may arise partly from the ground connection’s parasitics, rather than from a conventional spiral inductor.
Transmission lines and circuit topology
On-chip transmission-line choice affects loss, layout, grounding, and modeling. Microstrip is widely used and supported by established design tools, but its losses and substrate-related effects become more challenging as frequency rises. Grounded coplanar waveguide (GCPW) and coplanar waveguide (CPW) can reduce some ground-return and substrate-coupling challenges at higher frequencies. They can, however, require more complex layouts and may have less mature device and circuit models.
The circuit topology should match the receiver’s performance targets:
- Single-ended: A straightforward cascade of common-source stages. It is widely used when circuit simplicity and low noise are important; stage count depends on the required gain and operating frequency.
- Cascode: Combines common-source and common-gate devices to support wide bandwidth and gain, but requires careful DC biasing and design of the common-gate stage’s AC grounding.
- Balanced: Uses a splitter and combiner around two amplifier paths to improve input and output matching over broad bandwidths. The added components increase circuit area, and splitter losses can worsen noise figure.
As frequency rises, electromagnetic simulation becomes important for evaluating interconnects, small capacitors, and parasitic effects that can influence both matching and stability.
Performance trends and application implications
A demonstrated V-band, three-stage GaAs mHEMT LNA operating from 59 to 66 GHz achieved an average noise figure below 2 dB, with measured performance reported at approximately 1.9 dB. This illustrates that GaAs can deliver competitive results in a design constrained by industrial-grade technology and a defined operating bandwidth.
Published results summarized in the review show noise figure generally increasing with operating frequency. The reported trend is approximately 0.02 dB per GHz from around 80 GHz to much higher frequencies, while cryogenic equivalent noise temperature also rises with frequency, with results influenced by the device’s physical temperature. In the reviewed implementations, microstrip is common through W-band, while coplanar structures become more prevalent above it.
These figures are design context, not a substitute for evaluating a specific receiver. Measurement conditions, bandwidth, technology, temperature, and circuit architecture all matter when comparing LNAs.
Conclusion
Millimeter-wave LNA design is a system-level exercise in balancing sensitivity, gain, bandwidth, stability, power handling, and manufacturability. InP HEMTs are well suited to applications that place a premium on low noise, including cryogenic scientific receivers. GaAs mHEMTs offer a compelling alternative when production considerations and power handling carry greater weight. Careful transistor sizing and biasing, feedback, transmission-line selection, and topology are all essential to translating device performance into a reliable receiver front end.