GaN Power Transistors: How To Compare RF Devices

Many engineers working on RF amplifiers are easily led by the power figures on paper when selecting GaN (gallium nitride) tubes from a selection table. When two pipes with similar book parameters are actually welded to a plate, their measured performance often varies greatly. The problem actually lies in the test conditions that are ignored behind the numbers: test frequency band, excitation waveform, static bias, impedance matching, and junction bottom temperature. A truly reliable selection logic must be extrapolated from actual system indicators and all candidate tubes must be compared under the same set of benchmark conditions.

First, Find Out Where The Power Reference Surface Is Located

When evaluating power requirements, we cannot just focus on the center frequency point; we must take the entire operating bandwidth into account. More importantly, it is necessary to clarify whether the output power required by the system is set at the core pins, module connectors, or after the final filter. The transmission and filtering interpolation between these nodes is really there. If the interpolation difference is a few decibels zero, the power of the front-stage tube will have to be multiplied to the outside top.

If it is a multi-carrier or complex modulated signal, in addition to the nominal peak value, the peak-to-average ratio, available average power, and linearity bottom line must also be clearly defined; if it is a pulse application such as radar, in addition to the peak power, the pulse width and duty cycle must be strictly limited.

By clarifying this step, we can also avoid a common misconception: confusing RF GaN tubes with power switch GaN tubes. Although it is called gallium nitride, RF amplifiers value gain, saturation/linear power, efficiency, matching network sensitivity, stability, and intermodulation distortion throughout the entire operating frequency band; while the voltage resistance value of power converter components is usually (
) and on-resistance (
), which has little reference value for evaluating microwave band amplification performance.

Disengage The Test Waveform And Bias To See The Saturation Power (Psat), Which Is Basically Equal To The Blind Man Touching The Elephant

In the specifications
This represents only the device’s ultimate ability to push near the saturation edge, which is not equal to the 1 dB compression point (
), which does not even indicate how much effective average power can be output under certain linearity constraints. If it is a continuous wave (CW), it reflects the ultimate capacity under thermal equilibrium; if it is measured with pulse width and duty cycle, it is a completely different set of heat conduction and electrical stress models. It is very easy to choose the wrong device just by looking at the wattage.

Take the radio frequency selection manual of Neditek (Neditek) as an example:

• NDNM011813 covers 1.1–1.7 GHz and has a nominal CW saturation power of 220 W;
• NDNM01215 covers 1.2–1.4 GHz, nominal
  It reaches 400 W, but is limited to pulse conditions with a pulse width of 6 ms and a duty cycle of 30%.

The 400 W looks much better than the 220 W, but it cannot replace continuous wave operation at all. If it were replaced with other duty cycles or re-frequency, performance would also be questionable. Before locking the device, be sure to check the latest Datasheet to verify the test frequency point, drain bias, seat temperature, peripheral matching circuit topology, and precise power definition method corresponding to its calibrated power.

Figure Out Which Step The So-Called “Internal Matching” Has Reached

Many package-grade RF GaN tubes claim to come with either “pre-matching” or “internal matching”, which does help simplify the design of external matching networks, but this by no means means that it can directly present a perfect 50°C across the entire passband and at different input power levels
. During the selection phase, the manufacturer must obtain reference circuit board diagrams, small signal S parameters calibrated on the reference surface, and large signal load pulling (Load-pull) data in the target frequency band. When transplanting evaluation fixtures from the manufacturer to its own multi-layer production plates, differences in the test reference surfaces must be corrected in advance.

When making a wideband amplifier, you should be more vigilant: data with a beautiful center frequency point often masks problems such as drop in gain at the inner edge of the band, power shrinkage, or a significant drop in efficiency. Be sure to pull through the full frequency band to see the worst-case scenario (Worst-case), and compare it with the ambient temperature at the same bias. If it is a linear transmitter, directly using the target modulation waveform to measure the drain efficiency after EVM, adjacent channel leakage ratio (ACLR) and power backoff (Back-off), and using the CW saturation index to guess the linear working state has no engineering significance.

Starting From The Preliminary Assessment, The Heat Dissipation And Power Supply Timing Should Be Included

The junction temperature of GaN depends on the complete heat dissipation chain: core → package flange → thermally conductive interface material (TIM) → amplifier heat dissipation base plate. Assembly process, screw locking torque, thickness of thermal pad or thermal paste, and heat dissipation capacity of casing air duct and base plate are all directly reflected in the radio frequency output. Pulse width and duty cycle have a significant impact here——tubes that can reach very high peaks under short pulses, when the pulse is prolonged or the duty cycle increases, thermal resistance accumulation will cause the junction temperature to rise rapidly, and the output power and reliability will then plummet.

Bias timing must also be sloppy. Currently, most depletion-mode (RF GaN HEMTs) on the market are normally open devices. When powering up, the timing of “first negative bias of the gate voltage, then high voltage of the drain” must be strictly executed. The reverse is true when turning off, otherwise straight-through tube burning is very likely to occur. The driving habits of enhanced (normally closed) GaN commonly used in the power supply field must never be copied into the radio frequency system.

The final selection conclusion must not be based on just a paper parameter table, but must rely on a standardized and reproducible benchmark test record: clearly list the test frequency points, input and output reference surface positions, signal waveform standards, gate-drain bias voltage and static current, base plate controlled temperature, load standing wave conditions, and the power, gain, efficiency and linearity indicators measured under this operating condition. Only by establishing lateral comparisons under equivalent boundary conditions can the data on the selection table truly become stable performance in the actual circuit.