Magnetic resonance imaging (MRI) is widely used in hospitals and clinics for medical diagnosis. World­wide, there are approximately 36,000 MR machines. At pre­sent, about 2,500 MR imaging units are sold worldwide every year (1). How they work? Oversimplified, they identify the protons of hydrogen atoms in tissues containing water molecules. The energy of an oscillating magnetic field stimulates the molecules in briefly applied resonance frequencies to the patient. The stimulated hydrogen atoms emit an RF signal that is then measured by a receiving coil. Gradient coils vary the main magnetic field to determine positional data, which is translated by a computer. And by the way, those coils, briskly switching on and off, produce the loud and, as anyone who has received the procedure would tell you, obnoxiously noisy environment of an MRI scan. The rate at which the stimulated atoms return to stability, shows up as contrast between tissues, thus providing an image used for diagnosis.

An MRI system requires a strong and uniform magnetic field for the procedure to work. The basic physical theory behind a MRI system was independently discovered in 1946 by both Felix Bloch (2) and Edward Purcell (3). Inside the nucleus of every atom, individual protons and neutrons spin around an axis. Since atomic nuclei have charge, this spinning motion produces a magnetic moment, which can be understood as a vector having a magnitude and direction, along the spin axis. The relative strength of this magnetic moment is a unique property for each element, and therefore determines its suitability for magnetic resonance energy absorption and detection. The hydrogen nucleus with one proton is the nucleus of choice in MRI because it possesses the strongest magnetic moment and is abundant in organic tissues. In Figure 1, we see that in the absence of an external magnetic field, the individual magnetic moments have no preferred orientation.

Figure 1. Magnetic dipoles in the absence of an applied magnetic field.

The field strength of the magnet is measured by Tesla (T). With the application of an external magnetic field B0 the magnetic moments will align with the external field with a preferred orientation to guarantee the lowest energy state. The spin angular momentum causes the magnetic moment to precess about the B0 axis as shown in Figure 2

Figure 2. Precession of the hydrogen nucleus in the B0 field

The frequency of precession is governed by the Larmor equation as:

Where f is the frequency of precession, γ is a characteristic constant that depends on the given nucleus, and B0 is the strength of the externally applied field. For hydrogen nuclei, B0  is given as 42.576 MHz/T. Thus, in a field strength of 3T, the hydrogen nucleus will precess with at a frequency of 127.728 MHz (B0=3T is about 60000 times the Earth's magnetic field strength).

Consider the application of RF radiation at the Larmor frequency to a sample of non-magnetic material in an applied static magnetic field. When the frequency of the RF radiation matches the Larmor frequency of the protons, we have resonance, the most efficient state to exchange energy with protons (absorption and emission).

The applied RF radiation is composed of coupled electric and magnetic field components. The magnetic field component is denoted by B1, and it resides in a plane perpendicular to B0 while rotating about B0 at the Larmor frequency as shown in Figure 3.

Figure 3. Magnetization M rotating about B1 when the RF pulse is present

Applying B1 causes M (magnetic moment) to rotate by a certain angle away from the B0 axis.

Hence, if B1 of the RF pulses persists for the appropriate duration of time, M can be made to rotate onto the transverse plane.

While in the transverse plane and rotating at the Larmor frequency, M will induce an NMR (Nuclear Magnetic Resonance) signal in the RF receiver coil, which is oriented in the transverse plane as shown in Figure 4.

Figure 4. M rotates on the transverse plane after the application of a 90◦ RF pulse

When the RF pulse vector is switched off, M rotates in the transverse plane at the Larmor frequency and gradually decays to zero with time constant T2. Similarly, we will have a decay to zero with the time constant T1 that returns the protons to the original direction of the field B1. Both time constants depend on the nature of the body tissues under investigation.

Figure 5. MRI System configuration.

The strength of the magnetic field can be altered electronically from head to toe using a series of gradient electric coils, and, by altering the local magnetic field by these small increments, different slices of the body will resonate as different frequencies are applied.

In 3-Tesla (3T) MRI scanners, the required frequency should be around 127 MHz; some manufacturers still consider a 3T MRI to be achieved even when the Larmor frequency considered is slightly lower. In the design example presented below, the central frequency is 123 MHz.

RFPA architecture for MRI systems

For MRI systems, RF power amplifiers (RFPAs) are optimized for pulse operation to provide high RF power pulses within a short period of time. This amplifier uses two STAC4932B-100V N-channel MOSFETs from STMicroelectronics (4) in a push-pull configuration capable of 1.2 kW, each under pulse conditions, with each STAC244B (5) using a bolt-down air-cavity package (see Figure 6).

The design specifications for the 3T MRI project are:


In order to reach an RF peak-power level of 2 kW, it was decided to employ two STAC4932B MOSFETs in a double push-pull class AB configuration (see Figure 6).

Figure 6. STAC® packaging.

Figure 7. Push-pull topology.

The STAC4932B is an N-channel MOS field-effect RF power transistor. It is intended for 100 V pulse applications up to 250 MHz. This device is suitable for use in industrial, scientific, and medical applications (ISM). The STAC4932B benefits from the STMicroelectronics air-cavity STAC® packaging technology. In the double push-pull schematic, each of the STAC4932B high- and low-side devices contains two dice. From the input RF point of view, each STAC4932B is seen as a single gate. The design must ensure each gate has an equal and independent signal level (about 20 W shared by four gates), and at the same time prevent any additional gate interaction that may cause dangerous instability issues. The PCB's layout was especially designed to minimize any electrical asymmetry at the input of the devices.

Biasing and Stability Control

Several circuit topologies were taken into consideration for synthesizing the double push-pull amplifier (PPA). To create a compact design, our conclusive idea was to realize one single RF output balun transformer to serve the four drains (T2 in Figure 8). For the input, we used an input balun transformer (T1) in combination with two in-phase power splitters (L4, L8, C16, C18, C20 and L12, L16, C36, C41, C45) in order to guarantee proper isolation while the RF signal drives each gate. The balun transformer T1 and the input power splitter sections were carefully designed to respect the electrical symmetry.

Figure 8. 2-kW PPA electric circuit.

The RF LC decoupling filters fed through the VG1 and VG2 connectors (see Figure 8) are needed to bias each of the STAC4932B gates. The output biasing network includes the center tap of the primary winding of T2 and several multilayer ceramic capacitors. The capacitors charged at 100 V deliver the high DC current to the device’s drains and also dampens the voltage overshoot generated by the pulsed RF modulation, with the electrical transients controlled by L10, R13, C29, C30 and C33. Two test points can be inserted between two calibrated Rm resistors for current / voltage monitoring and an LED D1 for safety purposes during testing.

Design and Implementation

The main features of the 2-kW PPA are the planar implementation of the balun transformers (T1 and T2 embedded in PCB) and the SMT components as well as the lack of the ferrite elements in either. The PCB is embedded on a robust base plate (100 x 150 mm) to form a compact board that can withstand high RF power and high voltage (Vcc = 100 V). The 2-kW PPA amplifier layout is built on the Roger 4350B substrate with separate input-output PCB cards. The input PCB (see Figure 9) integrates the RF balun transformer T1 and the RF decoupling networks.

Figure 9. RF balun transformer T1 (Top and Bottom view).

The balun transformer T1 is λ/4 - 25 Ohm transmission line @ 123 MHz made using a strip-line technique on a 3-layer substrate (Roger 4350B, with a thickness of 20 +20 mils, see Figure 9) and is fed by a suspended 50 ohm microstrip line ('line bridge' in Figure 8).

In summary, the transformer T1 plays several important roles:

  1. Balun function and lowering the impedance;
  2. A quasi-one-dimensional RF structure (the λ/4 - 25 Ohm transmission line) on the PCB surface that allows the electrical symmetry to feed the RF to all gates.

The PCB shown in Figure 10 is the layout of the output RF transformer T2 with the remaining input power splitter and the biasing/filtering networks for the STAC4932B’s gates and drains.

Figure 10. RF balun transformer T2 (Top, left, and bottom, right).

The transformer T2 (ratio 4:1) is designed on the top/bottom layers (Figure 10) using a substrate Roger 4350B of 60 mils thickness in suspended broadside coupled strips and acts as a composite transmission line transformer from balanced to unbalanced mode. The RF output (type N-female connector) is directly connected to the T2 winding output strip (top view, Figure 10) through an air-suspended strip-line (50 Ohm). This allows the differential current flowing on the primary winding strip (PCB top layer) to move to the unbalanced RF output without power losses.

In this way, transformer T2:

  1. Provides the optimal impedances to the DMOS drains;
  2. Feeds high DC current filtered at Vcc = 100 V through the center tap of the winding top strip of T2;
  3. Carries out a ground current path to the RF output load in order to circumscribe the RF grounding issues.

The transformer T2 has been designed using EDA Software (ADS Keysight), and the refinement between electromagnetic (HFSS) and circuit simulation (see Figure 11). T2 uses C25, C26, and C23 on the winding top strip, and C37 and C42 on the output secondary strip, which makes it possible to tune drain impedance levels. On the copper carrier, two deep cavities (7 mm) accommodate the transformers T1 and T2; additionally they minimize the parasitic impedance (leakage) to ground.

Figure 11. Architecture of T2 and its essential transmission line view.

Figure 12 shows the assembled board and Table 1 lists the parts.

12. 2-kW MRI final assembled board.


Component ID



Part code

C12, C13, C6

1000 µF, 100 V



C10, C11, C52, C53

100 nF




4.7 µF, 100 V




15 µF, 100 V



C15, C51

10 µF, 35 V




100 µF, 20 V



C4, C7, C48, C49

22 µF, 25 V



C18, C41

300 pF



C16, C20, C36, C45

68 pF




1000 pF



C57, C58

3.3 pF



C8, C31, C32, C50

470 pF


ATC 100B 471FWN200XC

C2, C17, C24, C34, C40, C55, C59, C60, C61, C62, C63, C64

2000 pF


ATC 200A202KTN50C

C9, C14

1 µF, 100 V




1 µF, 100 V




18 pF




75 pF



C26, C23

100 pF



C37, C42

56 pF




700 nH



L2, L5, L13, L15

82 nH



L3, L14

110 nH



L4, L8, L12, L16

5.4 nH



R1, R5, R28, R30

50 Ω, 100 W




5600 Ω

Tyco Electronics



22 Ω

Tyco Electronics


R7, R29

100 Ω



R11, R22

4.7 Ω


4.7 Ohm -1206

R3, R24

43 Ω



R32, R33, R34, R35

27 Ω



R4, R16, R17, R31

20 Ω



R6, R8, R10, R12, R14, R15, R19, R20, R21,R25,R26, R27

1 Ω



Rm x 2

0.001 mΩ

Tyco Electronics

TL3A R001 1%


1 double pole








3 poles



















Line bridge

Roger 4350B, three layers, 20+20 mils, 1 OZ Cu on top-mid-bottom layers, Finit. metal HAL LF; total Tk=1.2 mm max., top screen printing component, tin chemical surface deposition.

Board input

Roger 4350B, three layers, 20+20 mils, 1 OZ Cu on top-mid-bottom layers, Finit. metal HAL LF; total Tk=1.2 mm max., top screen printing comp., tin chemical surface deposition.

Fin fixing

Roger 4350B, two layers, Tk=60 mils, 1 OZ Cu on top- bottom layers, Finit. metal HAL LF; total Tk=1.6 mm max., top screen printing comp., tin chemical surface deposition.

Board output

Roger 4350B, two layers, Tk=60 mils, 1 OZ Cu on top-bottom layers, Finit. metal HAL LF; total Tk=1.6 mm max., top screen printing comp., tin chemical surface deposition.

Mechanical plate

PPAMRI_002-Rev B


Table 1. Bill of materials.

Thermal Features of Air Cavity packages

The ST air cavity (STAC®) technology allows very low thermal impedance: Ztjc = 0.075 K/W (with t = 1 ms pulsed RF, Duty Cycle = 10%). Combining it with an appropriate heatsink (Rth <0.2 K / W) lowers the junction temperature below the device’s maximum rating (Tjmax = 200 degC).

The ability of STAC® to dissipate a high power pulse makes it possible to reduce board dimensions and external heat-sinks; so that, by using the flangeless package STAC244F shown in Figure 13, one can design a new board with the necessary electrical characteristics but with a dimension target of 80 x 100 mm or less.

Figure 13. Air cavity package.

Prototype RF Performance Measurements

Figure 14 and Figure 15 show measurements taken in pulsed condition (tpulsed = 1 ms and period 10 ms) with the power amplifier operated in class AB (100 V x 200 mA). In Figure 14, the power gain and the input return loss is depicted versus frequency @ Pin = 20 W. In Figure 15, we see the power gain and the drain efficiency versus the output power @ f = 123 MHz. The drain efficiency of 60% is reached at 2.2 kW RF power.

Figure 14. Power gain and input return loss over frequency.

Figure 15. Gain compression and Eff (%) Vs Pout @ f = 123 MHz.


A pulsed RF high power amplifier (> 2 kW) has been described as a guideline-design, oriented to new high-voltage DMOS devices STAC4932B in an air-cavity package (STAC®) at Vcc = 100 V. In particular, the amplifier combines excellent high frequency response with an efficient use of DC power. Due to the combination of SMT components and the fully planar technology of the RF transformers, a very compact and robust design was produced. This amplifier is the single unit of a high-power RF chain, and by combining some of them, we can achieve the very high power level required for an RF pulse generator in 3T-MRI systems.


Author1Roberto Cammarata received his Master’s degree in Electrical Engineering from the University of Catania, Italy. He began his career in 1985 working in the microwave field on hearth satellite ground stations (Selenia - Marconi Comm., Catania). Afterwards, he worked on Space Satellite Projects (Alenia Aerospace , Rome) before joining STMicroelectronics (Catania) to design power amplifiers for Communications and ISM RF applications.


Author2Alfio Scuto received his Master’s degree in Microelectronics Engineering from the University of Catania, Italy. In 1999, he joined ST's RF Power Design Center in Montgomeryville (PA) USA. As an RF Application Engineer, he was involved in the characterization of high-power and high-frequency transistors. In 2002, he moved to STMicroelectronics in Catania (Italy) to support the RF products marketing group developing power amplifiers in order to evaluate and verify product performance in relation to customer specifications. Today he works as a High-Power Application Engineer supporting customers and exploring new applications for high-power MOSFET devices.


1. Magnetic Resonance, a critical peer-reviewed introduction. [Online] 17 November 2014.

2. Nuclear induction. Bloch, F. s.l. : Physical Review, 1946, Vols. vol. 70, pp. 460-473.

3. Resonance Absorption by Nuclear Magnetic Moments in a Solid. E. Purcell, H. Torrey, and R. Pound,. s.l. : Physical Review,, 1946, Vols. vol. 69, pp. 37-38.

4. STMicroelectronics. STAC4932B.

5. Mounting recommendatios for STAC and STAP boltdown packages. AN3232.

6. Grebennikov, Andrei. RF and Microwave Power Amplifier Design. s.l. : Mc Graw Hill, 2005.

7. Holzman, Eric. Essentials of RF and Microwave Grounding. s.l. : Artech House, 2006.

8. Magnetic Resonance Imaging. Wright, C. A. s.l. : IEEE SIgnal Processing Magazine, January 1997.

9. BRUSH UP THE THEORY TO DESIGN AN HIGH POWER CLASSE-E AMPLIFIER. Alfio Scuto, Roberto Cammarata. s.l. : MICROWAVE - Engineering Europe, June 29, 2016.