G16.4427 Practical MRI 1 Transmit Arrays

Outline Paper Review RF power amplifiers Dual-tuned coils

Paper Review

Sidebar: Component coil combinations in arrays ?

Sidebar: Component coil combinations in arrays = * + * + * + * Optimal combination: Sum of squares combination: =( * + * + * + * )1/2

Component coil combinations and signal-to-noise ratio Matched filter effect: Sum Coil #2 Coil #3 Coil #4 Unfiltered Coil #1 Filtered

Component coil combinations and signal-to-noise ratio Generalized quadrature effect:

SENSE as a generalized optimal coil combination Matched filter combination Noise decorrelation

RF Power Amplifiers RF Power Amplifiers (RFPAs) are a vital sub-system of any NMR spectrometer or MRI scanner The sole purpose is the amplification of the RF pulse Many power amplifiers may be required to achieve higher power output levels Divider/combiner networks sum multiple stages If the RFPA was ideal, the output would be an exact replica of the input waveform with greater amplitude Conventional RFPA are not ideal and distort the signal

RFPA Architecture Pre-Driver and Driver are low power amplifier stages that raise the power level of the input signal from mW to a level high enough to drive the high power PA sections Directional coupler separates out proportional samples of forward and reflected power for internal/external power monitoring and fault detection The microcontroller is a micro-computer that continuously runs a fixed program loop that monitors several vital operating parameters (e.g. DC voltages, currents, pulse width, etc.) If there is a risk of damage, it will put the system into fault mode The DC power supply converts AC line voltages into DC voltages that are suitable to operate the Pre-Driver, Driver, Power Amplifiers and microcontroller

Actual RF Pulse It takes 19 parameters to characterize a pulse that has been through a non ideal amplifier It takes 4 parameters to define an ideal RF pulse. What are they?

Actual RF Pulse It takes 19 parameters to characterize a pulse that has been through a non ideal amplifier It takes 4 parameters to define an ideal RF pulse: Amplitude Frequency Pulse width Duty factor

Actual RF Pulse Parameters

RFPA Specifications: Time Domain We want very high RF power pulses with precise fidelity only for short periods of time Maximum pulse width (during which the RFPA can put out maximum power) is 20-300 ms for MRI Average power requirements (duty factor) ~10-15% maximum Pulse pre-shoot, post pulse backswing Distortion occurs after an RFPA has been un-blanked (or RF pulse is terminated) Appears as half or more cycles of a low frequency signal superimposed on the un-blanked noise voltage Low frequency, so it will be filtered out by the transmit coil

Pulse Pre-Shoot, Post Pulse Backswing pre-shot

RFPA Specifications: Time Domain Very high RF power pulses with precise fidelity for short periods of time Maximum pulse width (during which the RFPA can put out maximum power) is 20-300 ms for MRI Average power requirements (duty factor) ~10-15% maximum Pulse pre-shoot, post pulse backswing Distorsion occurs after an RFPA has been un-blanked (or RF pulse is terminated) Appears as half or more cycles of a low frequency signal superimposed on the un-blanked noise voltage As low frequency, it will be filtered out by the Tx coil Rise, fall time (transition duration) Time to transition from 10% to 90% of the voltage waveform Specification for MRI: 250 nsec to 10 μsec

Pulse Transition Duration Fall time Rise time

RFPA Specifications: Time Domain Overshoot, rising/falling edge Distortion occurs from inductively stored energy within the RFPAs circuitry (transition from zero to full power in ~100 ns  voltage spike due to large current changes in inductors get superimposed on the RF pulse) Specification for MRI: < 13%

Overshoot, Rising/Falling Edge pulse overshoot Pulse falling edge Pulse rising edge Falling pulse overshoot

RFPA Specifications: Time Domain Overshoot, rising/falling edge Distortion occurs from inductively stored energy within the RFPAs circuitry (transition from zero to full power in ~100 ns  voltage spike due to large current changes in inductors get superimposed on the RF pulse) Specification for MRI: < 13% Pulse overshoot ringing/decay time Energy being fly between inductive and capacitive circuits in the RFPA generates a lower frequency dumped sinusoidal wave that is imposed on the RF pulse after the rise time and modulates its amplitude Specification: time for the amplitude modulation to drop to less than 5% of peak RF pulse amplitude < 5 μsec

Pulse Overshoot Ringing/Decay Time

RFPA Specifications: Time Domain Overshoot, rising/falling edge Distortion occurs from inductively stored energy within the RFPAs circuitry (transition from zero to full power in ~100 ns  voltage spike due to large current changes in inductors get superimposed on the RF pulse) Specification for MRI: < 13% Pulse overshoot ringing/decay time Energy being fly between inductive and capacitive circuits in the RFPA generates a lower frequency dumped sinusoidal wave that is imposed on the RF pulse after the rise time and modulates its amplitude Specification: time for the amplitude modulation to drop to less than 5% of peak RF pulse amplitude < 5 μsec Pulse tilt (positive or negative) Gain change due to temperature increase in “on” RF transistors Specification for MRI: < 8% over 20 ms rectangular pulse

Pulse Tilt (Positive/Negative)

RFPA Specifications: Time Domain Long term amplitude/phase stability Ideally would amplify every pulse exactly the same way Changes in environment (e.g. temperature) can alter RFPAs Specifications for MRI: amplitude < 0.2 dB, phase < 3 degrees over 24 hours at constant temperature Phase error over-pulse Occurs as a phase shift across the duration of a rectangular pulse in cases when the pulse tilt is substantial Specification: < 5 degrees across a 10 msec pulse width Un-Blanking, Blanking propagation delay To reduce electronic noise during signal acquisition, the output stage of an RFPA are shut off Delay measures the ability of an RFPA to rapidly turn on/off Specification for MRI: 2 μsec

Pulse Overshoot Ringing/Decay Time Un-Blanked noise voltage Blanked noise voltage

RFPA Specifications: Frequency Domain Generic frequency domain specifications The bandwidth is the range of frequencies for which the RFPA complies with output power, linearity, etc. Power gain Specification: maximum peak power when maximum output power is required Gain flatness The wider the bandwidth the harder is to maintain constant power gain  flatness at key frequencies Specifications: broadband = ± 3 dB, nuclei centered = ± 0.2 dB at ± 500 kHz

RFPA Specifications: Frequency Domain Harmonic content Practical RFPAs are not perfectly linear  output frequency spectrums also at integer multiples of the input frequency Mostly filtered out by the transmit coil Specification: even/odd order harmonics = -20 db/-12 dB Spurious RF output emissions (oscillation) Erratic frequency components that the RFPA puts out (e.g. DC feed that couples RF power from output to input) Specification: < -50 dBc Input VSWR How close the input impedance is to an ideal 50 Ω resistor Specification: < 2:1 (perfect match = 1:1)

RFPA Specifications: Frequency Domain Output noise (blanked) To minimize electronic noise, the bias of the transistors of the final stages of power amplification are shut off (there will still be some tolerable noise output) Specification: -20 dB over thermal noise Noise Figure In applications where RFPA is transmitting at one frequency and RF receivers are listening at another, the less NF an RFPA has, the less will interfere with this second frequency Specification: < -10 dB

RFPA Specifications: Power Domain The input power to the RFPA is swept across a range of power levels (usually 30-40 db) E.g.: if an RFPA is driven to full power at 0 dBm input (i.e. 1 mW), the unit will be tested for input -40 to 0 dBm to check for phase linearity and gain RF power output 1T-3T: 0.5-2 kW extremities (legs and arms), 4-8 kW head, up to 35 kW whole body Higher field strengths: 10-20 kW is common Multi-channel: 4 kW (3T) and 1 kW (7T) per channel

RFPA Specifications: Power Domain Gain linearity Defined in terms of dynamic range (from maximum specified output power level to some dB down from such level) Specification: ± 1 dB gain variation over 40 dB dynamic range Phase linearity Although it takes few nanoseconds for the signal to go from input to output of the RFPA, there is a propagation delay In ideal case the phase shift is constant across the dynamic range Phase non-linearity is a due to parasitic junction capacitance present in all types of RF power transistors (change with output power) Specification: ± 7.5 degrees phase variation over 40 dB dynamic range

Gain Linearity Pulse sequences can contain RF waveforms that have precisely proportioned amplitude ratios, which can change dramatically in case of severe deviation from ideal gain linearity In the non-ideal case, the transfer function changes over the dynamic range of the amplifier  power levels will be amplified by different power gain factors

Troubleshooting Amplifier Performance Anomaly Symptom Excessive gain non-linearity Slice profile distortion Excessive phase non-linearity Excessive rise/fall time Gain instability Image ghosting/shading Phase instability Image ghosting Excessive pulse overshoot Spurious oscillation Image artifacts/streaking Low power output Inability to achieve desired flip angle

Any questions?

Dual-Tuned Coils A major problem of implementing multinuclear MRI is the construction of a probe capable of operating at more than one frequency To make a single-tuned coil resonant, we normally add a tuning circuit (a capacitor in series with the coil): In order to multiple tune a coil, we need to make the reactance curve of the tuning network cross the anti-reactance curve of the coil more than once

Double Resonant Circuit A useful tuning network consists of a parallel LC trap in series with the tuning capacitor network The reactance, as a function of frequency, will begin capacitive, then pass through a pole (trap resonant frequency) and then become capacitive again The reactance curve crosses the anti-reactance curve of the coil twice  two resonances are established

Matching Normally a reactive element is added in parallel to the series tuned network so that the input impedance to the entire network is real and equal to the generator impedance For dual-tuned coil, we can use a parallel LC matching network At the low frequency C is large enough so that we may consider only the inductor and adjust its value for proper matching The capacitor can then be tuned so that the parallel combination of C and L has the required reactance for matching at the higher frequency

Example: Dual-Tuned Birdcage at 1.5 T The fourth harmonic of the sodium frequency is very close to the proton frequency at 1.5 T (67.8 MHz vs. 64 MHz) It is challenging to decouple the two channels in a birdcage Modified inductive coupling circuits (with baluns) are used to provide better decoupling The trap circuit method is used to obtain identical current distributions for both resonance frequencies Same B1 field distribution

Any questions?

See you next week!