Control synchronization of power supplies and gradient amplifiers for better utilization of energy storage in MRI scanner
Yash-Veer Singh1, Juan A Sabate1, Ruxi Wang1, Viswanathan Kanakasabai2, Krishna Mainali1, and Huan Hu1

1GE Global Research, Niskayuna, NY, United States, 2GE GR JFW Technology Center, Bengaluru, India


The magnetic resonance imaging (MRI) gradient driver is a high-performance inverter that provides large peak currents, >1000A, for the gradient magnetic fields used for imaging. MR system has capacitors for energy storage at the amplifier stage input voltages. Volume/weight reduction in energy storage elements is one of the challenging issue in design of compact and efficient MRI gradient system. It is possible to reduce capacitance value in energy storage elements by synchronizing the control of the gradient driver different subsystems. Control synchronization is achieved by providing reference value of coil current simultaneously to the different controller of the MRI system.


Gradient amplifier (GA) to control magnetic field gradient in MRI scanners are switched power amplifiers implemented with full bridge inverters connected in series [1]-[2]. The amplifiers are fed from power supplies (PS) with multiple isolated outputs. The pulsating power requirements for the gradient sequences make energy storage a key component in the circuit structure of these gradient amplifiers [3]-[4]. Volume/weight reduction in energy storage elements is one of the challenging issue in design of compact and efficient MRI gradient system. The energy storage elements between PS and GA act as a buffer during fast transients, high di/dt. Energy storage at the input of the PS provides power support during the imaging sequence. Power supply stage regulates the front-end voltages of GA and delivers the power corresponding to the losses in the gradient coil and the GA. Energy storage elements will supply peak power demanded by the imaging sequence. It is possible to reduce capacitance value in energy storage elements by synchronizing control different subsystems of MRI scanner. Control synchronization is achieved by providing gradient coil current refence to all the stages -GA, PS and PDU- controllers simultaneously to calculate control response needed without waiting for the voltage dip during normal operation. The synchronization helps to limit the peak power flow from the utility with less capacitance.

Method of Control Synchronization

Block diagram of different subsystems in gradient chain of MRI system is shown in Fig.1(a). Power distribution unit (PDU) is connected to the grid. This PDU can be either low frequency transformer with taps on primary to serve different input voltage conditions or it can be another power conversion stage (high frequency). Each GA in Fig.1(a) has capacitive energy storage element, C1 for GA-X. Conventionally independent controls are used in gradient chain of MRI system as shown in Fig.1(a). These controllers utilize local reference value and feedback signals to generate controller output. Synchronized control platform as shown in Fig.1(b) enables faster calculation of required controller output and reduces the energy storage need in gradient amplifier by supplying coil current reference to all the different control blocks simultaneously. Schematic of the synchronized control is shown in Fig.3. Common energy storage, either capacitors or capacitor with electronic converter, is placed at the input of PS. Mathematical model of loads in each of the power stages is included in the control and reference value of coil current, Icoil_reference, is used as common signal to all three controllers and processed with these mathematical models to calculate feedforward control command for their respective controller.

Results and Discussion

Simulation of full gradient chain is designed in PLECS (circuit simulator). Reference value of PS output voltage is 1250V dc. Output voltage of PS drops by 75V on each dc port when coil current reference, 1300A flat-top, is provided as shown in Fig.2(a). Using synchronized control, for the same gradient coil current command, voltage dip in output voltage of PS is reduced to 32V as shown in Fig.2(b). Because now voltage dip is reduced significantly and therefore requirement of capacitance in dc link of amplifier also reduced. For the same gradient coil current command and with 3 mF capacitance on each input dc port of GA instead of 7mF, output voltage dip to 75V is shown in Fig.2(c). Synchronized control enables reduction of capacitance in gradient amplifier. A full-scale lab prototype has been tested. For the same coil current reference (1300A flat-top), output voltage of PS falls to 192V in case of conventional controller and with passive energy storage of 54mF in front of PS as shown in Fig.4(a). Voltage drop of 80V is noticed at the output of HFPDU in the common ES stage. Simulation of the same test condition is shown in Fig.4(b). Experimental results show good match with simulation results.


No acknowledgement found.


[1] J.Sabate, et al., Proc. ISMRM. 2007. [2] R.Wang. et al., Proc. ISMRM 2018. [3] J.Sabate, et al., Proc. European Power Electronics Conf. 2007. [4] Y.Singh, et al., Proc. IEEE-Applied Power Electronics Conference 2013.


Fig.1: (a) Gradient driver system with individual control. (b) Gradient driver system with synchronized control.

Fig.2: (a) simulation results with individual control. (b) simulation results with synchronized control. (c) simulation results after capacitance reduction (7mF à 3mF on each dc port of gradient amplifier).

Fig.3: Schematic of synchronized control scheme for different subsystem of MRI scanner.

Fig.4: (a) Experimental result of 1300A long flat-top pulse. (b) Simulation results of 1300A long flat-top pulse.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)