WO1991019993A1

WO1991019993A1 - Drive circuitry for field-generating coil of magnetic resonance imaging system

Info

Publication number: WO1991019993A1
Application number: PCT/US1991/004238
Authority: WO
Inventors: Stephen Crump; Richard Rzedzian
Original assignee: Advanced NMR Systems Inc
Current assignee: Advanced NMR Systems Inc
Priority date: 1990-06-13
Filing date: 1991-06-13
Publication date: 1991-12-26
Anticipated expiration: 1992-12-13

Abstract

Feedback control is provided for the resonant circuit (22) formed by the capacitive element (12) and field-generating coil (10) of a MRI system. A current sensing circuit (24) is connected to the resonant circuit for sensing the current flowing through the coil (10). Feedback circuitry compares the sensed current to a reference signal (28) (e.g., the desired sinusoidal current waveform), and provides an output signal prescribing the power to be supplied to the resonant circuit (22) to cause the current flowing through the coil to more closely approximate the reference current.

Description

DRIVE CIRCUITRY FOR FIELD-GENERATING COIL OF

MAGNETIC RESONANCE IMAGING SYSTEM

Background of the Invention

This invention relates to producing modulated gradient fields in a magnetic resonance imaging (MRI) systems.

A variety of modulated gradient field waveforms are known to be useful in MRI systems. For example, pulses of constant amplitude are used in certain

applications, while pulses of sinusoidal oscillations are used in others.

Previous MRI systems using constant amplitude pulses have employed feedback control to achieve accurate current delivery to the gradient coil, but feedback control has not heretofore been achieved in MRI systems in which sinusoidal oscillations are produced. In these systems, such as disclosed in Rzedzian U.S. Patent No. 4,628,264, the magnetic coil forms a resonant circuit with a storage capacitor, and the gradient coil has been driven "open loop" i.e., with a predetermined sinusoidal signal that, based on calculation and experience, is expected to produce the desired sinusoidal oscillation in coil current.

Summary of the Invention

The invention features feedback control of the resonant circuit formed by the capacitive element and field-generating coil of an MRI system. A current sensing circuit is connected to the resonant circuit, for sensing the current flowing through the coil. Feedback circuitry compares the sensed current to a reference signal (e.g., the desired sinusoidal current waveform), and provides an output signal prescribing the power to be supplied to the resonant circuit to cause the current flowing through the coil to more closely approximate the reference current. In preferred embodiments, the

capacitive element and power source are connected in series; the reference signal comprises a plurality of sinusoidal segments and at least one flat segment; the error between the sensed current and the reference signal is integrated.

In a second aspect, the invention features a method for charging a storage capacitor connected in series with the magnetic coil of an MRI system. A switch circuit connected in series with the capacitor and coil is opened to isolate the capacitive element from the coil. A charging circuit is connected across the

capacitive element. The level of energy stored in the capacitive element is compared to a reference level, and, based on that comparison, additional energy is delivered from the charging circuit to the capacitive element, until the energy stored in the capacitive element

approximates the level prescribed by the reference signal. In preferred embodiments, the capacitor is charged at two charging rates, to reduce parasistic effects and the errors they induce; the capacitor is first charged to a fraction of the desired level at a relatively higher rate, and then charged to the desired level at a relatively lower rate.

The invention provides more accurate control of the current delivered to the coil than has heretofore been achievable.

Other advantages and features of the invention will be apparent from the following description of a preferred embodiment, and from the claims. Description of the Preferred Embodiment

Fig. 1 is a simplified block diagram helpful for understanding the invention.

Fig. 2 is a block diagram of a preferred embodiment of the invention.

Fig. 3 is a timing diagram of signals produced by the structure depicted in Fig. 1 during the precharging of the capacitor and during the delivery of a sinusoidal pulse of current to the magnetic fieldgenerating coil.

Fig. 4 is timing diagram of signals produced by the structure depicted in Fig. 1 during precharging and during the delivery of a constant amplitude pulse to the coil.

As explained in detail in U.S. Patent Number 4,628,264 issued to Rzedzian, incorporated herein by reference, it is known in the art of MRI systems imaging to superimpose a modulated gradient magnetic field over a static magnetic field. Two useful waveforms for

generating the gradient field are a pulse of sinusoidal oscillations and a DC pulse of constant amplitude.

To achieve a sinusoidal oscillation in the gradient field, a sinusoidal current I_o is driven through the gradient coil. Ideally, I_o has the waveform: I_o = I_x · sin (W_ot) where W_e is the desired angular frequency and I_x is the peak current needed to yield the desired gradient field.

For a DC pulse of constant amplitude, I_o should ideally have a perfectly rectangular waveform.

However, since instantaneous changes in coil current are impossible, perfect rectangularity cannot be achieved. Thus, the drive current must gradually rise and fall at the beginning and end of the pulse.

Shown in Fig. 1 is a simplified block diagram of an embodiment of the invention. Coil 10, for

generating the gradient field, is connected in series with capacitor 12 and current-measuring shunt resistor 16. Parasitic resistor 14 (also shown in series with coil 10) represents the total circuit losses, which are primarily due to coil 10. Capacitor charging circuitry 18 is connected across capacitor 12 to pre-charge the capacitor prior to passing any current through coil 10. Control of coil current I_o is handled by control

circuitry 20. The load driven by control circuit 20 is represented by load 22, which consists of the series combination of coil 10, capacitor 12, parasitic resistor 14, and shunt resistor 16. Circuitry 20 controls coil current I_o by operating run switch 32 and by controlling voltage V_p across load 22. The resultant current I_o in coil 10 has a waveform determined by voltage V_p , the initial capacitor voltage V_co, and the impedance

characteristics of the load.

In one mode of operation, to energize coil 10 with sinusoidal current I_o, voltage V_p is chosen to have a sinusoidal waveform. To minimize the power voltage demand on control circuitry 20, load 22 is designed to have a resonant frequency equal to the desired frequency W_o of sinusoidal current I_o and to have a high quality factor Q. This allows control circuit 20 to achieve the desired high drive current I_o by maintaining V_p at the resonant frequency with a relatively low amplitude. More specifically, since the impedance of load 22 at its resonant frequency equals the pure load resistance, the drive voltage V_p needed to yield I_o is given by:

V_p = I_o · R_sum = I_x · R_sum · Sln (W_ot) where R_sum is the sum of the resistances of parasitic resistor 14 and shunt resistor 16. Accordingly, to minimize the power required from control circuitry 20, R_sum should be kept as small as practical.

To yield the desired magnetic field, a coil with a relatively large inductance is required. In preferred embodiments, a gradient coil with a measured inductance L of 1,025 microHenries is used. Accordingly, for load 22 to have a resonant frequency equal to a desired current frequency of 1.00 KHz, capacitor 12 must have a capacitance C of 24.7 microfarads.

To achieve accuracy in an MRI system using a sinusoidal waveform, I_x must remain constant. Toward this end, the series resonant system of Fig. 1 uses a feedback architecture which monitors coil current I_o and adjusts driving voltage V_p to maintain the desired coil current I_o. Current sensing amplifier 24 detects voltage V_s across a small, precisely known, shunt resistor 16 and amplifies V_s with gain G_a to produce feedback voltage V_f,, which is proportional to coil current I_o. Difference node 26 subtracts voltage V_f, from sinusoidal reference voltage V_R provided by signal generator 28. The gain G_a of current sensing amplifier 24 places feedback voltage V_f on the same scale as reference voltage V_R so that any difference between voltages V_f and V_R represents undesired error V_a in the magnitude of coil current I_o. This scale is chosen to maximize the dynamic range of the system.

Difference node 26 provides error signal V_a to the input of run controller 30, which operates on the error signal with a phase compensating transfer function designed to provide appropriate gain and phase

compensation to yield maximum stable feedback loop gain. In preferred embodiments, the phase compensation is provided with a transfer function of (s + w_o) /s, where s is the Laplace operator. This amounts to an integration of all error signal frequency components from DC to w_o. The resultant output control voltage V_b is applied to the input of power cell 34. The power cell 34 drives load 22 with voltage V_p, which tracks control voltage V_b. This feedback architecture extends the high precision of reference voltage V_R to coil current I_o.

As explained in Rzedzian U.S. Pat. No.

4,628,264, undesirable transients in coil current may occur when a stimulus voltage, such as V_p, is first applied to a tuned circuit such as load 22. Rzedzian discloses that such transients can be avoided by properly initializing the tuned circuit. More specifically,

Rzedzian discloses precharging a capacitor in the tuned circuit to a voltage corresponding to the peak capacitor voltage in the desired steady state.

To avoid transients, charging circuitry 18 precharges capacitor 12 to an initial voltage V_co equal to the peak capacitor voltage V_x generated when load 22 resonates with the desired coil current I_o. In the series topology of the present invention, the capacitor voltage signal V_c is given by:

where Z_c represents the capacitor's impedance. The resonant frequency W_o equals 1/√/LC, where L is the inductance of gradient coil 10 and C is the capacitance of capacitor 12. Thus it follows that:

V

To initiate the precharge mode, charging switch 36 is closed to connect current source 38 across capacitor 12. A feedback architecture is used to control current source 38 to closely match the magnitude of the initial voltage V_co to voltage V_x calculated by the above equation.

More specifically, voltage sensing amplifier 40 reads the voltage on capacitor 12 to provide a

measured capacitor voltage V_cf. The difference between measured capacitor voltage V_cf and a reference voltage V_L which prescribes the desired capacitor voltage) is fed as error signal V_e. to controller 42, which, in turn, supplies charge control voltage V_d to current source 38.

Fig. 3 illustrates the sequence in which the feedback architecture adjusts capacitor voltage V_c until the measured capacitor voltage V_cf precisely corresponds to reference voltage V_L. Controller 42 (Fig. 1)

initiates the precharge operation at a time t₁. by

asserting control signal C1 to close charge switch 36. A dual-rate charging scheme is used, in which the capacitor is first charged to within a few percent of the desired level at a high rate H (V_d at a relatively high

amplitude) and then charged more precisely at a lower rate L (V_d at a relatively low amplitude). The resultant reduction in charging current minimizes parasitic effects and the errors they induce. Once the capacitor voltage is substantially equal to V_x, charging is terminated at time t₂, by controller 42 opening charge switch 36, removing current source 38 from the circuitry driving capacitor 12.

Once capacitor 12 has been charged, run controller 30 initiates a sinusoidal pulse of current through gradient coil 10 by asserting control signal C2 at time t₃ . This closes run switch 32, and allows capacitor 12 to discharge into coil 10. Load 22

immediately begins to resonate at the resonant frequency W_o. with energy being transferred back and forth between the capacitor and the coil. At the same time, run

controller 30 provides energy, in the form of voltage V_p, to compensate for resistive losses. Run controller 30 terminates at time t^ the sinusoidal pulse by opening run switch 32 at the precise instant when capacitor 12 is fully charged and coil current I_o is zero, thus avoiding the need to recharge the capacitor for subsequent pulses.

The series configured drive circuitry of Fig. 1 is also capable of providing DC pulses of drive

current. This is possible in view of the fact that with series topology, the power cell carries the entire load current. As in the sinusoidal mode of operation,

capacitor 12 must be initialized prior to commencement of a run operation. In the DC mode, capacitor 12 is precharged as for a sinusoidal pulse to the V_co required to generate the desired coil current I_x equal to I_DC.

Referring to Fig. 4, at time T_Q, run

controller 30 asserts run control signal C2 to close run switch 32, allowing capacitor 12 to discharge into coil 10 with the same resonant current flow which produced the sinusoidal cycles described above; but the tuned circuit is only allowed to oscillate for a quarter cycle. At the end of the quarter cycle time segment (T₁), when capacitor voltage V_c has reached zero and coil current I_o has reached I_De' run controller 30 asserts capacitor bypass control signal C3, thereby closing bypass switch 46 to effectively short capacitor 12. Once capacitor 12 is bypassed, power cell 34 maintains the coil current at level I_De.

At time T₂, when it is desired to end the constant amplitude segment of the pulse, run controller 30 turns off bypass control signal C3, thereby opening switch 46. This initiates another sinusoidal pulse segment, in which capacitor 12 charges through a quarter cycle of resonant current flow, until capacitor voltage V_c reaches its maximum and coil current I_o reaches zero. At this point (T₃), run controller 30 turns off control signal C2, thereby opening run switch 32 and terminating the run operation.

In this manner a positive pulse is achieved having three segments: a constant amplitude segment bounded by two sinusoidal segments each consisting of a quarter cycle of a resonant sinusoid. Other waveforms can also be constructed. For example, the sinusoidal segments can be larger than a quarter cycle; an integral number of quarter cycles can be used. The inverse of the above described positive pulse can be achieved by

initiating capacitor 12 with a negative voltage.

Rectifiers and bridge circuits, such as disclosed in Rzedzian U.S. Patent No. 4,628,264, may be employed to achieve half and full wave rectification of the

sinusoidal segments.

As explained earlier, a separate precharge feedback architecture comprising capacitor charging circuitry 18 must control the initial capacitor voltage V_co to within a small allowable error. The degree of precision required relates in part to the fact that load 22 is designed to have a high quality factor Q. This results in high voltages across capacitor 12 and coil 10 using only a relatively low voltage V_p. As a result, a small error in initializing capacitor 12 can

significantly increase the power demand on power cell 34.

Charging circuit 18 is accordingly designed to precharge capacitor 12 to within a fraction of 1% of the ideal level. This is achieved by using the feedback architecture described above. To assure precision in the feedback circuitry, voltage sensing amplifier 40 is a high quality differential amplifier with common-mode rejection. This assures that the feedback signal V_cf accurately represents the capacitor voltage. The series inductances in charging circuit 38 are also carefully minimized to avoid overcharging. To minimize any

remaining parasitic effects, a dual-rate charging scheme is used in which the charging rate drops to a relatively slow rate as the capacitor voltage approaches the desired level.

Another important consideration in

implementing a series resonant feedback circuit is assuring that the frequency of drive voltage V_p matches the resonant frequency of the load. Given the high quality factor of the load, a mismatch between the drive and resonant frequencies will significantly increase the power demands on power cell 34.

To assure that the resonant frequency of load 12 remains stable over time and with temperature changes, the temperature of the capacitor is kept constant by being positioned in the incoming airstream of the power cell chassis. Further, the capacitors themselves have a low dissipation factor and low temperature coefficient of capacitance. A bank of twelve 2.0 microfarad capacitors with a 2KV peak rating are combined with sufficient trim capacitors (all of General Electric series 28F5600). Steps are also taken to assure stability in control circuitry 20. Run controller 30, difference node 26, and reference signal generator 28 are implemented with a digital computer clocked by a highly stable crystal oscillator.

Another consideration is DC error in the current sensing amplifier 24. As explained earlier, run controller 30 integrates all low frequency components of error signal V_a. Since run controller 30 cannot

distinguish DC errors in V_a introduced by sensing

amplifier 24 from real errors in current I_o, controller 30 responds to DC errors in amplifier 24 by distorting I_o and its integral. Furthermore, any offset current I_os introduced by controller 30 creates a voltage ramp on capacitor C with the slope: dv_c/dt = I_os/c

To control these errors, current sensing amplifier 24 is a high quality differential amplifier chosen to have DC stability. Further, as with capacitor 12, airflow is provided over amplifier 24 to minimize temperature variations and further reduce drift.

Fig. 2 shows the preferred embodiment. A single, bipolar power cell 110 is used to provide energy for both precharge and run operations. Digital

controller 112 controls the operation of power cell 110 during both operations. A digital control signal on line 111 prescribing the output of the power cell is delivered by controller 112, and converted by DAC 114 into a corresponding analog voltage V_B. The power cell

generates a high power output across terminals 116, 118 in precise conformity to voltage V_B.

Differential output terminals 116, 118 are connected across the series resonant load, which consists of capacitor 122, shunt resistor 124, and coil 126.

Terminals 116, 118 are also connected to capacitor charge circuit 120 to supply energy during the precharge

operation.

The multiplexing of the output of power cell 110 between the two operations of the precharging and running is controlled by digital controller 112. The controller activates capacitor charge circuit 120 by asserting control signals CS1, CS2, CS3 to close charging switches 130, 132, 134 (SCR1, SCR2, SCF3, SCR4). To select a run operation, controller 112 asserts control signal CS4 to close run switch 136 (SCR5, SCR6).

To execute a precharge operation, controller 112 supplies a 15 KHz sinusoidal input to the power cell input line 111 causing the power cell to produce a corresponding 15 KHz sinusoidal voltage V_p across

terminals 116, 118. Controller 112 next asserts control signal CSl to close charging switch 130. Upon the closing of switch 130, the 15 KHz current flows in primary coil 138 of step-up transformer 139 and provides a 15 KHz voltage at the output of each of secondary coils 140, 142.

Low inductance connections are used on the primary side of transformer 139 (e.g., for lines 116, 118) to minimize parasitic inductance.

The voltage generated by secondary coil 140 is applied to high-voltage rectifier 144 which produces a corresponding positive DC voltage across terminals 146, 148. Similarly, the voltage generated by secondary coil 142 is applied to high-voltage rectifier 150 to produce a negative DC voltage across terminals 148, 152.

Controller 112 selects a positive precharge voltage by asserting control signal CS2 to close charging switch 132. The voltage V_c across capacitor 122

accordingly rises to the voltage of rectifier 144 at a rate limited by current limiting resistor 154. To select a negative precharge voltage, controller 112 instead asserts control signal CS3, thereby causing the capacitor voltage to reach the negative level set by the output of rectifier 150.

As explained above in connection with Fig. 1, feedback architecture is used to precisely control the magnitude of the capacitor precharge voltage. In this regard, the voltage across capacitor 122 is read by voltage sensing amplifier 156 and converted to digital form by ADC 158. Controller 112 compares the sensed capacitor voltage with a predetermined digital reference voltage stored in the controller's memory. Based on the result of that comparison, controller 112 adjusts the power cell output. Using a dual rate charging scheme, controller 112 brings the capacitor voltage to the

desired level.

Once capacitor 122 is precharged, controller 112 begins a run operation by asserting control signal CS4 to close run switch 136 and thereby initiate current flow into coil 126. Voltage sensing amplifier 160 reads the voltage across shunt resistor 124 to provide a

voltage representative of the coil current. After being converted to digital form by ADC 162, the voltage is compared with a preset digital signal stored in the controller.

Controller 112 internally calculates an error signal (V_a, Fig. 1) and an appropriate correction signal (V_b, Fig. 1). The appropriate correction signal is then supplied to power cell 110 on input line 111. In this manner, controller 112 implements the operation of run controller 30, charge control circuitry 42, voltage references V_R and V_L, and difference nodes 26, 44 (Fig. 1) . Status monitor 180 provides an operator interface. To minimize the DC offset problems discussed above, both sensing amplifier 160 and ADC 162 should be chosen to minimize DC offset. Further, proper airflow over these components should be provided to minimize temperature variation.

Other embodiments are within the following claims.

Claims

1. Circuitry for driving a magnetic fieldgenerating coil in a magnetic resonance imaging system, said circuitry comprising:

a capacitive element for connection to the coil, said coil and capacitive circuit forming a resonant circuit having a resonant frequency;

a current sensing circuit connected to said resonant circuit for sensing the current flowing through said coil;

feedback circuitry for comparing the sensed current flowing through said coil to a reference current represented by a reference signal and, based on the result of that comparison, providing an output signal prescribing the power to be supplied to said resonant circuit to cause the current flowing through the coil to more closely approximate the reference current; and

a power source for delivering power to said resonant circuit in response to said output signal.

2. The circuitry of claim 1 wherein the

capacitive element and the power source are connected in series and configured to be connected in series with the coil.

3. The circuitry of claim 1 wherein said reference signal comprises a plurality of sinusoidal segments having said resonant frequency.

4. The circuitry of claim l or 3 wherein said reference signal has at least one flat segment.

5. A method of generating a time-varying magnetic field for magnetic resonance imaging, said method

comprising the steps of: (A) connecting a capacitive element, magnetic field-generating coil, and power source to form a

resonant circuit having a resonant frequency;

(B) storing energy in the capacitive element;

(C) releasing the stored energy to cause said resonant circuit to resonate;

(D) sensing the current flowing through said coil;

(E) comparing the sensed current to a reference current represented by a reference signal; and

(F) based on the result of that comparison, delivering power to said resonant circuit in such a manner as to cause the current flowing through said coil to more closely approximate the reference current.

6. The method of claim 5 wherein step A comprises connecting the capacitive element, coil, and power source in series.

7. The method of claim 6 wherein step A further comprises connecting a shunt resistor in series with the coil to form the resonant circuit, and wherein step D comprises:

connecting input terminals of a voltage sensing amplifier across the shunt resistor,

connecting an analog to digital converter to an output of the voltage sensing amplifier, and

connecting a digital processing circuit to the output of the analog to digital converter.

8. The method of claim 7 wherein steps D and E are implemented by said digital processing circuit.

9. The method of claim 8 wherein step E further comprises the steps of: (1) subtracting the sensed current from the reference signal to form an error signal, and

(2) integrating the error signal.

10. The method of claim 5 wherein step B comprises:

(1) sensing the voltage across the capacitive element,

(2) comparing the sensed capacitive voltage to a reference level,

(3) based on the result of that comparison, driving current through the capacitive element in a manner to cause the capacitive voltage to more closely approximate the reference level.

11. The method of claim 10 wherein step B(2) comprises the step of:

(a) determining whether the sensed capacitor voltage is within a specified fraction of the reference level; and wherein step B(3) comprises

(a) driving a first current through the capacitive element if the result of step B(2) (a)

indicates that the sensed capacitive is not within a specified fraction of the reference level; and

(b) driving a second current, lower than said first current, through the capacitive element if the result of step B(2) (a) indicates that the capacitor voltage is within a specified fraction of the reference signal.

12. A method of generating a magnetic field for magnetic resonance imaging, said method comprising the steps of:

(A) connecting in series a capacitive element, a magnetic field-generating coil, switch circuit, and power source to form a series resonant circuit having a

resonant frequency;

(B) opening the switch circuit to isolate said capacitive element and prevent said resonant circuit from resonating;

(C) connecting a charging circuit across said capacitive element;

(D) measuring the level of energy stored in said capacitive element;

(E) comparing the measured level of energy to a reference level;

(F) based on that comparison, delivering

additional energy from said charging circuit to said capacitive element;

(G) continuing steps D, E, and F until the level of energy stored in said capacitive element approximates the level prescribed by said reference signal.

13. The method of claim 12 wherein the level of energy in said capacitive element is measured by sensing the voltage across said element.

14. The method of claim 13 further comprising the steps of

determining whether the sensed capacitor voltage is within a specified fraction of the reference level; and driving a first current through the capacitive element if the result of that determination indicates that the sensed capacitive is not within a specified fraction of the reference level; and

driving a second current, lower than said first current, through the capacitive element if the result of that determination indicates that the capacitor voltage is within a specified fraction of the reference level.