Primary Goals
In hyperthermia, non-invasive ultrasound surgery, and other applications of therapeutic ultrasound, the primary effect which one wishes to regulate accurately is the temperature elevation in tissue. Ideally, at every point in the tissue, the temperature elevation is well known and accurately controlled. Various factors influence the actual temperature elevation achieved resulting from the sonications of a therapeutic ultrasound transducer. Some of these effects are tissue based, such as rate of blood perfusion, tissue thermal properties such as conduction, and overlying tissue geometry. These factors are generally beyond the control of the treatment administrator. However, other factors such as acoustic field shape and total acoustic power delivery have a fundamental effect on induced temperature elevations and can be directly controlled by the treatment administrator.
Simple desired acoustic field shapes can sometimes be achieved by using single element transducers, either alone or with the aid of field shaping mechanisms such as transmissive lenses or reflectors. However, to create complex field shapes that can be modified online, phased arrays are necessary. It is well known that, in general, the higher the number of elements in the phased array, the greater the inherent flexibility in acoustic field shaping the phased array will posses. While greater flexibility is no doubt an advantage when it comes to accurate and controlled acoustic power delivery and thermal elevation, this flexibility can only be truly actualized by accurately controlling the phase and power delivered to each element in the array. In either case -- single-element or multi-element transducers -- the total acoustic power output has a direct linear relationship with the induced temperature elevation, and thus acoustic power must be accurately controlled.
Extensive simulation studies have shown that for larger arrays, i.e. arrays with higher numbers of elements, phase uncertainties of plus or minus only a few degrees can have a noticeable effect on the acoustic fields. In addition, variations in applied RF power to the elements from the desired powers can also result in noticeable field effects. In the case of an array geometry such as the concentric-ring or the sector-vortex design pioneered by Cain and Umemura, the large differences in individual element sizes require that the driving system have a sufficient range in power control from channel to channel and that the system be able to accurately control the power delivery to elements with significantly different intrinsic impedances.
Thus, a system is needed that:
 | can accurately power the huge multitude of array designs, element sizes, and piezoelectric materials |
| can accurately control the phase of the RF signals arriving at each element surface |
| is expandable to include the number of channels necessary to drive different sized arrays |
| is safe and reliable in order to allow for future implementation into clinical systems |
A common way to accurately measure the forward and reflected power transmitted to a load, a dual-directional coupler is commonly used. However, the dual-directional coupler relies on known input and output impedances in order to accurately measure the power. Thus, in order to drive elements with significantly different impedances and be able to accurately control the power deposition, it is necessary to impedance match the elements to the standard output impedance of the amplifier system. Once the elements are matched at least close to the output impedance of the amplifier system, it is then possible to drive the load and to accurately measure the forward and reflected power transmitted to each element. The accuracy of the dual-directional coupler based measurements decreases as the matching drifts from the desired impedance.
For more information regarding accurate power delivery, please refer to the Power Control via Forward Power Feedback section of this document.
As was mentioned above, matching circuitry is necessary in order to drive elements with various intrinsic impedances and accurately control and measure the RF power delivery. However, the inclusion of matching circuitry results in phase shifts across each matching circuit that are dependent on the intrinsic impedance of the element. Thus, the relative phases that are found at the outputs of the amplifier channels are not the same phases that are found at the surface of the piezoelectric elements. These phase errors resulting from the matching circuitry will be apparent in the acoustic fields produced by the array elements.
Thus, phase control must be implemented after the matching circuitry, not before. This has been accomplished by implementing phase feedback circuitry. This feedback measures the relative phases at the output of the matching circuits and applies a phase correction to the driving signals in order to achieve the desired/programmed phase shifts at the surfaces of the elements.
For more information regarding phase control, please refer to the Phase Control via Phase Feedback document located elsewhere at this site.
The current trend in therapeutic ultrasound (and a trend that is expected to continue for some time) is the use of phased-arrays with an increasing number of individual elements. A common problem when driving many-element arrays is that excessive demands are placed on the controlling circuitry which must set the power and phase, monitor the system for fault conditions, and measure the power delivery. Placing these excessive demands on the controller results in low system safety, poor reliability, and slow and cumbersome interfacing.
One solution to this problem is to use a system of distributed control. The controlling circuitry for the UDS systems consists not only of an embedded controller, but also microprocessors distributed throughout the system. Specifically, each amplifier card is controlled by a microprocessor which in turn communicates with the embedded controller. Thus, the embedded controller is not responsible for directly monitoring for fault conditions, but can rather delegate this processing to the individual microprocessors. With one microprocessor monitoring and controlling only four amplifier channels, the control loop is very tight and the response rate is extremely high -- much higher than could be achieved with a single controller. This method of distributed control, while increasing safety and reliability, also provides the means for expanding the system to the number of channels necessary while only minimally increasing the burden on the embedded controller.
For more information regarding distributed control, please refer to the System Control/Software Design document located elsewhere at this site.
Amplifier Classes
The high voltages necessary to drive ultrasound transducer elements are typically produced by some form of RF amplifier. All amplifier designs rely on either one or two (depending on the desired output power) high power transistors in the output stage. The primary difference between the designs is the operating point of the amplifier. This 'operating point' determines the 'class' of the amplifier. Most amplifiers can be classified as A, B, A/B, C, D, and E. The classifications extend well beyond E, but all subsequent classes are basically variants of the Class E design.
Class A
An amplifier operating in Class A mode is biased such that the output stage transistors are in the active region at all times. As a result, power is dissipated even in the absence of an input signal. In fact, when there is no input signal, ½ of the maximum amount of drain current is dissipated by the output transistors. It is only when there is some sort of input signal that some of the power is dissipated in the load rather than in the amplifier itself. The advantage of this mode of operation is that since the output stage transistors are always on the amplifier is extremely quick to respond to changes in the input signal. The disadvantage is that the output transistors must be large and have sufficient heat sinking to remove the output power of the amplifier when no signal is present.
Class B
Class B amplifiers are biased right at the turn on point of the output transistors. This means that current flows in the output stage only when the input signal is greater than the turn-on voltage of the output transistors. This results in a little over 25% in improved efficiency, but at the cost of higher distortion, especially around the turn on point of the output transistors. Class A/B amplifier are biased somewhere between the turn-on point and the half linear point (i.e. Between Class A and B operation). Amplifier distortion can be reduced at the cost of reduced efficiency. When zero input signal is present a Class B amplifier should have no current flow in the output stage, but current should flow if the input signal becomes slightly positive. Class A/B amplifiers would have some current flow with no input signal present since they are biased above the threshold voltage of the output stage.
Class C
Class C amplifiers are biased at zero volts, which is below the turn-on voltage of the output transistors. The input voltage must exceed the turn-on threshold of the output transistors before any current flows in the output stage. The result is an amplifier with very high efficiency and very high distortion since the output signal does not begin to rise until the input signal has exceed the turn-on threshold.
Switching Amplifiers
Class D and E amplifiers work on a somewhat different principle than the previously described modes of operation. Class A-C amplifiers are typically described as "linear amplifiers" since their output voltage is proportional to the input signal. Class D and beyond amplifiers are described as operating in "switch" mode. The input signal is simply used to turn on the output stage transistor to it's full output power. Once the input signal falls below the turn-off threshold, the amplifier output immediately falls to zero. The result is that the output waveform is typically a square-wave with an amplitude equal to the voltage placed across the output stage. In Class E amplifiers, the output stage is designed into a special filter circuit that converts the square-wave into a sine-wave with a great deal of efficiency. A perfectly designed Class E amplifier will have no voltage across the output stage transistors while current is flowing in that transistor. Since there is no voltage drop, no power is dissipated in the output stage. Class E amplifier can operate at efficiencies approaching 100%.
Class E Amplifiers
As mentioned above, Class E amplifiers have very high efficiency and are capable of producing sine-wave output signals. Additionally, since no power is dissipated in the output stage during switching, the output stage requires minimal heat sinking. This is a distinct advantage since most linear amplifier require very large heat sinks. A disadvantage of Class E amplifiers is that the output signal amplitude is independent of the input signal amplitude, so how do you control the output power? As it turns out, this is an advantage rather than a limitation for phased array amplifiers. First, the amplitude of the input signal does not have to be carefully controlled, so simple digital signals can be used. Second, the output power can be simply controlled by controlling the voltage across the output stage of the amplifier. The primary disadvantage of Class E amplifiers is that, because of the tuned filter in the output stage, the amplifiers have a much narrower bandwidth than linear amplifiers. Linear amplifiers can have bandwidths of several decades while the bandwidth of a Class E amplifier is limited by the Q of the tuned filter. The use of multiple filter circuits that are switched in and out of the circuit can increase the bandwidth of the amplifier, but at the expense of size.
Driving Ultrasound Arrays
What is the best amplifier for driving ultrasound phased arrays? For any application which requires a large number of amplifiers in a small area, Class E amplifiers are the design of choice. The need for minimal or even no heatsinking allow Class E amplifiers to be constructed in significantly less space than linear amplifiers with equivalent output power. Additionally, since the input signal does not need to be amplitude controlled, the input electronics can be simplified by using standard logic components to control the phase shifting rather than precision operational amplifiers and analog phase shifters.
Power Feedback
If the efficiency of a phased array is known, then the amount of deposited power can be determined if the transmitted RF power is known. For linear amplifiers, the output voltage is determined by the magnitude of the input signal. Unfortunately, it is also at least somewhat dependent on the load. This means that the same amplitude input signal can result in different amounts of RF power will be delivered to the load depending on the impedance of the load. This makes it difficult to deliver a precise amount of RF power to a transducer element. If RF power measurement devices are available, then the power can be measured, and the amount of RF power delivered to the load can be calculated after the fact, but it is very difficult to control the delivered power during short sonications.
Controlling the power of switching amplifiers is typically less predictable than with linear amplifiers. The voltage across the output transistors on the amplifier determines the output amplitude. Previous switching amplifiers designed for driving phased arrays utilized a single variable power supply to vary the output amplitude of all of the amplifiers simultaneously [Buchanan and Hynynen, 1994]. Individual power control was achieved by varying the duty-cycle of the driving signal. This technique turned out to be impractical since each amplifier channel produced different output amplitudes and varying the duty-cycle caused significant phase shifts.
Since the output RF power from each channel is one important variable that needs to be controlled for precisely driving ultrasound phased arrays, it seems logical that the user only need to provide the desired output powers to the amplifier. Rather than trying to control the amplitude of the output signal, which would result in a variation of delivered power depending on the load, it would be far more useful to regulate the output amplitude based on RF power. This can be achieved by measuring the RF power delivered to the load using a dual-directional coupler. The forward power could then be used as a feedback signal to control the output amplitude.
The amplifiers offered by Advanced Surgical Systems, Inc. were designed with the ability to accurately control power. Highly efficient switching power supplies provide an adjustable voltage across the output stage of the amplifier. Dual-directional couplers measure the forward RF power and adjust the switching regulator outputs so that the desired RF power is delivered to the load. Since directional couplers are used, it would be possible to regulate the transmitted (forward-reflected) RF power, but this would result in an unsafe situation when the amplifier outputs are disconnected from the load. Thus, the amplifier outputs are controlled by forward power. An on-card microcontroller continuously monitors the performance of the amplifier and shuts down the amplifier when large errors in the forward power are detected. The microcontroller also shuts down the amplifier if significant amounts of reflected power are detected, indicating a loss of load. While the amplifiers can safely operate with no load (open circuit fault) or a grounded load (short circuit fault), the microcontrollers will disable the amplifier output channel for safety considerations.
Phase Feedback
For optimal phased array performance, the desired relative phases of the electrical driving signals must be controlled at the transducer element surface and not just at the output of the amplifiers. Errors in the phase of the driving signal at the transducer will result in phase errors in the acoustic field, resulting in sub-optimal acoustic fields and reduced array performance. Since element phase is not an absolute quantity but is measured relative to the phases of the driving signals of the other elements, the relative phases that are correct at the output of the amplifiers might not be correct at the transducer. Differences between electrical signal path lengths or differences in the impedance of the elements will cause potentially significant errors in the acoustic field.
The primary reason for loss of relative phase coherency is differences in the impedance of the transducer elements. This is, in part, a result of the need to accurately control the power delivered to the array elements. In order to insure maximum power transfer to the elements and to insure the ability to measure this power accurately, the transducer elements must be impedance matched to the output impedance of the amplifier system. (A standard amplifier output impedance is 50 Ohms, which is the output impedance of the Advanced Surgical Systems' UDS series amplifiers.) However, while impedance matching is necessary for accurate power delivery, a phase shift is introduced by each matching circuit that is dependent on the intrinsic impedance of the transducer element. This is a significant problem with arrays with differing element sizes, such as the concentric-ring phased array, where the differences in impedance can easily cause phase errors in excess of 45°. Even arrays with identically sized elements suffer from changes in impedances from element to element due to differences in the piezoceramics. Small elements with relatively high impedances are extremely susceptible to phase errors caused by element-to-element impedance variations.
In order to eliminate phase errors caused by variations in impedances and electrical path length, it is possible to measure the acoustic field produced by each element with a hydrophone and determine the difference between the measured and desired phase. This information can then be used to correct the desired phase files. Unfortunately, the correction factors are not array independent and must be measured for each array. In order to guarantee phase coherency over time, the arrays would need to be recalibrated over time to compensate for changes in impedance that result as the piezoceramic ages.
In order to eliminate the need to conduct periodic phase calibrations, Advanced Surgical Systems, Inc., has introduced a novel phase feedback technique that totally eliminates phase errors. For each amplifier channel, the phase of the RF driving signal is sampled after the electrical impedance matching for the respective transducer element and is synchronized with the phase at the output of the phase shifter. (See block diagram below.) This feedback mechanism results in a real-time phase correction that eliminates phase shifts caused by the varying impedances of the array elements.
This phase feedback technique can reduce the phase error between any two array elements to less than 5°.
Block Diagram of Phase Feedback
UDS Series Driving System
System Control and Software Design
Overview
For therapeutic ultrasound, it is necessary that the driving system be safe, reliable, and expandable, in addition to providing the basic control necessary to accurately drive a phased array. Thus, there are several tasks which must be performed by the controlling hardware and software. These tasks include:
| Set the power and phase |
| Measure the forward and reflected power |
| Monitor the operation of the amplifier circuitry |
| Shut down the system in the event of an amplifier fault condition |
For a system containing only a few channels, the control architecture is not as major an issue -- a single embedded controller can easily control a few channels. However, for a system that may possess over a thousand channels, the control system must be designed in such a way that no single part of the system is excessively burdened. This means that the number of automatic tasks performed by a single controller (such as fault detection) be reduced to an amount that can be comfortably performed by the controller and still allow sufficient resources for communication and asynchronous demands placed on the controller which originate from the user's console (such as measuring power, setting power, etc.) One solution to this problem is to use the method of distributed control throughout the driving system.
In the UDS amplifier systems, there are four independent amplifier channels on each amplifier card. These four channels are controlled and monitored by their own microcontroller. An embedded controller interfaces with the host computer/user interface, and informs the individual microcontrollers of the power, phase, etc., that the user has selected. In this architecture, each microcontroller need only be concerned with the operations of the four specific channels on its card. A block diagram of the distributed control architecture of the UDS series is shown in the following figure:
Block Diagram of Distributed Control Architecture
This distributed control architecture greatly reduces the burden on any single controller. For example, the user may wish to set maximum or minimum levels of either forward or reflected power on some or all of the operating channels. This information is sent to the embedded controller via a simple, English language command structure. The controller then disseminates the information to the appropriate microcontrollers via a custom designed high speed serial bus with an application specific protocol. During operation, each microcontroller will closely monitor the operation of four channels, and the embedded controller will continually poll each card for flags indicating fault conditions. Before sonication, the individual microcontrollers can be programmed by the embedded controller to disable any channels outside of the preset limits or to merely notify the embedded controller. In either case, if a fault is detected by the microcontroller, a flag indicating the type of fault detected is set. The embedded controller, upon polling and receiving this flag, can take appropriate action, including notifying the operator of the specific fault condition.
A variety of fault conditions may be detected:
|
Over
or under forward or reflected power limits |
| Over temperature (on the UDS 6460 systems) |
| Loss of phase feedback signal |
| Several other internal operation safeguards have also been implemented. |
For safety, the internal serial bus was specifically designed use in the UDS system. Besides the necessary serial lines, several lines have been implemented in a parallel manner in order to avoid any problems that may be associated with a blocked serial bus. To help avoid this condition, the bus is configured as a single master bus (the master being the embedded controller). Thus, polling is used to detect fault conditions detected by the individual microcontrollers. In addition, the embedded controller, as well as each individual microcontroller, includes a watchdog timer to prevent a loss of system control. Individual microcontroller resets will be detected by the embedded controller, and an embedded controller reset will be detected by the user interface console. Microcontroller resets automatically disable the affected channels until they are manually re-enabled via software. An embedded controller reset disables the entire system. Thus, the probability of unwanted power delivery is almost completely eliminated.
Besides monitoring the individual channels for safety considerations, distributed control also offers performance increases that could not be easily achieved with a single controller. For example, the UDS 6460 systems allow the pre-programming of power and phase steps via an internal microcontroller stack stored in non-volatile reprogrammable memory. During operation where it is desired to rapidly switch between various phase/power settings, a single parallel line can be toggled (either via an external input or via software commands to the embedded controller) which will inform all microcontrollers simultaneously to step to the next phase and power settings stored on the stack. With the current control architecture of the UDS systems, phase/power setting switching speeds in excess of 40 Hz can be achieved.
On the UDS 3225-HF systems, power and frequency steps may be programmed into the stack. This method is used in order to implement amplitude and/or frequency modulation.
