Tracewell Systems, a custom power supply manufacturer and engineering services company, builds power supplies and other electronic equipment for a variety of customers. One contract came from a major manufacturer of magnetic resonance imaging (MRI) systems to build axis amplifiers for its imagers and a system to test them.
The load is an MRI gradient tube, and a cooler in the left cabinet bay supplies water for the power supply and the gradient tube. Instrumentation is located in the middle cabinet bay. The system’s modular design makes field service easier.
While building to the customer’s design was a straightforward job, testing the completed amplifiers was not. Because of the power levels involved, none of our automated power supply test machines could do the job.
The answer was a custom automated test system using PC-controlled instrumentation in a hot mockup configuration. Economics called for a fully automated test with a low capital cost and quick development time. An opportunity existed to build the full gradient amplifier cabinet, so expansion of the test system had to be considered. Operator safety also was a prime concern.
How an MRI Works
An MRI uses two magnetic fields: a high-intensity static field produced with superconducting coils and a variable field produced by three sets of orthogonal windings mounted inside a gradient tube into which the patient is placed for scanning. The static field is on the order of 0.5 to 2 Tesla (5,000 to 20,000 gauss); the variable field typically is 180 to 270 gauss (18 to 27 mT).
When subjected to a magnetic field, the nuclei of hydrogen atoms become frequency-sensitive. They resonate with an applied pulse of radio frequency (RF) energy and reradiate that same frequency when the RF pulse ends.
The amplitude of the signal emitted by an individual atom depends upon the physical and chemical nature of its immediate environment (the tissue of which it is a part). Its resonant frequency relies on the local value of the magnetic field, which varies with position in a known way. An antenna, a receiver, and a computer capture the emitted radio waves, correlate them with the instantaneous value of the magnetic field at each location, and develop a series of cross-sectional images of the body structure being scanned.
Testing with the main magnet in place would be very dangerous because, in full operation, an MRI can cause objects to become projectiles as they are sucked into the gradient tube. There is a recorded case of a fully loaded pallet truck being pulled violently into the tube of an MRI. Fortunately, our system uses just the gradient tube and no magnets, so the field strength is low enough to be safe.
Test-System Requirements
The test system had to be fully automated to ensure repeatable test conditions and keep costs down. The operator’s only task would be to load and unload units, with no need for constant monitoring. The system had to automatically datalog and print out any failure reports. For safety reasons, there should be no need to get close to the machine during testing.
The system also had to be flexible to easily handle changes in test specifications and product configurations and expandable for full cabinet testing in the future.
System Description
Each set of coils is powered by a separate microprocessor-controlled power amplifier that runs on ±350 V, produces currents of up to 440 A peak, and is mounted on rack slides in its own drawer in the cabinet. Each axis amplifier acts as a voltage-to-current converter.
Internal circuitry in the amplifier compares a low-level control signal from the signal generator with a current feedback signal from the load. It produces pulse-width modulated inputs to the amplifier’s power stage, which consists of blocks of high-power MOSFETs arranged in an H-bridge configuration. Switching rapidly on and off, the MOSFETs, along with an IGBT, convert the ±350-VDC input from the power supply into a high-level current that flows into the low impedance (about 0.15 W and 0.5 mH) of the axis coils in the gradient tube.
Initial plans called for a test load using resistors and inductors to simulate the coils in the gradient tube. But as the requirements for the system developed, the power level became too high for that.
In the end, the customer supplied us with a gradient tube that we used for the load. The gradient tube is rated for 200 Arms continuous if one coil is driven and 175 Arms per coil if three coils are driven, for a maximum total dissipation of 13.8 kW.
Keeping It Cool
Both the power supply and the gradient tube are water-cooled. A heat exchanger/chiller mounted in the left cabinet bay circulates distilled water at 18°C through those units. Figure 2 is the block diagram of the complete test system including instrumentation, safety curtain, and cooling provisions.
For a source of water to take the heat out of the heat exchanger we selected the plant fire pond. The water is brought in from the pond through a filter to keep out aquatic life and debris, passed through the heat exchanger, and pumped back to the pond. Most of the time, the pond water does a good job. But during hot weather when the pond becomes too warm, we switch to city water to cool the heat exchanger.
Paying Attention to Safety
With the voltages involved, operator safety is vital. For that reason, it is important to keep all personnel away while the system is running. To ensure that, we enclosed the system with a light curtain. Breaking a light beam immediately removes all input power from the system and begins a controlled shutdown.
While the light curtain solved one problem, it created another: how to turn the system on and off. In a complete MRI unit, it’s possible to control everything via an RS-232 link. But in this test system, we use two manually operated rocker switches on the front panel, one to turn on voltage to the system and the other to start the coil current.
So, how could the operator throw the switches without touching them? The answer was remote control. The company that builds our PCB test fixtures designed a remote-switch actuator using a pair of small pneumatic cylinders for each switch, one for on and one for off, controlled by electrically operated valves.
Test-System Configuration
The test system’s structure is similar to a standard power supply test setup. There is a power source, a switch matrix, monitoring instrumentation, a DUT interface, and a test load. The entire system took between nine and 12 months to assemble.
A Compaq Deskpro 650-MHz PC with a National Instruments’ IEEE board serves as the system controller. To keep the PC outside the light curtain and the IEEE instruments inside, a 50-ft IEEE 488 cable was strung up and over the light curtain. This cable is longer than recommended, but we have had no problems with it.
System instrumentation consists of an oscilloscope, a switching system, a Keithley 7002 ATE Switch/Control System to route the test signals to the oscilloscope and handle other signal-routing and control chores, a Keithley 2000 DMM, and an Agilent 33120 Arbitrary Waveform Generator that produces the input signal to the amplifiers. The Agilent 33120 was chosen for its capability to generate the complex signals required in this application as well as its low cost.
For the oscilloscope, we looked at several with similar specifications. We chose the same type the customer already was using—a Tektronix 3012B dual-channel unit.
The Keithley 7002 handles all signal routing. It sends the feedback and monitor signals from the amplifiers to the oscilloscope, inputs and outputs from various system controls and sensors to their various destinations, and the trigger signals to the pneumatic valves. The multiplexer, oscilloscope, DMM, and function generator connect to the PC via the IEEE 488 bus.
Control Software
We selected National Instruments’ LabWindows/CVI® as the programming language based on its virtual-instrument feature that aids programming and its graphical user interface development to ease preparation of a control and monitor display. Since it is based on C, it encourages a modular design in which programming can be added with a reduced chance of affecting existing code.
Test Procedure
The amplifier has three outputs monitored by the oscilloscope: command, the input signal; output, the current feedback signal sensed by the 600-A current transducer; and error, the difference between the command and output signals. After a 5-minute wait after system turn-on to allow the water cooler to stabilize, a three-cycle, low-amplitude trapezoidal signal is sent to each amplifier to verify that it is alive and connected properly to the test system.
The command signal is checked to verify that it matches the output of the signal generator, and the output and error signal amplitude and phase angle are checked. During the full power tests, only the output and error signal amplitudes are monitored.
The signals from the amplifiers, a maximum of three units under test at one time, are switched to the scope through a Model 7017 Coax Switch Card in a Keithley 7002 Switch/Control System. In the event of a fault, the scope signals are saved to the scope waveform memory.
The four test sequences are as follows:
-
225 A peak for one hour using a 250-Hz, 160-ms sine-wave burst and a pulse repetition rate of 0.5 Hz.
-
440 A peak for one hour using the same burst as Test 1.
-
370 A peak for one hour using a 250-Hz, 45-ms sine-wave burst and a pulse repetition rate of 7 Hz.
-
150 Arms for 0.5 hour using a continuous 100-Hz sine wave.
During the tests, the signals are monitored and recorded every 15 seconds and displayed on the PC monitor. The primary water temperature and flow rate, the gradient tube thermal switches, and the cabinet power supply also are monitored. The 480 VAC input, the light curtain, and the cooler status are monitored every second using the TTL sense feature of the Keithley 7037 Switch Card in the 7002. A sense relay on the 480-VAC line trips on an undervoltage or phasing error.
The test time is three hours for a single drawer; if the system were fully loaded, the test would take about eight hours. Some of the basic tests can be run simultaneously for all axes. The high current tests must be run one axis at a time because of limitations in the amount of power available.
Conclusions
We learned several lessons from this project. First, maintain good communications with the customer to prevent misunderstandings and keep up with design changes. Secondly, learn as much as possible about the final application of whatever equipment is being built so testing can be as realistic as possible. And last, it’s vital to plan for change and not just let it happen.
About the Author
Joseph Hupf is a senior test engineer at Tracewell Systems. He has more than 30 years of automated test experience with transformers, circuit board assemblies, and power supplies. He received a B.S.E.E. from the State University of New York at Buffalo. Tracewell Systems, 9962 Rte. 446, Cuba, NY 14727, 585-968-2400, e-mail: [email protected]
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2003 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
November 2003