Research & development
Highly integrated frequency
converters for pump motors
The ever increasing demand for improving the cost to performance ratio motivated Grundfos in the mid eighties to start on integrating electronics into pumps. In 1985 it was decided that the company should be able to develop and manufacture frequency converters for integration in motors and pumps. A mature decision for a mechanical company only with experience in developing and manufacturing pumps and induction motors. A development report from Pierre Vadstrup.
Huge investments were released for this project and the result ten years after is a factory with 400 employees of which 100 have their daily duty in research and development. The factory contains the most up today equipment for manufacturing die-bonded thick film hybrids and direct copper bonded modules. Today more than 20000 highly integrated frequency converters are manufactured each month.
In the mid eighties technology for developing highly integrated power electronic products was present, the MOSFET transistor was available, Smart-Power technology was in embryo and ASIC technology was commercially available. Up to that time only frequency converters realized by discrete components had been developed and manufactured, so little help was available in the literature concerning manufacturing methods and processes. The result of the development effort is highly integrated hybrid modules known as the X3000 series modules. Today modules are commercial available from Toshiba, Mitsubishi and others [1], [2] and [3], but these modules do not contain the same complexity as the X3000 concept.
System integration
In this chapter the level of integration in the X3000 module will be described and the reasons for initiating the project will be discussed. A frequency converter for induction motor drives consists of the following electrical parts: see Fig. 1.
The items in Fig. 1 can be separated in two parts namely the energy storage components and the remaining non-energy storage components. The non-energy storage components are integrated in the X3000 module (see Fig. 2). Grundfos’s compact module is able to handle up to 10 kW output power with up to 100% overload capability. In Grundfos’ applications this module is integrated into pumps, and it is common practice to use the water in the pump as cooling media for the power module even if it is hot water with temperatures higher than 100 °C. To obtain the high level of integration ASIC (Application Specific Integrated Circuits) have extensively been used in the realization of both the driver and control circuitry. In applications supplied from single phase mains, drivers implemented in smartpower technology are used. In applications supplied from three phase mains a pulse transformer is used as communication link to the highside driver, which in this case are implemented in a BICMOS process.
The driving circuits are in both cases implemented as one common lowside and three separated highside drivers, where the highside drivers are supplied by boot-strapping techniques. Initially the goal in the development was to implement the hole drive system in one chip, but problems in getting sufficient voltage isolation on the chip prevented the fulfilment of this goal. Today drivers are available on the market from for example International Rectifier in 600V and 1200V versions.
The modules are either implemented by use of MOSFET’s for low power applications which means power less than 10kW or by use of IGBT’s for high power applications.
When using MOSFET’s, the intrinsic diode in the transistor structure is used as a free-wheeling diode, this was rather new in the middle eighties and this intrinsic diode was electrically poor, but during the late eighties this diode was improved in order to cope with inductive loads.
The control chip consists of a modulator part, a measurement part and a microprocessor part. The modulator part generates three-phase sinusoidal modulated PWM signals which are used to control the voltage amplifying transistor/diode bridge.
Control of ramping, V/F ratio control, motor protection, alarm handling and so on are implemented in the code and speed optimized microprocessor part. Normally the control part of a motor control system is implemented in a standard microprocessor or if servo performance is needed a signalprocessor. These processors often contain built-in peripherals not needed, but often additional hardware is needed in order to fulfill the applications needs. Speed of the processor is often a limiting issue, and quite often a faster processor than initially planned is required, because of lag of suitable hardware peripherals. Protection systems relying on software and robustness can be a problem in these microprocessor based systems These inexpediencies result in additional direct costs but also indirect costs to mechanics and power supply.
In the next generation of the X3000 modules this inexpediency is faced by introduction of the CPC (C-Processor Core) concept combined with dedicated hardware structures. The philosophy in the method is to let the hardware architecture of the processor reflect the C-program it is going to execute via a retargetable C compiler in an optimal sense [4]. The architecture can be optimized due to speed, code efficiency and power consumption. Time consuming tasks like the Park-Blondel transformations can be executed in dedicated hardware structures. The developed processor core has been benchmarked against an Intel 8052, the result is a 10 times speed improvement, 50% reduction of needed code and all this at less than half dissipated power. Dedicated hardware structures improve these figures even more.
Finally the measurement circuit is used for measuring currents, voltages and temperature in the converter for protection use and for control of the application functionality. These three parts are integrated into a single chip called a MACE (Mature Application Control Engine). Information passed easily between the modulation part and the microprocessor, an example of this is a principle for reconstruction of the three phase motor currents by use of the information about how the PWM signals are generated related to the currents measured in the legs in the lowside half bridges.
Many more degrees of freedom are available in the modulator than in many commercial available modulators. This implies the possibility to execute experiments during startup of the motor and pump in order to detect the initial conditions. Examples can be the detection of a running motor, detection of non-decayed electromagnetic fields in the motor, measuring of stator resistance etc..
The inrush function is implemented in thick film technology, which includes control circuitry, inrush resistor and inrush transistor, this circuitry gives a smooth charging of the bus capacitor.
The module consists of a low level portion in contact with a copper base, this level is split up into a power part and a control part. On the power part the rectifier, inrush transistor and inverter power semi-conductor dies are all soldered to the DCB (Direct Copper Bonding) and the control part consists of a ceramic substrate on which die-bonded ASIC’s are attached together with SMD (Surface Mounted Devices) components via a multilayer thick film layer. The ceramic substrate is glued to the copper base.
On a second level in the module structure application specific circuitry can be implemented on thick film. After production the module is sealed with gel to protect it against humidity and dirt.
The module is a unit that together with the energy storage components form a full frequency converter. This unit can autonomously control a induction motor, or a PM synchronous motor or even a BLDC motor. If galvanic isolation is needed for the module, it can be controlled via a serial interface to a PC or to another microprocessor by a user-friendly protocol.
In Grundfos products the module operates normally autonomously but the serial interface is used for diagnostic purposes to monitor the historical values of a unit when it is serviced or in case of malfunctions. The serial interface can even be used for infrared communication with the module.
Benefits of integration
From a mechanical point of view the benefit of integrating the non-energy storage components in a module seems not to be so obvious, because the energy storage components are the space consuming parts. However there are other reasons that does the integration favorable, the following reasons are significant:
Unit cost is the market generating parameter, if the cost is not minimized the market share is reduced, so unit cost reduction is a must. In the mid eighties when the development of integrated modules started at Grundfos, existing frequency converters were scrutinized and analyzed in order to evaluate what was the cost factor. It was obvious that a lot of manual work was needed to manufacture these converters as they consisted of a lot of components. Furthermore these frequency converters contained a lot of functionality that from a pump control point of view was redundant.
One of the goals was to build a frequency converter at one tenth of the price of a standard converter and this goal has been reached.
The problem with standard electronic components available on the market is that they normally do not have the exact functionality you want, and they often have redundant functionality. Normally a bunch of external components are needed in order to program the correct functionality.
These facts justified the decision to use ASIC’s for realizing the circuits in the drive system and control system. In the X3000 module all integrated circuits are implemented in ASIC’s and during the development of the chips special attention has been paid to the minimization of the number of connections, because for every extra wire bond on a chip the reliability of the product is reduced. In order to reduce the amount of bondings, chip-to-chip bonds are used.
The driver and controller in the frequency converter consists of only 7 chips and a few external components, of which the resistors are implemented as laser trimable resistors. The chip set consists of 1 Low Side Driver, 3 High Side Drivers, 1 MACE and 1 Inrush Driver.
The small module area, the few components, thick film resistors and all active components in dieform are the ingredients that makes large scale production possible. All the processes in manufacturing the modules are highly automated with lasertrimmers, automated die bonding and automated Al and Au wire bonding.
Availability is a consequence of using ASIC’s, because when you control the process down to the mask level, as you do in full custom designs, it is no problem to develop the design to a secondary silicon foundry vendor, and it is up to you to control the delivery.
From a technical point of view there are some savings in integrating the frequency converter in a module. Distances inside the module is very short which eliminates the need for using snubbers and high di/dt ratings are possible. Further benefits obtained by integrating are that signal conditioning is not a problem due to the very small distances on the chips and between chips. Concerning EMC the module has a well defined geometry which makes it easier to make a more EMC correct design, and it is our experiences that much less voluminous components are needed to fulfill the demands for approvals.
By using soldering techniques for mounting the power components on the DCB, certain thermal transitions are avoided and high power cycle capability is obtained. The thermal resistance from the power component junction to the copper base is decreased. In some applications the module is cooled by water temperatures as high as 110 degrees C.
Now all the goodies about integration has been mentioned, but during development the high degree of integration is a problem, because it is not easy to make hardware debugging. When debugging is needed special fixtures have to be developed in order to enter the compact structure. Debugging on silicon is a tedious job so it is necessary to add additional circuitry on the ASIC’s in order to bring the chip in a state where internal parts can be accessed or observed.
Simulation tools can be helpful in generating the necessary test circuitry.
During development many solutions rely on simulations and not on practical experiments.
Design tools
Prototyping can be difficult in a project based on ASIC’s because a breadboard solution does not have the same properties as a fully integrated solution, but there has to be some confidence in the chosen circuit realization. It is not unusual that it takes several months from shipment of a full custom design layout to the day where engineering samples of the chip is available, and besides this fact the expenses are high.
To confirm solutions and to build up confidence, simulations are extensively used and three types of simulators are use for simulating the ASIC implementation:
n analog simulator
n digital simulator
n mixed mode simulator
It is obvious what the intention is for the two first simulators, but the purpose of the last one is to simulate connections between the digital and analog implementation. The question one could rise is if there is a need for the two first ones when the last one is available, but the answer is yes, because when you use this mixed mode simulator the digital part will be simulated electrically and not logically, which will increase the calculation time considerable. CADENCE is the simulation tool used for these three types of simulation.
Normally prototyping is used where it is possible and FPGA’s (Field Programmable Gate Array) are extensively used for implementing digital parts.
Power handling is another issue that up front has to be accounted for in the development of modules, because the responsibility for correct cooling is ours. Using naked dies means that you are in charge of choosing the thermal transitions from the chip down to the copper base. Grundfos has developed a simulation tool where the dynamic transfer function from power dissipation in the junction of a power component to the junction temperature has been modeled By this tool the peak temperature during severe load conditions can be simulated, an example of this can be a simulation of the situation when the motor suddenly stops due to blocking.
This simulation program shown in fig. 3 is based on MATLAB/SIMULINK which is a strong combination for making both discrete and continuos time simulations see [5]. The tool is very block oriented so the user does not have to worry about simulation technical issues concerning synchronization of the individual blocks. Very little programming is needed in the Simulink environment so the effort can be concentrated on the technical issues.
For modeling the thermal transfer functions in the power components, which as mentioned previously, is a part of the simulation program, the finite element program ANSYS is used.
The programs are used in the following sequence: first a thermal transient is calculated in ANSYS, then this response is fitted to a transfer function description in MATLAB and finally this transfer function is transferred to SIMULINK ready for simulation. (see fig 3.)
Functionality clarification is done in a signalprocessor based system called D-Space. This tool has some profitable links to the Matlab/Simulink environment. Programs developed in the simulation environment can be used directly in the D-Space environment, which gives full consistency. The D-Space system can even be part of the simulation system.
Severe situations, that can be difficult to establish in a laboratory, can by the simulator be analyzed and evaluated in order to implement the correct protection strategies in the modules.
The simulation tool has been of great help in explaining different observed phenomena and for evaluating different control strategies
Conclusions
The technologies for developing and manufacturing highly integrated frequency converters are available but there are still a conspicuous mismatch between the geometrical sizes of the components containing energy and the ones that do not. There is a need for research in the possibilities for making more energy dense DC-link capacitors.
Smart Power technology has not yet had its breakthrough especially not in integration of the driving system together with the power components in the high voltage range.
Finally there is a need for research in power components able to operate at higher temperatures than these available today, because the power concentration becomes more and more dense as integration improves implying that thermal problems becomes more and more severe.
Modules are becoming available on the market in the shape of six-pack’s, but using these modules requires still some external components and can cost wise not compete with the X3000 module, which recently have been launched for OEM customers.
References
[1] E.R. Motto, et. al. , "A New Generation of Intelligent Power Devices for Motor Drive Applications", IEEE IAS Conference October 1993.
[2] G. Majumdar, et. al. "A New Generation High Performance Intelligent Module" PCIM Europe May 1992
[3] E.R. Motto, et. al. , " A New Intelligent Power Module With Microprocessor Compatible Analog Current Feedback, Control Input and Status Output Signals", IEEE IAS Conference October 1996
[4] C.W. Fraser and D.R. Hanson, "A Retargetable C Compiler: Design and Implementation", New York, The Benjamin/Cummings Publishing Company, 1995
[5] Matlab/Simulink Reference manuals, Math Works, Inc. Mass. USA 1993
www.grundfos.com
