IEEE – APEC ‘99 March 14-18, 1999
AC Drives: Year 2000 (Y2K) and Beyond Russel. J. Kerkman, Gary L. Skibinski, and David W. Schlegel Rockwell Automation Standard Drives Division 6400 W. Enterprise Drive Mequon, WI 53092 (414) 512 – 8263 (414) 512 – 8300 fax
[email protected] Abstract: The unprecedented growth in industrial motor drives over the past decade resulted from the process improvements demanded by the automation industry. With the increased speed of modern micro-controllers and digital signal processors and the reduced Insulated Gate Bipolar Transistor (IGBT) drive package size, coupled with the low maintenance of ac motors, multiple machine configurability has been achieved with minimal process downtime. As we will see, this progress required significant effort in mitigating numerous technical problems associated with this rapidly expanding technology. The paper reviews the economic trends in the industry, performance enhancements for Volts per Hertz (V/F) drives and high performance Field Oriented Controllers (FOC), inverter hardware topologies, and IGBT induced system problems and solutions. Finally, the authors will look beyond the year 2000 (Y2K) and offer a glimpse at what is possible and practical over the next five years.
I. INTRODUCTION The growth in industrial motor drives over the past 10 years has exceeded 25%, a rate far exceeding the previous 30 years. This unprecedented growth results from the increased demand for efficient reliable process control, power train flexibility, process improvement, and process control complexity. With the huge induction motor installed base, the continued drop in dollar per instruction cycle of modern micro-controllers and Digital Signal Processors (DSP), and the reduced Insulated Gate Bipolar Transistor (IGBT) drive frame size has allowed end users to increase product quality with a marginal cost increase. This progress required significant effort in mitigating numerous technical problems associated with these new technologies. This paper will review the present state of the ac drives industry. Including the economic status and trends within the industrial drives market. The industrial workhorse, the Volts per Hertz (V/F) control, is examined in a historical context. Although the technical literature features tens of contributions each year to high performance controllers, very little exists on the performance enhancements incorporated in today’s V/F option. Standard V/F functions include current limit control, line loss, and reconnection to rotating motors. High performance drives are discussed next, especially Field Oriented Controllers (FOC) and sensorless vector control. The evolution of FOC and the development of hybrid controllers have resulted in reliable, high performance ac drives that have tackled application requirements formerly the domain of dc drives.
The inverters power topology has undergone constant change in order to fully optimize device capability and availability. Among the first topologies was the Current Source Inverter (CSI) which was later augmented with the Voltage Source Inverter (VSI), based on application. The control platform started with six step and transitioned into modern Pulse Width Modulation (PWM). Among the power devices that optimized each topology were the Silicon Controlled Rectifier (SCR), Gate Turn-off Thyristor (GTO), Bipolar Junction Transistor (BJT) and IGBT. Third generation IGBTs with rise times ranging from 5,000 to 10,000 volts per microsecond now provide designers with devices that switch at carrier frequencies above the audible range. With device ratings from 600 to 3300 V at currents between 400 and 2400 A, IGBTs now assume roles previously reserved for GTOs. Higher switching frequency, standardization in packaging concepts, silicon die geometry, performance enhancements, and cost structures are discussed. These advances have not occurred without significant difficulty. The paper reviews the attendant problems with the paradigm shift in technology, specifically motor overvoltage, bearing currents, and common mode or Electro Magnetic Interference (EMI), and some of the mitigation methods. Finally, this paper peers into the market beyond Y2K. This paper will try to place into context many of the recent advances and prognosticate on what can realistically be expected in the near future. II. PRESENT CONDITION OF AC DRIVES This section of the paper will examine the economic status and trends within the industrial drive market. Standard V/F and sensorless control platforms and relevant reliability and performance features mandatory for modern industrial drives will be discussed. Finally, the section will end with a discussion on inverter hardware topologies and power device technology. A. Economic Status. To assist in providing an economic status, the “1998 ac drive worldwide outlook” report from Automated Research Corporation was reviewed [1]. The report indicated the worldwide sales of ac drive hardware, software, and support services was 4.85 billion dollars. The report contained the breakdown of total sales by geographic region, Europe, the Middle East, and Africa are lumped together with 39%, Japan
460 V Drives 1000000 Volume (Cubic Inch)
with 27%, North America with 21%, Asia with 12%, and Latin America with 1%. The report also indicated the breakdown of total sales by drive output rating. The ac drives industry is dominated by the low power market, the 1-4 kW range with 21% and the 5-40 kW range with 26% of total sales. The higher power market consists of the 41-200 kW range with 26%, 201-600 kW range with 16%, and the 600 kW and higher range with 11% of total sales.
100000 1980's 1990's
10000
Modern 1000 100
B. Drive Industry Trends.
1
In today’s automation market, stand-alone drives dominate the marketplace. At an increasing rate, they are being replaced with intricately controlled drive systems using a computer based control platform with a multitude of drives, motors, sensors, and processing equipment. Automation has placed increasing demands on the industrial drive market regarding volume, functionality, configurability, and cost. The real estate required to mount a drive has become increasingly expensive and has forced suppliers to decrease the drive size, but with increasing functionality. A survey of three-phase industrial drives over the last 20 years indicates the volume of the drive package has dramatically decreased, primarily due to faster, rise-times leading to reduced switching losses as shown in Figs. 1 and 2. The volume is defined as the drive’s length times the width times the height, as provided by the drive user’s manual. Fig. 1 portrays the 230 volt class of ac drives. The top line a typical 1980’s drive, the middle line is derived from the offerings of several industrial drive suppliers in the early 1990’s, and the bottom line is derived from the offerings of several modern industrial drive suppliers. As shown, only one vendor is shown for 1980’s drive, the 1990’s drives had several frame breaks, while the modern drives had almost an incremental frame size available. Fig. 2 shows the 460 volt class of drives. The top line is for drives from the 1980’s, the middle line is derived from the offerings of several industrial drive suppliers in the early 1990’s, and the bottom line is derived from the offerings of several modern industrial drive suppliers. As shown, numerous frame breaks were present with all the drives, including the fractional offerings.
10
100
1000
Horsepower
Fig. 2. 460 volt drives volume trend.
A survey of three phase industrial drives over the last 20 years indicates the power electronics devices used to invert the dc link voltage back to ac sinewave voltage has varied from GTO’s, BJT, and IGBT’s, but the device current rating has remained fairly consistent per horsepower application over time. Fig. 3 demonstrates the power device current rating as a function of drive horsepower. The trend lines were derived from the offerings of several industrial drive suppliers, both past and present utilizing GTO, BJT, and IGBT technology modules. As expected, the top line which represents the 230 volt drives (typically using 600 volt modules) had current ratings higher than the bottom line which represents the 460 volt drives (typically using 1200 volt modules), due to the current loading. The 575 volt drives module current rating closely paralleled the 460 volt data. A survey of three-phase industrial drives over the last 10 years indicates the cost of a drive has decreased over time. Fig. 4 demonstrates the cost associated with a drive as a function of horsepower. The top line is from two 1990’s vintage drives, while the bottom line is derived from the offerings of several modern drive suppliers. C. Performance and Reliability. Although the technical literature is filled with contributions to the specific areas of FOC and sensorless operation, very little exists on the mundane drive requirements – functions all Power Device Rating
230 V Drives 800
1000
1980's 1990's Modern
100
10 0.1
1
10
100
Horsepower
Fig. 1. 230 volt drives volume trend.
Current Rating (Amp)
Volume (cubic inches)
10000 600 230
400
460
200 0 0
25
50
75
100
Horsepower
Fig. 3. Power device current rating trend.
Average Industry List Price 15000
1990's 460 V
Dollars
12000 9000
Modern 460 V
6000 3000 0 0
10
20
30
40
50
Horsepower
Fig. 4. Average industrial drive cost trend.
drives must contain [2-4]. The following are a few of the fundamental features: Line loss control; Reconnect to spinning motor; Dynamic braking; Flux braking; DC braking; and Dead time compensation. When the dc bus voltage dips below a predetermined level, an ac line disturbance is sensed and the drive generally faults. Critical processing applications override the fault until power is restored or the bus voltage falls below a minimum level, which faults the drive. Drive suppliers may provide an inertia ride through option, which is one method of limiting line disturbance faults. In this case, the load decelerates and its inertia regenerates power onto the dc bus until the minimum bus voltage condition trips the drive. The ability to lock onto the rotational frequency and regain motor control allows a processing line to remain operational. Of the possible approaches, the most common fall under two main categories – current or voltage frequency lock. The strategy, depending on the manufacturer, often ramps the applied frequency from either the maximum or last known frequency until the sensed frequency matches the command. Fig. 5 displays the results of a reconnect strategy where a Phase Lock Loop (PLL) on terminal voltage drives the operating frequency. The drive is interrupted and the motor coasts for approximately 1 second. Upon reconnection, the drive catches the motor and reverses direction in response to a commanded change in direction.
The inclusion of an external dynamic brake allows the rapid duty cycling or hard deceleration of an overhauling load without causing an overvoltage fault. An external dynamic brake consists of a low resistance, high wattage resistor bank that can be placed across the dc bus via a semiconductor. The rate of switching depends on the voltage across the dc bus. Below the threshold voltage, the dynamic brake is inactive. Above the threshold voltage, but below the overvoltage trip level, the resistive load is placed across the dc bus to bleed off the excess voltage. The duty cycle of the semiconductor depends upon the dc bus voltage and its duration. A second method of braking is flux braking, which raises the motor flux, thereby increasing motor losses and causing faster deceleration times. When below base speed, these additional losses assist in slowing the motor. When above base speed, the flux current is reduced in the field weakening control and flux braking is normally not used. As speed decreases, the flux current increases until there is enough voltage margin to run rated motor current. The maximum flux (d-axis) current is equal to the drive current limit. A third method of braking is dc braking, which injects dc current into the motor. The dc current sets up a stationary field, which induces a rotor current. The motor absorbs the rotational energy dissipating a substantial portion as heat. One the most critical technical problems faced in the development of ac VSI drives is dead time compensation. Dead time is the time between the turn on and off of the upper and lower devices in a leg of the inverter. This time is designed into the inverter control to prevent a short circuit in the dc bus. Dead time with inadequate compensation is a major cause of instabilities and inferior performance. The quantity of literature attests to the importance placed on dead time compensation in today’s drives [5-8]. Two basic approaches exist. One uses the polarity of the current and corrects the voltage with feedforward control. The other approach employs voltage feedback and corrects the voltsecond error. Fig. 6 shows the response when the former approach is implemented. Prior to activation the induction motor was unstable. Once the dead time compensation is active, the motor stabilizes [7].
Deadtime Comp Activated
Fig. 6. Experimental results without and with dead time compensation. Fig. 5. Experimental results of a reconnection to a spinning motor and then a reversal.
D. Sensorless Control.
r1
Today’s general purpose VSI drive has a user selectable sensorless vector option. Applications requiring low speed performance exceeding the performance of V/F but without the penalty associated with position feedback will often be met by selecting the sensorless option. Among the application characteristics warranting sensorless operation, controlled torque for smooth acceleration and running, less overshoot and reduced potential for bus over voltage, zero or low speed, and position control are most prominent. Specific applications with typical specifications include printing (speed range 100:1), coating (smooth accel/decel), pulp (speed +/- 1%), metal (speed loop 20 radian/sec, 50:1) [9]. The amount of literature dedicated to the pursuit of low speed sensorless operation is extensive [2,3,10,11]. Proposed control schemes may be classified into at least four main categories. The first category basically improves upon the V/F control. Essentially a simple stator resistance and slip compensator, the enhanced V/F sensorless is relatively simple in principle. Although requiring a determination of the stator resistance and nominal flux conditions, its primary advantages in addition to simplicity is no voltage feedback or outer velocity loop is needed. Fig. 7 gives a typical block diagram of the control. The voltage drop associated with each component of current – flux and torque – is compensated for and the resultant voltage vector applied to the PWM modulator. In this manner, the steady state stator flux is maintained at a fixed value over the speed range. In the past, the resolution of the feedback current into its flux and torque producing components was neglected. In this case a voltage equal to the product of the stator resistance and magnitude of the stator current was added to the voltage neglecting the phase angle. However, this technique is seldom employed even when a single current sensor is used in the dc link. The second category incorporates more details of the motor in calculating the applied voltage and frequency of operation. Unlike the enhanced V/F controller above, this sensorless approach also incorporates a velocity loop and a PLL. One of AC Line I feedback
Voltage Boost V/F
Slip Comp X
Σ
V dc V
ωr
ωr
*
Σ
1 S
θ
PWM Controller
Current Limit
Rr
IR Comp
Fig. 7. Enhanced V/F block diagram.
Motor
∆
V
ACR 2
Id*
Vu
Vd
Vv
PWM
Vw
Inverter
CT 1+ 2
Vq L1 Kv
ωr ∗
r1 ω 1∗
Iq* ASR
ωr
ω ∗∗ 1
ACR 1
θ
PRG 1
ωs
p Kω
P Iu
Iq CS
ω 1∗∗
Iv Iw
Id Induction Motor
Fig. 8. Block diagram for a sensorless drive with a velocity loop and a PLL.
the first disclosures of the approach appeared in [12]. Fig. 8 is a block diagram of the control. The main control loops consist of an Automatic Current Regulator (ACR) on the flux producing or d-axis stator current. This ACR is typically a Proportional Integral (PI) compensator. Feedforward terms r1*Id* for the stator resistance drop and a (l1+l2)*ω 1***Iq, speed voltage, are added to the ACR2 output to form Vd. The second channel consists of two compensator loops. First, there is a Velocity Regulator (ASR) formed by comparing the velocity command (ω r*) with the estimate of the rotor velocity (ω r). Velocity estimation is achieved by subtracting the slip from the electrical frequency (ω 1**). The slip is proportional to the torque current (Kω*Iq). The output of the ASR is the torque current command, which forms the command to the third regulator - a PLL current regulator ACR1, which is compared to the Iq of the motor. The output of ACR1 forms the nominal frequency ω 1* used for the feedforward terms of the voltage command and is modified by the derivative of Iq to form the electrical frequency (ω 1**) which is integrated to form the electrical angle and fed back to the velocity loop. The addition of the derivative term helps in stability and mitigates the algebraic loop. Finally, the qaxis voltage (Vq) is determined by summing a resistance drop with the speed voltage term (L1*ω 1**). The third category employs a PLL on either the flux [13] or voltage [14]. In the case of Fig. 9, a rotor flux field orientation is achieved by regulating the rotor flux, obtained from an observer, to one axis. This is accomplished by forming an error signal consisting of a derived and calculated flux, which is based on the feedback current and motor parameters. The error signal is fed to a compensator circuit (PI) with the output an estimate of the rotor velocity. Like the second category, this approach requires a velocity regulator. Needing voltage feedback, this approach is more complicated and costly than either category 1 or 2.
R 1 + σL 1 p.
id
1 p.
Σ
X
λ' d
sin (ϕ)
vd Σ
Ka +
Kb
1
p.
p.
ψ 1*
ϕ
vq
iq
PWM Inverter
Motor
ψ1 Current feedback
λ' q
1 p.
Σ
Switching Table
Σ T
p. ϕ R 1 + σL 1 p.
Σ
T*
X
cos (ϕ)
Calculator
Voltage feedback
ϕ sin (ϕ)
cos (ϕ)
1 T2
Id ϕ Σ
X
Iq ϕ
Σ
X
1 p.
D N
X
Fig. 10. Block diagram of a simple DSC control.
ωr
Σ
X iq
Σ
( p. ϕ - ω r )
id
Fig. 9. Block diagram for a sensorless drive with a PLL on the flux.
The last category is direct self-control [15,16]. Although the control has evolved over the years, Fig. 10 shows the basic elements of the control. Two hysteretic controllers feed a lookup table that decides the inverter state. The control inputs are the torque and flux errors, which are derived from the current and voltage feedback signals. This control has been refined and is capable of torque control at zero mechanical speed. Recent contributions to sensorless control have explored the application of specialty motors and high frequency injection [17,18]. Both technologies attempt to use the drive motor as a resolver through subtle changes in the motor’s design and/or through high frequency injection of current or voltage. These techniques exploit the motor’s anisotropic characteristics, which if necessary may be accentuated in the design. A recent contribution promises zero electrical and mechanical speed operation without modifications to the machine geometry or design [19]. As mentioned above, the application determines the performance demanded from the drive system. Field experience confirms the performance from commercially available sensorless drives is application dependent. Table 1 lists values of merit for a number of drives. Speed range represents the experimentally determined lowest speed for
which acceptable performance was obtained with base equal to 60 Hz electrical. The quality of the remaining figures of merit is depicted in Figs. 11 and 12. From the torque vs. speed profiles, it is easy to estimate the relative speed range, starting capability, and dynamic speed accuracy. Clearly, drive C (Fig. 11) shows a larger speed range, better static and dynamic speed accuracy when compared to drive A (Fig. 12). As a result, an application requiring a slow speed reversal with good dynamic speed control through zero and subjected to a high dynamic load change needs the performance of drive C (Fig. 13). Clearly, disturbance rejection and good noise immunity are needed to ensure repetitive and reliable operation. E. Improving Reliability and Dynamic Performance. Early in the development of FOC and sensorless control, technologists recognized performance rests with accurate motor models and parameters [20]. Recent contributions in this area indicate the performance improvements require refined motor models, improved on-line adaptation, and commissioning procedures [14,21,22]. Fig. 14, for example, shows the variation in torque with stator resistance for a 5 hp motor; operating frequency is the parameter of variation. Accurate torque control at low speed requires knowledge of the machine’s stator resistance. One method reported on in the literature incorporates an angular change in the reference frame, thus eliminating the corrupting influence of the stator resistance. Fig. 15 demonstrates the dynamics of this control at low speed and sensorless operation. Inserting and removing a resistor bank artificially altered the value of the stator resistance. The disturbance to the torque was minimal and resistance estimation demonstrates rapid convergence.
TABLE 1. Comparison of Commercially Available Sensorless Drives Drive Supplier Speed Range
A 1:5 - :30
B 1:5
C 1:120
D 1:60
E 1:120
F 1:120
G 1:120
H 1:40
Static Speed Accuracy (0-100% rated) Dynamic Speed Accuracy (@ 30 Hz Speed) Speed Response (BW) (@ 30 Hz Speed) Starting Torque
-0.3/+0.0%
-3.0/+0.0%
-0.5/+0.0%
-0.12 /+1.6%
-1.0/+0.0%
-0.5/+0.0%
-0.55/+0.0%
-2.7/+0.0%
7.0% sec
no data
0.8% sec
2.7% sec.
0.55% sec.
0.39%sec
0.25% sec.
6 rads.
no data
13 rads.
13 rads.
11 rads.
20 rads.
10 rads.
no slip comp 10 rads.
50 – 150%
100%
150%
150%
150%
125%
150%
150%
Fig. 11. Experimental results of the torque vs. speed from drive C with good dynamic response. Fig. 14. Percent torque deviation resulting from stator resistance variation.
Fig. 12. Experimental results of the torque vs. speed from drive A with poor dynamic response. Fig. 15. Experimental results from a step change in stator resistance.
Fig. 13. Experimental results of a slow speed reversal from drive C with good dynamic response.
Commissioning procedures are necessary to determine controller gains and Model Reference Adaptive Controller (MRAC) parameters [14,22,23]. Accuracy of the parameters is dictated by performance requirements, which ultimately determines the commissioning tests and feedback resolution. Well established system identification procedures may be
adequate for a particular drive algorithm and performance, but prove inadequate or more importantly result in an unstable or under-performing drive in other applications. Thus, commissioning procedures are an integral and critical part of today’s drive system [14,23-25]. VSIs with 3rd generation IGBTs primarily control today’s industrial ac motor drives. These devices together with modern micro-controllers or DSPs provide reliable, cost effective industrial controllers. This has not occurred without overcoming significant technical hurdles, not fully anticipated or understood. These unintended consequences include but are not limited to terminal overvoltages, bearing currents, and common mode interference. References [26-28] are but a few of the vast amount of literature available. Fig. 16 is a classical example of the interaction of low rise time IGBTs and the PWM controller. This double pulsing condition produces terminal voltages in excess of two times the dc bus potential. The results clearly show the delay in arrival of the inverter line to line voltage at the motor terminals, the excessive terminal voltage, and a voltage disturbance at the inverter terminals resulting form the reflected wave returning to the inverter [26,29].
500 v/div
Inverter
0 1670 Vpk
Motor 500 V/div
0
0 50
5
10
15
20 25 Time (µ sec)
30
35
40
45
Fig. 16. Experimental results of IGBT double pulsing and the corresponding reflected wave at the motor terminals.
F. Power Topology Trends The evolution of power topologies has transitioned either by a new topology dictating a new device or a new device dictating a new topology. In the 1930’s, a patent assigned to Mittag of Germany described the first Auto Sequential Current Source Inverter (CSI) topology under six-step control using an Ignitron device with a 100 microsecond turn off time. Modern ac drives started with the SCR device, first introduced in 1957 at the GE research lab. The SCR (with reduced turn-off time) created the first of many topology and device “technology wave” trends over time. These trends are shown in Table 2. In the late 1950’s and early 1960’s, this revolutionary new SCR dictated the creation of VSI topologies. The variable dc link bus voltage in the VSI, evolved naturally from the 1930’s phase control technology, by changing converter gas filled Ignitrons to solid state SCR’s and adding a dc link inductors and capacitors. AC motor control was obtained by slaving dc bus voltage with the dc to ac inverter output frequency desired, while maintaining a constant V/F ratio on the motor to prevent iron saturation and improve torque output at variable speeds. Output frequency was transitioned to six-step control with SCR inverter grade devices (faster recovery times than converter SCRs) using analog Diode Transistor Logic (DTL) control boards. The McMurray auxiliary commutated inverter topology introduced in the mid 1960’s, suffered from inavailability of fast recovery SCR and fast, soft freewheel diodes devices. This resulted in large peak recovery voltages that were above the present SCR blocking voltage ratings. The new topology value was the potential to increase motor voltage fidelity by using (Abbondonti, Patel and Hoft) harmonic elimination schemes in the VSIs that utilized many devices. The power semiconductor industry responded to the challenge over the
next 15 years with faster SCR turn-on and turn-off time ( ~ 12 microseconds), higher SCR voltage blocking ratings, and better freewheel diodes for this topology. In the meantime, the device parasitics dictated building and selling inverters using a six-step McMurray-Bedford scheme, which included extra magnetics, resistors, and snubbers to deal with the device parasitics. Inverter control progressed in the analog world with the introduction of the Operational Amplifier (op amp) Integrated Circuit (IC) and Transistor Transistor Logic (TTL) ICs for fault logic and inverter firing control. In the 1970’s, the availability of high current converter SCR for rectifier phase control of the dc link current regulated link inductor and availability of the required high blocking voltage SCR for the inverter, this dictated a return to the old CSI topology. The simple, robust, six-step CSI structure in the 5 - 250 hp region was built by Landis, Rettig, Phillips, and Maslowski at Louis Allis, as well as at Robicon. Lipo at GE, as well as others at Westinghouse completed higher hp CSI modifications, such as the Load Commutated Inverter (LCI). Six-step analog control and digital firing logic progress continued with the introduction of improved op amps and further control integration using Complementary Metal Oxide Semiconductor (CMOS) ICs. During the late 1970’s, Novotny started to lay the analytical foundation of the dynamics of VSI and CSI ac drives. The dawn of present day topologies came in the early 1980’s, with lower cost, low hp, fixed dc bus, six switch PWM VSI. This topology slowly displaced the CSI topology and older VSI topologies that required auxiliary commutation. The reasons being: (i) the fixed dc bus front end was now a simple diode bridge rather than expensive converter SCRs used to vary the dc bus value. (ii) The availability of new power BJT devices in a TO3 package simplified the topology and allowed cost removal of the auxiliary commutation circuits, (iii) the arrival of Large Scale Integration (LSI) control ICs and microprocessors allowing the PWM work of Black at Bell labs in the 1930s and 1940s to become practical in a real time sense. One of the first transistorized 10 hp PWM inverter was done by Murphy, Stitch, and Gilmore at Allis Chalmers power electronic center as well as T B Woods. However, it was the smaller companies such as Lovejoy, Parametrics and Power Transistor Inverter (PTI) that were the leading forces to market. Plunkett at GE was one of the first to investigate PWM transistor drives for electric vehicles. In the mid 1980’s, the medium hp PWM VSI market was enhanced by paralleling TO3 package technology and the introduction of the D-60T dual Darlington at Westinghouse. The D-60T was one of the first high blocking voltage BJTs with fast risetime (~ 1 -2 microseconds) to reduce device switching loss. Topology drawbacks of these devices were the problematic paralleling of the TO3, due to interconnection parasitics and low tolerance to a short circuit. The D60T required costly snubber circuitry and a low current gain of 10, which required expensive base drivers. Fuji and Mitsubishi Semiconductor companies solved the low current gain problem by supplying silicon with a normalized gain of
100 for every device ampere rating. In addition they improved the short circuit protection of the device. Module popularity exploded in incremental ratings up to 1000 A and 1200 V ranges, due to easy planar thermal interfaces to heatsinks and that the problematic interconnection parasitics of parallel silicon die were transparent to the user. Improvements have continued into the 1990s in short circuit protection, improved Reverse Bias Safe Operating Area (RBSOA), reduced topology snubber requirements and device thermal cycling of the bond wires and substrate. In the mid 1980’s, the 100 kVA to 900 kVA range was initially dominated by GTO devices. An Undeland regenerative snubber topology helped to reduce high snubber loss due to the GTO’s slow risetime and especially long turnoff times. The LCI remained the Mega power topology. In the early 1990’s, the 1st generation epitaxial based IGBT gate voltage driven devices (~ 400 nanosecond risetime) were directly replacing the gate current driven BJT modules (~ 2 microsecond risetime) in the PWM VSI topologies. The IGBT in the same PWM VSI topology offered reduced switching times and therefore reduced switching losses, reduced heatsink size and lower cost base drives rather than the BJT gate current drivers. As a result, overall drive cost reduction potential was possible. IGBT users have asked for faster risetimes, up to the devices limit, along with better freewheel diode performance to reduce switching loss. Also, control of IGBT inverter topologies have progressed further into full scale integration with DSPs in both the analog current loop and velocity loop. In the late 1980’s, a case of new device technology dictating a new topology is the MOS Controlled Thyristor (MCT) by Temple at GE. The MCT’s lower RBSOA is well suited to switching in the Auxiliary Commutated Resonant
Pole Inverter (ACRPI) by De Donker and Lyon. In the late 1980’s and early 1990’s, new soft switch resonant inverter topologies such as the Resonant DC Link (RDCL), the Active Clamp Resonant Link Inverter (ACRL) and Quasi-Resonant Inverter were introduced by Divan with a continuous following of other resonant inverter topologies. These topologies could use existing device parasitics in the design and possibly extend the device current range due to lower switching losses. Presently, the device of choice for most hp sizes is clearly the IGBT. This is due to its low gate drive cost, low switching loss - a result of 50 nanosecond rise and fall times, improved freewheel diodes and extended high blocking voltage capability now possible with the Non Punch Through (NPT) device technology of Siemens/Eupec. Presently, the topology wave of choice remains the PWM VSI due to demonstrated cost and reliable performance over the last decade. Although, the beginnings of a resonant inverter wave are possible, the addition of extra silicon devices or resonant components makes it difficult to compete cost wise with the six, single, snubberless devices in the hard switch PWM VSI. G. Inverter Power Device Technology. The evolution of power modules has transitioned through many power devices including the BJT, GTO, Metal Oxide Semiconductor Field Effect Transistor (MOSFET), and finally to the IGBT. Simplistically, the IGBT is a monolithic silicon element with a fast switching MOSFET controlling the base of a high power BJT with conductivity modulation reducing the MOSFET drain to source on-time resistance. Numerous improvements have been made since Baliga at GE
Table 2 Historical time line showing when the Device has dictated the Topology required and when the Topology has dictated the Device Required
Device Technology Waves
? time
Time: Topology:
1930’s CSI
1950s VSI
Device:
Ignitron
SCR
SCR no fast SCRs no diode yet
Control:
6 step
6 step
6 step
Analog/ DTL
Analog/ TTL
Control Type: mechanical / Commutator
1960s VSI VSI McMurray McMurray - Bedford
1970’s CSI
1980’s VSI
2000 & BEYOND ? VSI ?
IGBT 1 ST - 4th modules, PT & NPT
IGBT ? SiC? Modules?
PWM
PWM
????
A/Digital LSI / Micro’s
A/ more Digital DSP
????? ?????
high Voltage BJT SCR T03
Programmed 6 step PWM Analog/ CMOS
1990’s VSI
BJT moduleS
and Russel at RCA introduced IGBTs in the early 1980’s [30]. Device latch-up, parasitic capacitance, switching losses, current sharing, and the RBSOA are all issues that have received attention. Although the most attention may have been spent in the characteristic modeling area [31-36]. When the IGBT was first introduced, the general consensus was to switch smaller drives at 20 kHz to minimize the switching losses and increase drive efficiency. A second advantage was to remain above the audible frequency range for air handling applications. Time has indicated that increased switching frequency has brought diminishing returns due to the dv/dt induced EMI. Thus, high power and general performance applications use slower switching speeds, while high performance uses higher switching speeds. A major change has occurred in module packaging. The main driver has been to ease the manufacturing process with a thru hole concept and shift away from intricate bus work. The second driver is a standardized package from the American, European, and Japanese vendors. One of the many standardized packages is a design with the electrical connections on the top and a flat metallic surface for thermal conduction via a heatsink on the bottom. A major improvement has been made in Intelligent Power Modules (IPM), which is an IGBT with added functionality integrated into the package. Besides the increase in voltage and current ratings, several features have been added. An additional PN junction may be placed on the die for thermal sensing. The PN junction exhibits a voltage change with temperature providing real-time monitoring of the actual junction temperature [37]. The gate drive circuit has also been placed on the silicon, therefore a reverse bias voltage isn’t needed to collapse the gate channel due to the direct contact with the die. Gate charge control is used to optimize switching loss and minimize deadtime. All these improvements are complemented with higher current density, as more power can be obtained from less silicon area. The reliability of the power devices has increased greatly due to changes implemented in the manufacturing process and better quality initiatives. New bond wire techniques and vacuum soldering process are used at the die to substrate junction, which decreases the chance of solder voids and increase the reliability. Power device levels have increased to include medium voltage applications. The High Voltage IGBT (HVIGBT) looks to take over the higher voltage applications to 3 kW. The HVIGBT has a similar package and low power gate driver circuitry like the IGBT. The Gate Commutated Thyristor (GCT) look promising for a GTO replacement with increased switching frequency and improved efficiency in a pressure contact package for medium voltage applications. The pressure contact package has minimized the bond wire inductance and provides two thermal conductive surfaces with stacking capability for three level designs [38]. The GCT is a single element, cost effective, gate controlled device that looks to minimize the gate driver complexity and snubber circuitry associated with GTO’s, although still in a tiered design [39].
The cost structure associated with the semiconductor industry has been driven down, primarily due to the dramatic price swing in the memory segment. However, the power device market is demonstrating sufficient growth to warrant the 4th generation IGBT product development and manufacturing process improvements. As with most aging products on the market, traditional price erosion has occurred. A better IGBT comparison method is to compare device performance vs. cost, rather than the dollars per amp traditionally associated with power devices. III.
Y2K AND BEYOND
In this section of paper, the authors will try their luck at prognostication. Specifically, this paper will try to examine, with as realistic an eye possible, the following topics: architectures, communications, control, power devices, power topologies, and finally power quality. A. Architectures Drives of the ‘90s, based on micro-controllers and DSPs with Field Programmable Gate Arrays (FPGA) or Application Specific Integrated Circuits (ASIC) for glue logic, will give way to Drives on a Chip (DOC). This will be possible because of the relentless increase in complexity, density, and speed of chip technology together with the rapid reduction in cost with the increasing die size of modern fabrication houses. The complexity and size of the today’s IGBT based VSI drive control board increases with performance. A basic low performing drive with minimal functionality would have a control board of approximately 6.5 cm2. A high performance drive’s control board requires more connectors for option boards, voltage feedback for example, interface hardware, communications options, etc and is approximately 8x11 inches. In addition to increasing cost, the additional complexity in hardware and software extends development time and increases resource requirements. B. Communications Y2K drives will see an integrated communications strategy, wireless communication, and Internet connectivity. Today’s drive, having a minimum 125-500k baud communication interface dependent on length, will see two orders of magnitude increase in communication speed. Amorphous software will adapt to user entered application data, thus new software structures and management process will be necessary to organize and control software development. C. Control Advances in power device modeling will improve the performance of control algorithms based on voltage observers. These improvements will be dependent on two technology drivers in the development of device behavior. First new devices with reduced turn on/off times will necessitate EMI mitigation through the design of PWM
controllers that restrict the voltage spectrum and common mode. If achieved, many practical limitations of present observers based on inverter switching states will be eliminated. Second, if in an effort to attenuate high frequency differential and common mode interference devices are controlled through a soft turn on gating sequence, accurate device modeling is then needed to account for volt-second turn on/off. Changes in device behavior will create opportunities for control engineers to improve the performance of voltage observers and incorporate more sophisticated controllers. New topologies with environmentally friendly characteristics will require development of control structures compatible with the topology. As was the case with soft switching inverter development, this aspect of developing commercial drives may delay or limit the introduction of the technology. Encoderless positioning, using the motor as a resolver through high frequency injection, will become the controller of choice for positioning. Sensorless positioning, a technical area of immense interest will appear as a user selectable option, similar to sensorlesss vector control is today. D. Power Devices Enhancements will be implemented in several areas for the 4th generation of IGBT’s, the power device endorsed for the near future. Higher breakdown voltage (1700 V) and current levels are being released for general distribution by numerous vendors. New technology has allowed the free wheeling diodes to be optimized for EMI/RFI noise reduction. The diode will exhibit lower dv/dt and oscillation free recovery at all operating temperatures and currents. In IPMs, gate charge will be used in conjunction with current monitors in the free wheeling diode path to minimize loss and preventing snap off of the diode during lower device turn on. The IGBT will incorporate new trench gate or advanced gate technology. Trench gate technology minimizes the J-channel Field Effect Transistor (JFET) resistance and increased cell density, resulting in reduced switching losses and a higher current density [40]. The issues surrounding the trench gate technology consist of lower short circuit capability and higher transconductance. The planar device will continue to use a finer pattern in the die geometry to allow the lower Vce sat characteristics. Either technology promises increased current density. The most promising die material technology has been placed on Silicon Carbide, but is several years from mass production and may experience price constraints. Beyond the traditional price and size limitations, the power device arena’s future demands the lowest possible dv/dt with the minimal switching losses in order to provide the best efficiency and maximum current density. More integration is necessary to minimize drive hardware (gate driver, snubbers, and fault detection) and ease manufactureability.
E. Power Topologies In the Y2K and beyond, drive cost and size will continue to decrease, while reliability and performance trends will increase due to further integration. In the short term following Y2K, the PWM VSI will remain as the predominant power structure due to cost constraints. However, the device used in the topology will see a paradigm shift to lower switching and conduction losses to further reduce heatsink size and drive cost. The low hp stand alone drive market will become ever increasingly a commodity purchased item. To reduce material, manufacturing and assembly costs, power structure integration with gate drivers, sensors and smaller control boards may be required. The first wide spread use of soft switch high frequency resonant topologies may be in the control board power supply to further reduce size, cost and EMI constraints. The distinguishing mark between drive vendors may be what features are supplied as standard. Once the cost and reliability of next generation Integrated Drive/Motor package (IDM) is proven in the hostile environment, there may be a large untapped user market switching to these packages as the millions of induction motors operated across the ac line are replaced by IDM’s to improve plant efficiency. The bulky, linear power supplies of the 1960’s had to quickly progress to smaller, lower cost Switch Mode Power Supplies (SMPS) to keep pace with computer (Volume / in3) and semiconductor industry dynamics. Likewise, the bulky induction motor in the Y2K may see a faster technology wave into material investigation and increased output torque production to keep pace with the ever-decreasing drive size. However, based on the fast 3 year response of the motor industry in switching from 1,000 Vpk insulation technology to 1,600 Vpk insulation technology to reduce motor failures on sub 1 microsecond risetime drives, the motor group will meet this challenge as well. The medium hp topology may see an onset of soft switch resonant topologies, if they can compete with the hard switch PWM VSI cost plus the cost of required external filters to reduce dv/dt. Further work on EMI, bearing current problems and motor insulation failures on long drive output cables is required to see if soft switching can avoid these problems. The high hp market will see further improvements drive size and increased carrier frequencies due to better IGBT devices. The medium voltage market is and will continue to undergo major topology changes as VSI technology is embedded with high voltage IGBTs. The long term view after Y2K is that power structure reliability will be the next distinguishing feature in this saturated commodity market. A new intensified focus on reliability science will provide a much higher Mean Time Between Failure (MTBF) number for the drive, approaching 20 year life. Focus on control and topologies to provide a “tripless” drive to customers during power line disturbances may also be an area of focus, if one wishes to replace the
across the line induction motors in critical applications with a drive and motor. The Integrated Regenerative Drive (IRD) presently uses an IGBT output inverter and an IGBT input converter on the front end that is interconnected with a common capacitor bank. This topology is presently used only in specific high regeneration applications due to cost of the front end, the associated magnetics, and switching EMI filters connected to the customer ac line. However, conversion of low cost, dc drive business to ac business in Y2K will require advanced IRD work on devices, control, magnetics, components, and filters to compete with dc. The ac/ac matrix converter of Venturini and Wood of Westinghouse in the 1970’s has a possible chance of arriving in the later years of Y2K. This is due to the sophisticated real time integrated control chips now available that were required to properly control the topology in the past. The new device required for this topology also may be fabricated as offshoots of present IGBT technology.
ACKNOWLEDGMENTS The authors would like to thank the many reviewers who provided comments and gave this paper a better perspective of the past, present, and future. REFERENCES [1] [2] [3]
[4]
[5]
[6]
F. Power Quality As the quantity of drives proliferate in Y2K and beyond, the issue of increased EMI will force most drive vendors to further integrate EMI filters into the product, if they have not already done so, for the European CE certification mark. Also, as the quantity of drives proliferate in Y2K and beyond, the secondary issues of drive efficiency and input harmonics, will become more important. As the cost of energy in Y2k and beyond increases, drive efficiency may become an important parameter for drive specification. The PWM VSI ac drive front end is typically a diode bridge, which contains 5th and 7th harmonic currents. The proliferation of many drives on a common ac bus may exceed IEEE 519 harmonic standards. However, real problems with VSI drives have only been observed in the field at off shore or local power generator sites. The high generator source impedance is up to 25% of base impedance and interacts with drive harmonics to flat top the ac line voltage. This results in reduced dc link voltage in the capacitor input filter circuit, which may trip the drive on undervoltage faults. Another observed failure modes is the harmonics interacting with high generator source impedance to distort the ac line voltage waveform, which is sensed by an older peak detect voltage regulator on the generator. This scenario causes wild voltage variations and oscillations. Active filter harmonic injection units are now possible to compensate for these ac line harmonic problems and may be the preferred and lowest cost solution to harmonics in Y2K and beyond.
[7]
[8]
[9]
[10] [11] [12]
[13]
[14]
[15]
[16]
[17]
[18]
IV. CONCLUSION This paper has reviewed past, present and future trends with respect to drive economics, architecture, control schemes, performance, power device, and power topology aspects pertinent to the industrial drive market.
[19]
“AC Drive Worldwide Outlook – January, 1998”, Automated Research Corporation, Dedhem, MA 02026 (781) 461-9100 A. Abbondanti. 1977, Conf. Rec. IEEE ISPCC, pp. 177-184. A Abbondanti and M.B. Brennen, “Varaible speed induction motor drives use electronic slip calculator based on motor voltages and currents”, IEEE Transactions on Industry Applications, Vol. IA-11, No. 5, 1975, pp. 483-488. T. Kataoka, Y. Sato, and A. Bendiabdellah, “A novel volts/hertz control method for an induction motor to improve the torque characteristics in the low speed range”, EPE 1993, Brighton, pp. 485-488. R. Ueda, T. Sonoda, Y. Inoue, and T. Umezu, “Unstable oscillating motor in PWM variable speed drive of induction mode and its stabilization,” IEEE Industry Application Society Conference, San Francisco, CA, October 4-7, 1982, pp. 686-691. Y. Murai, T. Wantanabe, and H. Iwasaki, “Waveform distortion and correction circuit for PWM inverters with switching lag-times”, IEEE Transactopns on Industry Applications, Vol. 23, No. 5, Sept./Oct. 1987, pp. 881-886. D. Leggate and R. Kerkman, “Pulse based dead time compensator for PWM voltage inverters”, IEEE Transactions on Industrial Electronics, Vol. 44, No. 2, April 1997, pp. 191-197. A. Munzo-Garcia and T. A. Lipo, ”On-line dead time compensation technique for open-loop PWM-VSI drives,” IEEE Applied Power Electronics Conference, Anaheim, CA, Feb. 15-19, 1998, pp. 95-100. I. Miyashita, A. Imayanagida, T. Koga, “Recent industrial application of speed sensorless vector control in Japan,” IEEE IECON’94, pp.1573-1578. “Sensorless Control of AC Motor Drives” Tutorial A, IECON’96, Taipei, Taiwan, August 5, 1996. “Sensorless Control of Electric Drives”, Tutorial Course, IEEE Industry Application Society Conference, San Diego, CA, October 6-10, 1996. T. Okuyama, N. Fujimoto, T. Matsui, and Y. Kubota, ”A high performance speed control scheme of induction motor without speed and voltage sensors,” IEEE Industry Application Society Conference, Denver, CO, September 28 - October 3, 1986, pp. 106-111. C. Schauder, “Adaptive speed identification for vector control of induction motors without rotational transducers”, IEEE Transactions on Industry Applications, Vol. 28. No 5, September/October 1992, pp. 1054-1061. R. J. Kerkman, B. Seibel, T. Rowan, and D. Schlegel, “A new flux and stator resistance identifier for ac drive systems”, IEEE Transactions on Industry Applications, Vol. 32, No.3, May/June 1996, pp. 585-593. I. Takahashi and T. Noguchi, “A new quick response and high efficiency control strategy of an induction motor,” IEEE Industry Application Society Conference, Toronto, Canada, October 6-11, 1985, pp.496-502. M. Depenbrock, “Direct Self-Control (DSC) of inverter-fed induction machine,” IEEE Transactions on Power Electronics, Vol. 3, No. 4, October 1988, pp. 420-429. P. Jansen and R. Lorenz, “Transducerless field orientation concepts employing saturation-induced saliencies in induction machines”, IEEE Transactions on Industry Applications, Vol. 32, No. 6, Nov/Dec 1996, pp. 1380-1393. F. Blaschke, J. van der Burgt, A. Vandenput, “Sensorless direct field orientation at zero flux frequency,” IEEE Industry Application Society Conference, San Diego, CA, October 6-10, 1996, pp. 189-196. J. Ha and S. Sul, “Sensorless field orientation control of an induction machine by high frequency signal injection,” IEEE Industry Application Society Conference, New Orleans, LA, October 2-6, 1997, pp. 426-432.
[20] L. J. Garces, “Parameter adaption for the speed controlled static ac drive with a squirrel-cage induction motor,” IEEE Transactions on Industry Applications, Vol. IA-16, Mar/Apr 1980, pp. 173-178. [21] L. Garces, R. Kerkman, T. Rowan, D. Schlegel, B. Seibel, and K. Stachowiak, “Operation of induction machines with electromagnetic irregularities”, IEEE Industry Application Society Conference, New Orleans, LA, October 2-6, 1997, pp. 151-159. [22] H. Schierling, “Self-Commissioning – A novel feature of modern inverter-fed induction motor drives,” IEE Conference on Power Electronics and Variable Speed Drives, U.K., July 13-15, 1988, pp. 287-290. [23] T. Okuyama, N. Fujimoto, and H. Fujii, “A simplified vector control system without speed and voltage sensors – effect of setting error of control parameters and their compensation,” Electrical Engineering in Japan, Vol. 110, No. 4, 1990, pp. 129-139. [24] N. R. Klaes, “Accurate off-line identification of the operating point dependent induction machine parameters,” EDS’90 – Electrical Drives Symposium, Capri, Sept. 25-27, 1990. [25] D. Borgard, G. Olsson, R. Lorenz, “Accuracy issues for parameter estimation of field oriented induction machine drives,” IEEE Industry Application Society Conference, Denver, CO, October 2-6, 1994, pp. 593-600. [26] R. Kerkman, D. Leggate, G. Skibinski, “Interaction of drive modulation & cable parameters on ac motor transients,” IEEE Transactions on Industry Applications, Vol. IA-33, No. 3, May/June 1997, pp. 722-731. [27] D. Busse, J. Erdman, R. Kerkman, D. Schlegel, and G. Skibinski, “Characteristics of shaft voltage and bearing currents”, IEEE Industry Applications Magazine, Nov./Dec. 1997, pp. 21-32. [28] G. Skibinski, J. Pankau, R. Sladky, J. Campbell, “Generation, control and regulation of EMI from ac drives,” IEEE Industry Application Society Conference, New Orleans, LA, October 2-6, 1997, pp. 15711583 [29] D. Leggate, J. Pankau, D. Schlegel, R. Kerkman, and G. Skibinski, “Reflected waves and their associated current”, IEEE Industry Application Society Conference, St. Louis, MO, October 12-15, 1998, pp. 789-797.
[30] B. Baliga, M. Cheng, P. Shafer, and M. Smith, “The insulated gate transistor (IGT): a new power device”, IEEE Industry Applications Society Conference, October 3-7, 1983, pp. 794-803. [31] A. Hefner, “An investigation of the drive circuit requirements for the power insulated gate bipolar transistor”, IEEE Transactions on Power Electronics, Vol. 6, No. 2, April 1991, pp. 208-219. [32] D. Berning and A. Hefner, “IGBT half-bridge shoot-through characterization for model validation”, IEEE Industry Application Society Conference Proceedings, San Diego, CA, October 6-10, 1996, pp. 1491-1499. [33] D. Metzner, T. Vogler, and D. Schroder, “A modular concept for the circuit simulation of bipolar power semiconductors”, IEEE Trans on Power Electronics, Vol. 9, No. 5, September 1994, pp. 506-513. [34] J. Hsu and K. Ngo, “Behavioral model of the IGBT using the Hammerstein configuration”, IEEE Transactions on Power Electronics, Vol. 11, No. 6, November 1996, pp. 746-754. [35] Z. Shen and T. P. Chow, "Modeling and characterization of the insulated gate bipolar transistor for SPICE simulation”, International Symposium on Power Semiconductor Devices ’93, May 18-20, 1993, Monterey, CA, pp.165-170. [36] F. Blaabjerg, U. Jaeger, S. Munk-Nielsen, and J. Pedersen, “Power losses in PWM-VSI inverter using NPT or PT IGBT devices”, IEEE Trans on Power Electronics, Vol. 10, No. 3, May 1995, pp. 358-367. [37] T. Kajiwara, A. Yamaguchi, Y. Hoshi, K. Sakurai and J. Gallagher, “New intelligent power multi-chips modules with junction temperature detecting function”, IEEE Industry Application Society Conference, St. Louis, MO, October 12-15, 1998, pp. 1085-1090. [38] H. Kirihata, Y. Takahashi, H. Wakimoto, and F. Niino, “Investigation of flat pack IGBT reliability”, IEEE Industry Application Society Conference, St. Louis, MO, October 12-15, 1998, pp. 1016-1021. [39] E. Motto, and M. Yamamoto, “New high power semiconductors: High voltage IGBTs and GCTs”, Powersystems World '98 Conference, Santa Clara, CA, November 7-13, 1998 [40] E. Motto, J. Donlon, H. Takahashi, M. Tabata, and H. Iwamoto, “Characteristics of a 1200 v PT IGBT with trench gate and local life time control”, IEEE Industry Application Society Conference, St. Louis, MO, October 12-15, 1998, pp. 811-816.
