BACKGROUND OF THE INVENTION
The invention relates generally to a semiconductor power device that controls AC power to a load, and more specifically to such a device having integral current monitoring.
In controlling the application of electrical power to a load, it is known practice to use a switching device in series with the load. Power MOS devices have proven to be very useful in this regard, at least for DC applications.
The terms MOS and MOSFET (which originally stood for metal- oxide-semiconductor) are used to refer to insulated gate devices generally, notwithstanding the fact that most modern devices have polysilicon gates rather than metal gates. AC switches using MOS technology are more complex in nature than DC switches, since AC switches need to conduct in both directions. It is known to create an AC switch by hooking up two MOSFET's as DC switches, either in a common drain or in a common source configuration. A somewhat simplified explanation for the AC operation of such a switch is that in a given polarity, one MOSFET conducts as a MOSFET while the other acts as a forward-biased diode in parallel with a MOSFET.
In most power electronics applications where AC power is controlled and delivered to a load, it is important to monitor and control the load current (which corresponds to the switch current) and the switch temperature. Certain properties of power MOSFET's make it possible to sense current, using the current mirror technique. Power MOSFET's
are normally implemented as a plurality of discrete cells connected in parallel with gates in common, sources in common, and drains in common. The current mirror (or current sensor, as it is sometimes known) comprises a relatively small region of cells having their gates and drains in common with those of the main transistor and having their sources commonly connected to a separate output terminal. For example, the current mirror might comprise ten MOSFET cells while the main transistor comprised 10,000. The current, flowing through the current mirror is found to be a predetermined fraction of the current flowing through the main transistor. If the mirror cells and the main cells are configured geometrically alike, the fraction corresponds roughly to the ratio of the numbers of cells (e.g., a current ratio of about 700:1 for a cell ratio of 1000:1).
SUMMARY OF THE INVENTION The present invention is drawn to a technique for providing dual direction current sensing in a power MOS device. This is accomplished with a single current mirror which is configured to provide the same current ratio in both directions for at least one predetermined temperature. The invention finds applicability in such applications as AC switches and synchronous rectifiers. The ability to configure a current mirror such that the ratio of main transistor current to mirror transistor current remains the same in both directions relies on a recognition of the reasons that the current ratio is normally different for the two directions of conduction. For drain-source current flow, conduction is solely through the MOS channel. Since a greater fraction of the mirror cells are on the periphery, the mirror tends to conduct more current per cell than the main transistor conducts per cell. For source-drain current flow, the
conduction is through the parasitic diode, and possibly through the MOS channel depending on the gate voltage. In general, the diode conduction tends to dominate. Diode conduction is largely a function of cell area, so the mirror tends to conduct the same amount of current per cell as the main transistor conducts per cell for this direction.
The invention contemplates any of a number of techniques for providing relatively increased diode conduction in the mirror in order to provide the same current ratio as when channel conduction is the sole mechanism. These include increasing the doping of the cell body, increasing the diode area per cell relative to the amount of MOS channel area, providing extra diode cells in the mirror, or locating the current mirror in the hottest part of the chip where diode conduction is greatest.
A monolithic AC switch having the single current mirror that provides current sensing in both directions requires only five external connections (two gate terminals for the two main transistors, two source terminals for the two main transistors, and a single source terminal for the current mirror) . While reduced pin count is a generally desirable objective, it has particular significance in this case since it allows the device to be packaged in a commercially available standard five-pin package. This provides a low packaging cost and low pin count, which render the AC switch suitable for the industrial, commercial, and consumer markets.
The monolithic AC switch may be configured to provide conductivity modulation so that the first and second main transistors operate like first and second insulated gate bipolar transistors ("IGBTS") . In such a case, some of the cell bodies of the first main transistor cooperate with cell bodies of the second main transistor to define the emitter of the bipolar transistor that forms part of the
second IGBT for current flow in one direction, and vice versa for the other direction.
A further understanding of the nature and advantages of the present invention may be realized by reference to the remaining portions of this specification and to the attached drawings.
BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1A is a simplified top view of a chip containing a MOSFET with a current mirror according to the present invention;
Fig. IB is a simplified cross-sectional view of one MOSFET cell;
Fig. 1C is a simplified top view of the MOSFET cell;
Fig. ID is an equivalent circuit schematic of the chip of Fig. 1A, plus external circuitry;
Fig. 2A is a simplified top view showing the use of special cells in the current mirror; Fig. 2B is a simplified cross-sectional view showing of a special cell in the current mirror;
Fig. 2C is a simplified top view of a special cell having reduced MOSFET conduction;
Fig. 2D is a simplified top view of a special cell having no MOSFET conduction;
Fig. 3A is an equivalent circuit schematic of an AC switch using discrete power MOSFET's in a common source configuration;
Fig. 3B is an equivalent circuit schematic of an AC switch using discrete power MOSFET's in a common drain configura ion;
Fig. 4 is a simplified top view of a monolithic AC switch according to the present invention;
Fig. 5A is a simplified view of an IGBT cell;
Fig. 5B is a simplified equivalent circuit schematic of an IGBT; and
Fig. 6 is a simplified equivalent circuit schematic of a monolithic AC switch with conductivity modulation.
DESCRIPTION OF THE PREFERRED EMBODIMENTS Fig. 1A is a simplified top view of a power MOSFET 10 formed on a semiconductor chip. MOSFET 10 can be used as a stand-alone device such as a synchronous rectifier suitable for use in such applications as high frequency, low voltage switching power supplies, or can be combined with another MOSFET (either discrete or monolithically formed) to define an AC switch, as will be described below. While the description will be in terms of n-channel devices, the application to p-channel devices is clear.
The chip is divided into a main transistor region 12 and a current mirror region 20. Region 12 defines the main transistor, designated T,, while region 20 defines the current mirror transistor, referred to as the mirror and designated T«. Transistor T, is implemented as a large number of individual cells (say 5000-10,000) connected in parallel, while mirror T« consists of a much smaller number of cells (say 20-30) . As will be described below, at least some of the mirror cells are different from the main transistor cells in order to provide a current mirror that behaves comparably for both directions of current flow. These mirror cells will be referred to as special cells.
Fig. IB is a simplified cross-sectional view of a normal MOSFET cell 35 as is found in transistor T, and mirror T-,. The normal cell is formed in an n- epitaxial ("epi") layer 40 which is deposited on the top side of an n+ substrate 42. Normal cell 35 comprises a p/p+ body 45 formed in epi layer 40 and a peripherally extending n+
source 47 formed within the perimeter of body 45. The source is slightly spaced in from the body perimeter so as to define a peripheral channel 48 at the epi surface. A polysilicon gate 50 overlies the channel (as well as the portions of the epi surface between cells) and is insulated therefrom by a thin layer 52 of gate oxide. A metal connection 53 connects the source and the central portion of the body to a source node S. A metal connection 55 connects the polysilicon layer to a gate node G. A metal layer 57 formed on the lower surface of substrate 42 defines a drain node D. Body 45 and epi layer 40 define a p-n junction (parasitic diode) that conducts when the source potential is sufficiently more positive than the drain potential.
Fig. 1C is a simplified top view of a normal cell, taken at the epi surface and shows channel region 48 extending around the periphery, source 47 spaced inwardly therefrom, and the central portion of body 45. The polysilicon layer overlies the plane of the figure, and its inner boundary is shown in phantom. Fig. ID is an equivalent circuit schematic of power MOSFET 10, together with external circuit elements normally connected to the chip during operation. Transistor T1 and mirror TM have common gate and drain nodes G and D, and separate source nodes S, , and S-,. Nodes G, S, , S„, and D are brought out as accessible pins to provide a 4-pin device. Each of the transistors is shown as having a parasitic diode connected in parallel. These diodes, which conduct in the source-drain direction are defined by the junction between the p/p+ body and the n-epi layer. Source nodes S, and D are connected in-line in a circuit. A resistor 60 is connected between source nodes S, and SM, and an operational amplifier 62 provides an output signal corresponding to the voltage drop across resistor 60.
The resistor may be off the chip as an external element, or may be formed on the chip.
Fig. 2A is a simplified top view illustrating a typical configuration of the layout of one group of embodiments of current mirror T . As with the main transistor, the mirror comprises an array of cells; however, at least some of the mirror cells are different from the main transistor cells. In the illustrative embodiment, most of the mirror cells are so-called normal cells 35, configured like the cells of transistor T, , while a number of the mirror cells are special cells 65 which have a different configuration.
As an aid to understanding the reason for configuring some mirror cells differently, it is helpful to consider how mirror T„ would behave if its cells were all the same as the cells in transistor T1. When conduction is from drain to source, the parasitic diode is reverse biased and conduction is solely via the MOS channel. Since the mirror region has a higher periphery-to-area ratio than the main transistor, the current through the peripheral cells can spread laterally beyond the cell boundaries. Thus, the current per cell in the mirror exceeds the current per cell in the main transistor, which consists of a greater fraction of interior cells. When conduction is from source to drain, the parasitic diode is forward biased and conduction is through the diode and the MOS channel in parallel. Conduction through the diode tends to dominate, and since it is primarily a function of cell area, the mirror and main transistor carry approximately the same current per cell. Therefore, the current ratios are different for the two directions of current flow, which tends to make the use of a single current mirror unsuitable for bi-directional current sensing. However, by configuring the special cells to provide relatively greater diode conduction, equal current
ratios in both, directions can be achieved, at least for a particular temperature.
Fig. 2B is a simplified cross-sectional view of one of special cells 65. Special mirror cell 65 is formed in a generally similar manner to the normal mirror cell, but is altered in one of several ways to provide proportionately more diode conduction (i.e. proportionately less MOSFET conduction). The special cell has a p/p+ body 45' and a metal connection 53' to the common source node SM for the normal and special cells in mirror M. However, the cell is formed either with a source region that extends only part way around the periphery of the cell or without a source region (making it a pure diode cell) . Special cells 65 can alternatively be configured with the same geometry as normal cells 35, including a source extending around the entire periphery, but with enhanced p+ doping in the body region. Fig. 2C shows one embodiment wherein the source designated 47' is formed only part way around the cell periphery so that MOSFET conduction only occurs for a portion of the periphery. Fig. 2D is a top view of another embodiment of the special cell in which no source region is provided so that no MOSFET conduction occurs.
A somewhat different approach requires no structural differentiation of the mirror cells. Since the diode conduction increases with temperature, it is sometimes possible to achieve the required enhanced diode conduction in the mirror by locating the mirror at the hottest portion of the chip (generally near the center) . To the extent the temperature differential on the chip is insufficient to provide the necessary degree of current differential, it is possible to combine other techniques for enhancing diode conduction.
Power MOSFET 10 (with its special current mirror) may be connected to another power MOSFET chip 70, to define
an AC switch. MOSFET 70 defines a transistor T2 with respective source, gate, and drain nodes S2, G2, and D2. Fig. 3A shows MOSFET's 10 and 70 in a common source configuration to define an AC switch 75. The common source configuration allows the use of a single gate drive. Fig. 3B shows the MOSFET's in a common drain configuration to define an AC switch 80. Fig. 3B also applies to a monolithic AC switch as described below.
Fig. 4 is a simplified top view of an AC switch 100 monolithically formed on a semiconductor chip. The monolithic version implements the common drain configuration of Fig. 3B. The same designations of the transistors and their nodes will be used.
The chip is divided into first and second transistor regions 102 and 105 separated by guard rings 107. Located within first transistor region 105 is a current mirror region 120. Regions 102 and 105 define the two main transistors, designated T, and T2, of the switch, while region 120 defines the current mirror, designated TM. Each of transistors T, and T2 is implemented as a large number of individual cells (say 5000-10,000) connected in parallel, while current mirror TM consist of a much smaller number of cells (say 20-30) . As will be described above in connection with MOSFET 10, at least some of the mirror cells are different from the main transistor cells in order to provide a current mirror that behaves comparably for both directions of current flow.
The circuit schematic of monolithic AC switch 100 (together with external circuit elements) is the same as discrete AC switch 80 except that the common drain need not be brought out as an accessible pin. As above, current mirror TM is associated with transistor T^, with which it shares gate and drain connections but has a separate source node S„. Nodes G-^, S-^, SM, G2, and S2 are brought out as
accessible pins to provide a 5-pin device. In the case, of each of the AC switches, the series-coupled MOSFET's are connected to a series-coupled AC source 130 and load 132, shown in Fig. 3B. In operation, when the output of AC source 130 is applied to load 132 and AC switch 75, 80, or 100, the current through the current mirror will be the same fraction of the current through the switch in both directions. Thus, the output of operational amplifier 62 will be a sine wave having an average value of 0. In the event that conditions change (such as overheating of the chip) , the diode conduction will increase, and the current flow in the mirror will increase for one direction. The output from operational amplifier 62 will then have a DC offset, which can be communicated to suitable control circuitry. It is also possible to configure the mirror for symmetric operation at a temperature other than the intended operating temperature. In such a case, normal operation would be characterized by a DC offset in the output of operational amplifier 62. It is also possible to provide multiple current mirrors, each configured for symmetric operation at a different temperature. This would allow a user to use the chip under various operating conditions.
The monolithic AC switch illustrated in Fig. 4, while having an operation similar to the discrete version illustrated in Fig. 3B, has an inherent characteristic due to the fact that transistors Tχ and T~ are fabricated on the same substrate. More particularly, if the two regions 102 and 105 are sufficiently interdigitated that a significant number of the MOSFET cells in transistor 1.1 are reasonably near the MOSFET cells in transistor T~, a degree of conductivity modulation occurs, and the MOSFET's behave to some extent like insulated gate bipolar transistors ("IGBT's"). The conductivity modulation occurs when the
conductivity of one MOS device is modulated by the injection of minority carriers from the other MOS device. It should be noted that guard rings 107 do tend to reduce the effect. Fig. 5A is a simplified cross-sectional view of a vertical IGBT cell 135 as is known in the prior art. The IGBT cell is formed in an n- epitaxial ("epi") layer 140 which is deposited on an n+ layer 141 formed on the top side of a p+ substrate 142. IGBT cell 135 comprises a p/p+ body 145 formed in epi layer 140 and a peripherally extending n+ source 147 formed within the perimeter of body 145. The source is slightly spaced in from the body perimeter so as to define a peripheral channel 148 at the epi surface. A polysilicon gate 150 overlies the channel (as well as the portions of the epi surface between cells) and is insulated therefrom by a thin layer 152 of gate oxide. A metal connection 153 connects the source and the central portion of the body to a source node S. A metal connection 155 connects the polysilicon layer to a gate node G. A metal layer 157 formed on the lower surface of substrate 142 defines an emitter node E. Body 145, epi layer 140, and P+ substrate 142 define the collector, base, and emitter of a pnp bipolar transistor.
Fig. 5B is a simplified equivalent circuit schematic of an IGBT as is known in the prior art. It does not show the parasitic NPN bipolar transistor. For a discrete IGBT, current flow occurs in one direction only (emitter to source) . Thus an AC switch using discrete IGBT's back to back would require external diodes.
Fig. 6 is an equivalent circuit schematic of the monolithic AC switch where significant conductivity modulation (IGBT action) occurs. The switch includes IGBT,, IGBT2, and IGBT^, associated with MOS transistors 1^ , T2, and M respectively. In this case, the bodies of at least some of the cells of transistor T2 act as the collector for
the PNP transistors in IGBT1 and IGBTH. The bodies of at least some of the cells of transistors T^ and T„ act as the collector for the PNP transistor in IGBT1. In the schematic, the bodies are denoted P1 P2. and PM. If the switch is made on a p+ wafer, the effect is to limit the potential drop in the n- epi layer for current flowing laterally.
In operation, current flowing from S- to S, goes through MOS transistor T2 and through IGBT.^ and IGBTM. Current flowing from S1 to S2 goes through MOS transistors T, and TM and through IGBT2.
As in the case of the switch with no IGBT action, symmetrical behavior of the current mirror is achieved by decreasing its MOSFET flow or enhancing its IGBT current flow relative to that through IGBT, by the use of special cells in the mirror. Reference to Figs. 2A-C is appropriate.
In conclusion, it can be seen that the present invention provides an AC switch with current sensing requiring only a single current mirror that operates symmetrically in both directions. While the above is a complete description of the preferred embodiment of the invention, various modifications, alternative constructions, and equivalents may be used. Therefore, the above description and illustrations should not be taken as limiting the scope of the invention which is defined by the appended claims.
Source: http://www.wipo.int/pctdb/en/wo.jsp?WO=1989012361&IA=US1989002371&DISPLAY=DESC
No hay comentarios:
Publicar un comentario