Integrated circuits often require voltage reference circuits. Reference circuits may be used to establish known voltage levels for controlling power supplies and other circuits. Ideally, reference circuits should exhibit good immunity to changes in process, voltage, and temperature (so-called PVT variations).
One popular type of reference circuit is the so-called bandgap reference circuit. Bandgap reference circuits exhibit stable behavior with respect to PVT variations, but are limited to producing output voltages at about 1.2 volts. Threshold-voltage-based complementary metal-oxide-semiconductor (CMOS) reference circuits have been developed that are capable of operating at lower output voltages, but this type of reference circuit tends to exhibit large amounts of process dependence, due to the dependence of threshold voltage on process (implant) variations.
As power supply voltages in modern circuits are scaled to lower voltages, there is a need to produce reference circuits that operate at voltages below one volt. It would be therefore desirable to be able to provide improved integrated circuit voltage reference circuits.
A reference circuit may be provided that has a pair of semiconductor devices. Each semiconductor device may have an n-type semiconductor region, an n+ region in the n-type semiconductor region, a metal gate, and a gate insulator interposed between the metal gate and the n-type semiconductor region. The metal gate may have a work function that matches the work function of p-type polysilicon. The gate insulator may have a thickness of less than about 25 angstroms. The metal gate may form a first terminal for the semiconductor device and the n+ region may form a second terminal for the semiconductor device. The second terminals may be coupled to ground. When a voltage is applied across the first and second terminals, current may tunnel through the gate insulator and the semiconductor device may exhibit a turn-on voltage of between 0.3 and 0.5 volts.
The reference circuit may have a biasing circuit that is coupled to the first terminals of the semiconductor devices. During operation, the biasing circuit may supply different currents to the semiconductor devices and may provide a corresponding reference output voltage at an output terminal. The reference voltage may have a value that is less than one volt.
Further features of the invention, its nature and various advantages will be more apparent from the accompanying drawings and the following detailed description of the preferred embodiments.
Voltage reference circuits are commonly used in integrated circuit designs where a stable voltage of a known magnitude is required. For example, some integrated circuits have power supply circuitry in which the magnitude of the power supply voltage that is produced by the power supply circuitry is regulated using a bandgap reference circuit.
Biasing circuit 32 may have an operational amplifier such as operational amplifier 28. The positive input terminal of operational amplifier 28 may be coupled to node 24. The negative input terminal of operational amplifier 28 may be coupled to node 26. During operation, operational amplifier 28 provides a corresponding output voltage Vout on output terminal 30 while maintaining the voltages on nodes 24 and 26 at equal values.
Diode MGLD1 has an anode coupled to terminal 24 and a cathode coupled to ground. Diode MGLD2 has an anode coupled to terminal 26 and a cathode coupled to ground. In one embodiment, diodes MGLD1 and MGLD2 are formed from metal-gate leakage diode structures that exhibit a relatively low turn-on voltage. The turn-on voltage of diodes MGLD1 and MGLD2 is generally about 0.3 to 0.5 volts, as opposed to the 0.7 volt turn-on voltage associated with conventional p-n junction diodes of the type used in bandgap reference circuits. The low turn-on voltage of diodes MGLD1 and MGLD2 (e.g., 0.3 to 0.5 volts, 0.4 to 0.5 volts, less than 0.5 volts, etc.) allows reference circuit 22 to produce a voltage Vout on terminal 30 that is about 0.8 to 0.9 volts. This sub-one-volt reference signal may be used in circuits that require low-voltage references such as low-voltage power supply circuits and other circuits.
During operation, diode MGLD1 is characterized by a junction voltage of VGB1 and diode MGLD2 is characterized by a junction voltage of VGB2. Biasing circuit 32 and operational amplifier 28 hold the voltage at both nodes 24 and 26 at about VGB1. The resistor network made up of R1′, R2′, and R3′ then ensures that the currents I1 and I2 have appropriate magnitudes (and I1/I2 has an appropriate non-unitary ratio) to set a desired value for ΔVGB=VGB1−VGB2. The value of ΔVGB is proportional to absolute temperature (PTAT), whereas the value of VGB1 is complementary to absolute temperature (CTAT).
The PTAT characteristic associated with ΔVGB (line 34 of
Vout=R2′/R3′(ΔVGB1)+VGB1 (1)
If desired, other biasing circuits may be used. The biasing circuitry that is used in the illustrative configuration of
The performance of low-voltage reference circuit 22 of
Metal-gate leakage diode 44 may be formed from a semiconductor substrate such as a silicon substrate. An n-type doped region such as n-well 50 may be formed in the silicon substrate. One or more heavily doped n+ regions 52 may be formed in n-well 50 (to form Ohmic contacts with the n-well) using ion implantation or other suitable doping techniques. The n+ regions are electrically connected to the n-well and therefore both the n-well and n+ regions form part of one of the terminals for diode 44 (i.e., its cathode). The n+ regions in cathode C may have associated metal contacts or other conductive terminal structures that are coupled to n+ regions 52 and that also form part of cathode C. As shown in
Gate insulator 48 may be formed on the surface of semiconductor substrate 50. Gate insulator 48 may be formed from a layer of dielectric such as silicon oxide, a hafnium-based oxide, other metal oxides, a nitride, oxynitrides, or other insulating materials. Quantum mechanical tunneling may allow current to pass through insulator 48 during operation of diode 44.
Conductive gate 46 may serve as anode terminal A. Conductive gate 46 is preferably formed from metal. If desired, conductive gate 46 may be formed from doped semiconductor. For example, conductive gate 46 may be formed from a p+ polysilicon layer when region 50 is an n-well. Such polysilicon-based gate structures are typically formed using self-aligned semiconductor fabrication processes and may involve an undesired amount of process complexity. Formation of gate 46 from metal, which generally avoids the need for self-aligned techniques, is therefore generally preferred.
In configurations in which gate conductor 46 is formed from metal, the work function of the metal is preferably chosen to approximately match that of p-type polysilicon. The work function of the metal may, for example, be within +/−0.5 eV of the work function of p-type polysilicon. This type of metal is depicted in
As shown in
The thickness of gate insulator 48 and the work function of gate conductor 46 may, if desired, be adjusted to adjust Vout and the amount of current that passes through diodes MGLD1 and MGLD2 (e.g., to produce a circuit configuration that exhibits reduced power consumption). With one suitable arrangement, the thickness TOX of insulator 52 may be about 13 angstroms (e.g., about 13 to 20 angstroms, less than 15 angstroms, less than 20 angstroms, about 13 to 25 angstroms, less than 25 angstroms, etc.). If desired, gate insulator 48 may be formed on an integrated circuit as part of a standard CMOS semiconductor fabrication process (e.g., when forming gate insulators for metal-oxide-semiconductor transistors elsewhere on the integrated circuit), thereby avoiding the need to include additional process steps (e.g., gate insulator removal steps) as part of the process of forming diodes MGLD1 and MGLD2.
At values of TOX below about 25 angstroms, the conduction mechanism in diodes MGLD1 and MGLD2 is believed to be by direct tunneling of carriers (electrons) between the n-wells of the diodes to their p-metal gates. The total current that tunnels through the gate insulator during operation of the diode includes a contribution from both the conduction band and the valence band. The structure used for diodes MGLD1 and MGLD2 resembles that of a p-metal gate varactor device having a gate insulator that is thin enough to permit quantum-mechanical tunneling of carriers and in which no current flow between the diode terminals is possible until the gate voltage on the p-metal gate is approximately equal to the flat-band voltage of the device (i.e., the turn-on voltage is approximately equal to the flat-band voltage VFB). With a metal gate, the magnitude of flat-band voltage VFB is typically smaller than that for a polysilicon gate (at 0K) and is smaller than the bandgap of silicon by about 0.3 volts. At values of TOX above about 25 angstroms, the conduction mechanism involves other mechanisms such as Fowler-Nordheim tunneling and does not generally result in a good diode-like characteristic of the type shown by curve 56 of
The foregoing is merely illustrative of the principles of this invention and various modifications can be made by those skilled in the art without departing from the scope and spirit of the invention.
Number | Name | Date | Kind |
---|---|---|---|
3975648 | Tobey et al. | Aug 1976 | A |
5229711 | Inoue | Jul 1993 | A |
5471131 | King et al. | Nov 1995 | A |
5530394 | Blossfeld et al. | Jun 1996 | A |
5610550 | Furutani | Mar 1997 | A |
5867013 | Yu | Feb 1999 | A |
6933770 | Ranucci | Aug 2005 | B1 |
7009444 | Scott | Mar 2006 | B1 |
7768343 | Sinitsky | Aug 2010 | B1 |
20020079876 | Eguchi | Jun 2002 | A1 |
20050106765 | King | May 2005 | A1 |
20050231270 | Washburn | Oct 2005 | A1 |
20060197584 | Hsu | Sep 2006 | A1 |
20070080740 | Berens | Apr 2007 | A1 |
20080042737 | Kim | Feb 2008 | A1 |
20090001958 | Fujii | Jan 2009 | A1 |
20090189590 | Ide | Jul 2009 | A1 |