The technical field of this invention is power supply control in integrated circuits.
The current surge during power up or power down of a power domain on a large integrated circuit can cause functional failures to the device and/or cause reliability issues in the transistors. The prior art generally considered this a problem only during power up. Power supply instability is an issue during power down. When the current drawn suddenly drops oscillations in the supply may occur producing undershoots and overshoots. This behavior is observed in prior designs and is expected to get worse with future larger designs.
This invention makes the change in current drawn from the power grid gradual by sequencing the power switch chains differently for both power up and power down. During power up, this invention establishes a reasonable connection with the power grid through a series of weak power switches and then starts turning on the strong power switches. Thus the current load of a power domain is ramped up. During power down, this invention reverses the process. Strong switches are all turned off before turning off the weak switches.
The prior art comprehends only the power up case but this invention handles both power up and power down scenarios.
These and other aspects of this invention are illustrated in the drawings, in which:
129 is supplied with an individual signal from power supply controller 142. This invention controls the sequence of power switch activations.
The transistors 210, 220, 230 . . . 290 are sequentially energized via an inverter chain. Drive signal 201 from a corresponding output of power supply controller 142 is input to inverter 211. The output of inverter 211 is connected to the gate of transistor 210 and to the input of inverter 212. The output of inverter 212 is connected to the input of inverter 221. The output of inverter 221 is connected to the gate of transistor 220 and to the input of inverter 222. The output of inverter 222 is connected to the input of inverter 231. The output of inverter 231 is connected to the gate of transistor 230 and to the input of inverter 232. The output of inverter 212 is connected to the input of a next inverter. This inverter chain continues to inverter 291. The output of inverter 291 is connected to the gate of transistor 290.
An input from drive signal 201 causes inverter 211 to switch transistor 210 ON. Inverter 211 also switches inverter 212. This input caused inverter 212 to switch inverter 221. Inverter 221 to switch transistor 220 ON. Each inverter in the chain causes a delay from its input before its output switches. This causes a propagation delay before the next transistor switches ON. Thus switches 210, 220, 230 . . . 290 switch ON sequentially as the input travels the inverter chain. The delay of each inverter in the chain depends upon the size of the transistors used in the inverter (bigger transistors switch faster) and the load on the output. Larger transistors 210, 220, 230 . . . 290 have larger gate capacitance requiring the corresponding driver to move more charge to turn the transistor ON. Thus larger transistors 210, 220, 230 . . . 290 cause the inverter chain to propagate slower than smaller transistors. Thus transistors 210, 220, 230 . . . 290 turn ON sequentially. When turning OFF a similar delay occurs in the inverter chain causing a corresponding sequential action in turning OFF transistors 210, 220, 230 . . . 290. This causes transistors 210, 220, 230 . . . 290 to turn OFF sequentially.
In accordance with this invention the power switches 121, 122, 123 . . . 129 are not identical. Instead, power switches 121, 122, 123 . . . 129 are constructed from a variety of strengths. A strong switch can carry a large current and produces a small IR voltage drop. A weak switch carries a smaller current and provides a larger IR voltage drop. The strength of these switches is controlled by the width of the source-drain channel of the corresponding transistors 211, 212, 213 . . . 219. In general a wider source-drain channel produces a stronger transistor than a narrow source-drain channel.
This example embodiment shows p-channel metal oxide semiconductor (PMOS) transistors controlling conduction of the voltage supply (Vdd) to the power domain. Those skilled in the art would realize this invention could be practiced using n-channel metal oxide semiconductor (NMOS) transistors to control conduction of ground (Vss) to the power domain. Such a change would require inversion of the drive voltages (
In accordance with the preferred embodiment of this invention power switches 121, 122, 123 . . . 129 are arranged in strength from weakest transistor to strongest transistor. During power up, weak switches are turned ON first, followed by strong switches. During power down, strong switches are turned OFF first, followed by weak switches.
State machine 500 remains in ON state 520 until receipt of an external OFF command. On receipt of the OFF command state machine 500 advances toward state 510 periodically turning OFF power switches starting with the strongest. State machine 500 advances to state 527. In state 527 power switches 121, 122 and 123 are ON and all stronger power switches are OFF. State machine 500 advances to state 528 upon reaching time Dt3 following the OFF command. In state 528 power switches 121 and 122 are ON and power switches 123 . . . 129 are OFF. State machine 500 advances to state 529 upon reaching time Dt2 following the OFF command. In state 529 power switch 121 is ON and power switches 122, 123 . . . 129 are OFF. State machine 500 advances to state 510 upon reaching time Dt2 following the OFF command. As described above, in state 510 all power switches 121, 122, 123 . . . 129 are OFF.
State machine 500 may be implemented by special purpose hardware or by a suitably programmed microcomputer. If the integrated circuit including power control system 100 includes a programmable central processing unit, some portion of the computing capacity could be devoted to this power supply control.
This invention provides controlled shutdown of power switch chains and hence avoids functional failure or damage to the transistors. Controlling the power down of a power domain, ensures proper functionality of other power domains that share the power grid. This permits proper functioning of dynamic power management. The main differentiation of this invention over the prior art is making the transient change in current consumption slowly not only during power-up but also during power down.
This application claims priority under 35 U.S.C. 119(e)(1) to U.S. Provisional Application No. 61/859,378 filed Jul. 29, 2013.
Number | Date | Country | |
---|---|---|---|
61859378 | Jul 2013 | US |