SRAM row redundancy

Details SRAM row redundancy SRAM row redundancy

2001-03-05

2004-02-17

Iqbal, Nadeem (Department: 2184)

Error detection/correction and fault detection/recovery

Data processing system error or fault handling

Reliability and availability

C365S201000

Reexamination Certificate

active

06694448

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to static random access memories (SRAMs) and, more particularly, to techniques for replacement of a defective row in a SRAM with a redundant row.
2. Prior Art
If one or more memory cells in a row of a SRAM array are defective, the entire row is considered to be defective and must be replaced with a redundant row in the array. The prior art uses two techniques for replacement of a defective row in a SRAM. One such technique is called a “shift” method and the other technique is called a “brute force” method.
FIG. 1
illustrates a typical global word line (GWL) control circuit
100
that is used to implement the prior art “shift” method that provides one type of row redundancy for the word lines in a SRAM memory. For an exemplary SRAM with a total of 2048 word lines, four word lines (WLs) are grouped into one global word line (GWL) so that 512 GWLs are used for the SRAM with 2048 word lines. Each GWL then uses a 4:1 post-decode circuit to access each of the four WLs in a particular GWL. This shift technique uses a separate GWL control circuit
100
for each of the 512 GWLs. Each control circuit
100
uses one GWL fuse element
102
for each GWL, so that a total of 512 fuses are used for the 512 GWLs and the corresponding 2048 WLs.
The 512 separate GWL control circuits are arranged in a cascade arrangement of separate GWL control circuits, one for each GWL. When a fuse
102
for a particular GWL control circuit is blown, each control circuit beyond the defective row shifts all of the memory cells up one row in the SRAM memory array. All of the addressing accommodates this shifted arrangement. Each control circuit
100
has a FH terminal that is connected to a corresponding FL terminal of the next higher GWL control circuit. Each control circuit
100
also has a XDH terminal that is connected to a corresponding XDL terminal of the next higher GWL control circuit. For the bottommost GWL control circuit, XDEC, FL and XDL are connected to VSS, or ground. Predecode address signals A
76
, A
54
, A
32
, and A
10
are all connected to VSS for the top GWL control circuit. Postdecode address signals XR[3:0] are provided to select one of four word lines WL associated with a GWL. Each of the output signals WL[3:0] activates one of the four WLs associated with a particular GWL.
There are several criteria for determining that a row in the SRAM is a defective row. A row is determined to be defective because it has a single bad bit, more than one bad bit, or the entire row is defective for some reason. An associated GWL fuse element
102
is blown for the defective row. This disables the GWL control circuit
100
and shifts the GWL control operation physically to a next GWL control circuit above. This shift continues up until the actual control circuit for the redundant GWL row is reached at the top of the rows of GWL control circuits.
An advantage of the shift method is that only one fuse needs to be to blown for the GWL control circuit of a defective memory row. However, an important disadvantage of the shift method is that it requires a fuse to be available for each GWL control circuit , so that, for example, 512 virgin fuses are required for the 512 GWL control circuits, even though only one fuse is to be blown. A further disadvantage is that this type of system is not flexible because only one defective memory row out of 512 memory rows can be replaced with this method.
For the shift technique, the decoder redundancy could be further subdivided by dividing the memory into several smaller memory blocks with a redundant GWL control circuit and a redundant memory row for each smaller memory block. For a memory with 512 GWL control circuits that is divided into four memory blocks, one redundant GWL control circuit is provided for every 128 GWLs. This still limits this technique to one redundant GWL for 128 GWLs. If the 128 GWL group had two bad GWLs, this technique would not work even though there may be three other unused redundant GWL control circuits and redundant memory rows provided for the other three 128-row groups.
FIG. 2
illustrates another redundancy control circuit
200
for providing row redundancy for a defective SRAM row. This circuit uses a so-called “brute force” technique because it disables all of the regular GWLs and activates a redundant GWL whenever an address for a defective row is detected. The input signals to the redundancy control circuit
200
are nine bits [A0:A8] of a nine-bit address signal. The output control signals of the redundancy control circuit
200
are a redundant GWL control signal REGWL and a RFLAG signal that turns off the regular GWLs. When the RFLAG signal goes to 0, it turns off all the regular GWLs and brings up the redundant GWL control signal REGWL. The defective redundancy control circuit
200
includes a fuse enable FENABLE circuit
202
and 9 fuse decoder circuits FRXDECFUSE circuits
204
a
-
204
i.
FIG. 3
illustrates the FENABLE circuit
202
. When it is not blown open, a fuse element
206
provides a VCC voltage to the drain terminal of a pull-down NMOS transistor
208
and to an input terminal of a first inverter
210
. The Gate terminal of the pull-down NMOS transistor
208
is connected to the output terminal of the first inverter
210
. The output terminal of the inverter
210
is connected to an input terminal of a second inverter
212
. The output terminal of the second inverter
212
is connected to a NORMH output terminal
214
and to an input terminal of a third inverter
216
, the output terminal of which is connected to a NORML terminal
218
. When the fuse element
206
is blown open, the drain terminal of the pull-down transistor
208
is at a low voltage, which is inverted through the first inverter
210
to turn on the pull-down transistor
208
. This provides a high signal at the NORMH terminal
214
and a low signal at the NORML terminal
218
.
FIG. 4
illustrates a typical FRXDECFUSE circuit
204
that includes a transfer gate
250
that is controlled by the NORMH and NORML signals at terminals
252
and
254
, respectfully. An AIB input terminal
256
to the transfer gate
250
is connected to the input terminal of the transfer gate
250
and to an input terminal of an inverter
258
. An output terminal of the transfer gate
250
is connected through a fuse element
260
to an RAIB output terminal. An output terminal of the inverter
258
is connected through another fuse element
264
to the RAIB output terminal. When the RFLAG signal goes to 0, it turns off all the regular GWLs and brings up the redundant GWL signal REGWL.
An advantage of the brute force method is that it is very flexible. If setup for 4 redundant GWLs per 512 regular GWLs, you are not limited to just one redundant GWL per 128 GWL as in the shift method. This method could accommodate bad memory rows associated with 4 GWLs in a 128 GWL group. This method also requires less total virgin fuses because 19 fuses are required per redundant GWL. A disadvantages of this method is that 10 fuses need to be blown per redundant GWL. The requirement of 19 virgin fuses for each of the four redundant GWL is a disadvantage.
Consequently, a need exists for an SRAM with a more cost-effective, efficient technique for controlling a redundant row substituted for a defective row.
SUMMARY OF THE INVENTION
It is therefore an object of the invention to provide a fuse-controlled redundancy control system for a SRAM that, overall, uses fewer virgin fuses, which equates to less fuse area that provides a smaller chip size and a lower chip cost.
It is another object of the invention to provide a fuse-controlled redundancy control system for a SRAM that has fewer fuses to blow, which results in less time required for a laser to blow fuses and which proves for lower chip cost.
It is another object of the invention to provide a fuse-controlled redundancy control system for a SRAM that provides flexibility in designating redundant rows, which results in more die area saved and a l

Affiliated with

Also associated with

Rating

SRAM row redundancy does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with SRAM row redundancy, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and SRAM row redundancy will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3351657