IP Library Granted Patent US 12,700,471
Granted Patent B2
US 12,700,471 · App. 18/207,482 · Granted Aug 4, 2026

Fully scannable memory arrays

Inventors: Thomas A. Ziaja (Austin, TX); Paul J. Jordan (Austin, TX)
Assignee: SambaNova Systems, Inc.
G11C29/30G11C29/12015G11C29/18G11C2029/1802G11C2029/3202
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,700,471
App. No.
18/207,482
Filed
Jun 8, 2023
Granted
Aug 4, 2026
Kind
B2
Art Unit
2824
USPC
365/201
Abstract

An array of memory cells can be configured into one or more scan chains that uses non-overlapping scan word line pulses in a direction opposite to a direction of the scan chain to shift scan bits in the direction of the scan chain from a scan chain input to a scan chain output. A memory cell may include a latch and a scan input multiplexer. The array includes a pulse generator to derive a pulse generator output from a clock pulse, and a digital delay line to generate the non-overlapping scan word line pulses from the pulse generator output. The scan chain may include a latch from an input buffer and may scan multiple columns or rows.

Claims (65)

1 . An array of memory cells, comprising a scan chain that shifts scan bits in adjacent memory cells in the scan chain using non-overlapping scan word line pulses to clock successive memory cells in the scan chain, wherein a shift direction of the scan chain is along a column of memory cells.

2 . The array of memory cells of claim 1 , wherein:

the memory cell includes a single latch.

3 . The array of memory cells of claim 1 , wherein the shift direction of the scan chain is from a scan chain input to a scan chain output, and wherein an update order of the adjacent cells in the scan chain is from a last cell in the scan chain to a first cell in the scan chain.

4 . The array of memory cells of claim 3 , further comprising:

a pulse generator to derive a pulse generator output from a clock pulse; and

a digital delay line with multiple outputs, configured to shift the pulse generator output along the multiple outputs to generate the non-overlapping scan word line pulses, wherein:

the multiple outputs are coupled with clock inputs of the memory cells.

5 . The array of memory cells of claim 4 , further comprising:

an input buffer including a series of input buffer cells that are aligned in a first dimension perpendicular to the shift direction of the scan chain; and

an address decoder to decode an address of a series of memory cells that are aligned in the first dimension;

wherein:

a scan shift action of an input buffer cell is triggered by the clock;

a shift action of a memory cell included in the series of memory cells is triggered by a scan word line pulse; and

the address decoder has outputs respectively coupled with clock inputs of memory cells included in the series of memory cells.

6 . The array of memory cells of claim 5 wherein:

the address decoder is a row address decoder;

the series of memory cells that are aligned in the first direction is a row of memory cells;

the series of input buffer cells that are aligned in a first direction is a row of input buffer cells; and

the shift direction of the scan chain is along a column of memory cells.

7 . The array of memory cells of claim 6 , wherein:

the scan chain input is included in a first buffer cell in a first column;

the scan chain continues from the first buffer cell in the first column via a first memory cell in the first column to a last memory cell in the first column;

the scan chain continues from the last memory cell in the first column to a first buffer cell in a second column;

the scan chain continues from the first buffer cell in the second column via a first memory cell in the second column to a last memory cell in the second column; and

the scan chain output is included in a last buffer cell in a last column.

8 . The array of memory cells of claim 5 , further comprising:

an additional buffer including a series of additional buffer cells that are aligned in the first direction, wherein the additional buffer separates a first part of the array of memory cells from a second part of the array of memory cells, and wherein an additional buffer cell includes a latch.

9 . A method of scanning an array of memory cells arranged in N rows and C columns, the array comprising a scan chain, the method comprising:

shifting scan bits in adjacent memory cells in the scan chain using non-overlapping scan word line pulses to clock successive memory cells in the scan chain, wherein a shift direction of the scan chain is along a column of memory cells.

10 . The method of claim 9 , wherein:

the shift direction of the scan chain is from a scan chain input to a scan chain output, and wherein the adjacent memory cells receive the non-overlapping scan word line pulses in a direction reverse from the shift direction of the scan chain.

11 . The method of claim 9 , further comprising:

copying a scan input value into an input buffer of a first column;

for each column c, copying a value from a memory cell (N−1, c) into an input buffer cell of column c+1, wherein c ranges from 0 to C−2;

providing a value of memory cell (N−1, C−1) to the scan chain output;

for each column c and each row r, copying a value from cell (r−1, c) into cell (r, c), wherein c ranges from 0 to C−1, and r ranges backward from N−1 to 1; and

for each column c, copying a value from an input buffer cell of column c into memory cell (0, c), wherein c ranges from 0 to C−1.

12 . An array of memory cells arranged in N rows, numbered 0 through N−1, and C columns, numbered 0 through C−1, each row comprising a word line multiplexer, and each column comprising a first scan multiplexer, a second scan multiplexer, a read multiplexer, and a column buffer cell, wherein:

each word line multiplexer, each first scan multiplexer, and each second scan multiplexer has a first input, a second input, and a scan enable input, and is configured to couple the first input with an output when the scan enable input is de-asserted and to couple the second input with the output when the scan enable input is asserted;

the second input of the first scan multiplexer in column 0 is coupled with a scan chain input;

in each column:

the output of the first scan multiplexer is coupled with an input of the column buffer cell;

an output of the column buffer cell is coupled with an input of the memory cell in row 0 and with the first input of the second scan multiplexer;

the output of the second scan multiplexer is coupled with inputs of the memory cells in rows 1 through N−1;

outputs of memory cells in rows 0 through N−1 are coupled with inputs of the read multiplexer; and

an output of the read multiplexer is coupled with the second input of the second scan multiplexer;

in each column c wherein c ranges from 0 through N−2, the output of the memory cell (N−1, c) is coupled with the second input of the first scan multiplexer in column c+1;

the output of memory cell (N−1, C−1) is coupled with a scan chain output; and

in each row:

the output of the word line multiplexer is coupled with CLK inputs of memory cells in columns 0 through C−1.

13 . A method of scanning an array of memory cells, wherein the array comprises a scan chain including at least two columns, coupled in series, and wherein each column comprises M of the memory cells, numbered 0 through M−1, each with a memory cell enable input, a first scan multiplexer, a second scan multiplexer, a read multiplexer, and a column buffer cell, and the array comprises M cell multiplexers, each with an output coupled with one of the M memory cell enable inputs of the M memory cells of the at least two columns, the method comprising:

copying a scan chain input value to a second input of the first scan multiplexer in a first column;

asserting clock inputs of the column buffer cells;

for each column:

copying an output value of the first scan multiplexer to an input of the column buffer cell;

copying an output value of the column buffer cell to an input of memory cell 0 and to a first input of the second scan multiplexer;

copying an output value of the read multiplexer to a second input of the second scan multiplexer;

copying an output value of the second scan multiplexer to inputs of memory cells m, wherein m ranges from 1 to M−1;

copying output values of memory cells m to inputs of the read multiplexer, wherein m ranges from 0 to M−1; and

for the first column copying an output value of memory cell M−1 to a second input of the first scan multiplexer in a second column;

for a last column, copying an output value of memory cell M−1 to a scan chain output value;

using a scan enable signal to select second inputs of the first scan multiplexer, the second scan multiplexer, and m cell multiplexers, wherein m ranges from 0 to M−1;

for each cell multiplexer m, wherein m ranges backward from M−1 to 0, applying a pulse to a second input of the cell multiplexer to enable memory cells m to copy a memory cell input value to a memory cell output; and

de-asserting the clock of the column buffer cells to enable the column buffers cells to copy their input values to their output.

Assignments (2)
INTELLECTUAL PROPERTY SECURITY AGREEMENT Recorded Apr 18, 2025
From: SAMBANOVA SYSTEMS, INC.
To: SILICON VALLEY BANK, A DIVISION OF FIRST-CITIZENS BANK & TRUST COMPANY, AS AGENT
Reel/Frame 070892/0001 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 8, 2023
From: ZIAJA, THOMAS A.; JORDAN, PAUL J.
To: SAMBANOVA SYSTEMS, INC.
Reel/Frame 063899/0306 →
Continuity (1)
Related Publication 20240412797A1 · Dec 12, 2024
References Cited (6)
US 5550843A · Yee · 1996 [cited by examiner]
US 10162005B1 · Alzheimer · 2018 [cited by examiner]
US 20170115342A1 · Wang · 2017 [cited by examiner]
WO 2010142987A1 · 2010 [cited by applicant]
M. Emani et al., Accelerating Scientific Applications With Sambanova Reconfigurable Dataflow Architecture, in Computing in Science & Engineering, vol. 23, No. 2, pp. 114-119, Mar. 26, 2021, [doi: 10.1109/MCSE.2021.30572… [cited by applicant]
Podobas et al, A Survey on Coarse-Grained Reconfigurable Architectures From a Performance Perspective, IEEEAccess, vol. 2020.3012084, Jul. 27, 2020, 25 pages. [cited by applicant]