IP Library › Granted Patent US 12,530,570
Granted Patent B2
US 12,530,570 · App. 17/882,341 · Granted Jan 20, 2026

Methods and systems of operating a neural circuit in a non-volatile memory based neural-array

Inventor: Vishal Sarin (Santa Clara, CA)
G06N3/063G06F7/49942G06F17/16G06N3/04G06N3/049G06N3/065G06N3/082
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 12,530,570
App. No.
17/882,341
Granted
Jan 20, 2026
Kind
B2
Abstract

In one aspect, a method of a neuron circuit includes the step of providing a plurality of 2 N −1 single-level-cell (SLC) flash cells for each synapse (Y i ) connected to a bit line forming a neuron. The method includes the step of providing an input vector (X i ) for each synapse Y i wherein each input vector is translated into an equivalent electrical signal ES i (current I DACi , pulse T PULSEi , etc). The method includes the step of providing an input current to each synapse sub-circuit varying from 2 0 *ES i to (2 N −1)*ES i . The method includes the step of providing a set of weight vectors or synapse (Y i ), wherein each weight vector is translated into an equivalent threshold voltage level or resistance level to be stored in one of many non-volatile memory cells assigned to each synapse (Y i ). The method includes the step of providing for 2 N possible threshold voltage levels or resistance levels in the 2 N −1 non-volatile memory cells of each synapse, wherein each cell is configured to store one of the two possible threshold voltage levels. The method includes the step of converting the N digital bits of the weight vector or synapse Y i into equivalent threshold voltage level and store the appropriate cell corresponding to that threshold voltage level in one of the many SLC cells assigned to the weight vector or synapse (Y i ). The method includes the step of turning off all remaining 2 N −1 flash cells of the respective synapse (Y i ). Various other methods are presented of forming neuron circuits by providing a plurality of single-level-cell (SLC) and many-level-cell (MLC) non-volatile memory cells, for each synapse (Y i ) electrically connected to form a neuron. The disclosure shows methods of forming neurons in various configurations for non-volatile memory cells (flash, RRAM etc.); of different storage capabilities per cell—both SLC and MLC cells.

Claims (13)

1 . A method of operating a neuron circuit in a non-volatile memory-based neural array, the method comprising:

providing a plurality of N many-level-cell (MLC) non-volatile memory cells for each synapse (Yi) connected to a bit line forming a neuron, wherein the plurality of non-volatile memory cells are ordered from a most significant levels (MSLs) to a least significant levels (LSLs), and wherein N is the number of bits in the weight vector of synapse (Yi);

providing an input vector (Xi) for each synapse (Yi), wherein each input vector (Xi) is translated into an equivalent electrical signal ESi;

providing an electrical signal to each synapse sub-circuit, the electrical signal comprising a set of input currents ((2 {circumflex over ( )}{0·k})·IDACi, (2{circumflex over ( )}{1·k})·IDACi, (2{circumflex over ( )}{2·k})·IDACi, . . . ), where k=log 2 (L);

providing a set of weight vectors, wherein each weight vector is translated into an equivalent threshold voltage level or resistance level to be stored in one of many MLC non-volatile memory cells assigned for each synapse;

providing for 2{circumflex over ( )}N possible threshold voltage levels or resistance levels in the (2{circumflex over ( )}N−1)/(L−1) non-volatile memory cells of each synapse, wherein each non-volatile memory cell is configured to store L possible threshold voltage levels or resistance levels, one of which is equivalent to an off state (0 state); and

converting the N digital bits of the weight vector into equivalent threshold voltage level or resistance levels and store the appropriate cell corresponding to that level in one of a plurality of MLC non-volatile memory cells assigned to the weight vector or synapse (Yi).

2 . The method of claim 1 , wherein weight level (W) is stored in sequence from least significant levels non-volatile memory cell to most significant levels non-volatile memory cell in a sequential manner such as if weight level (W) falls within level 0 to L−1, then store W in the least significant levels non-volatile memory cell and turn off the other remaining non-volatile memory cells forming the synapse by storing level 0, which is equivalent to an off cell in the remaining non-volatile memory cells; if weight level (W) falls within level L to 2 (L−1), then store L−1 in the 1st least significant levels non-volatile memory cell, and W-(L−1) in the 2nd least significant levels non-volatile memory cell and turn off the remaining non-volatile memory cells forming the synapse by storing level 0; and to implement the process sequentially for various possible values of weight levels (W) per synapse.

3 . The method of claim 1 , wherein the step of providing an electrical signal to each synapse sub-circuit further comprises:

providing an input current (IDACi) to each synapse sub-circuit.

4 . The method of claim 1 , wherein the step of providing an electrical signal to each synapse sub-circuit further comprises:

providing an input pulse (TPULSEi) to each synapse sub-circuit.

5 . A method of claim 1 , wherein the non-volatile memory cell is a flash memory cell.

Assignments (3)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jan 7, 2026
From: ANALOG INFERENCE INC.
To: SAGENCE AI CORPORATION
Reel/Frame 073397/0464 →
SECURITY INTEREST Recorded Dec 8, 2025
From: SAGENCE AI CORPORATION
To: CUSTOMERS BANK
Reel/Frame 073136/0296 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Apr 28, 2023
From: SARIN, VISHAL
To: ANALOG INFERENCE INC.
Reel/Frame 063488/0320 →
Continuity (6)
Division 16452308 · Jun 25, 2019
Provisional Application 62803562 · Feb 10, 2019
Provisional Application 62773773 · Nov 30, 2018
Provisional Application 62721116 · Aug 22, 2018
Provisional Application 62689839 · Jun 26, 2018
Related Publication 20230082240A1 · Mar 16, 2023
References Cited (9)
US 20070045711A1 · Bhattacharyya · 2007 [cited by examiner]
US 20140059406A1 · Hyun · 2014 [cited by examiner]
US 20140380129A1 · Shin · 2014 [cited by examiner]
US 20160148701A1 · Karakulak · 2016 [cited by examiner]
US 20170269669A1 · Choi · 2017 [cited by examiner]
US 20180314927A1 · Muralimanohar · 2018 [cited by examiner]
US 20190287631A1 · Tran · 2019 [cited by examiner]
Donato et al., “On-chip Deep Neural Network Storage with Multi-Level eNVM” (Jun. 24, 2018) (Year: 2018). [cited by examiner]
Renesas Application Note, “Basic DACs for Electronic Engineers” (Jul. 1997) (Year: 1997). [cited by examiner]