Apparatus and method with multiply-accumulate operation
A multiply-accumulate (MAC) computation circuit includes: a bit-cell array configured to generate an analog output corresponding to a MAC operation result of an input signal; a first analog-to-digital conversion (ADC) circuit configured to determine an upper part of a digital output corresponding to the analog output; and a second ADC circuit configured to determine a lower part of the digital output based on a reference voltage corresponding to the upper part.
1 . A multiply-accumulate (MAC) computation circuit comprising:
a bit-cell array configured to generate an analog output corresponding to a MAC operation result of an input signal;
a capacitor array configured to convert a current value of the analog output into a voltage value;
a first analog-to-digital conversion (ADC) circuit configured to receive the voltage value of the analog output and determine an upper part of a digital output corresponding to the analog output; and
a second ADC circuit configured to receive the voltage value of the analog output and determine a lower part of the digital output based on a reference voltage corresponding to the upper part,
wherein the first ADC circuit is a successive approximation register (SAR) ADC circuit, and the second ADC circuit is a flash ADC circuit.
2 . The MAC computation circuit of claim 1 , wherein
for the determining of the upper part, the first ADC circuit is configured to compare the analog output to coarse ranges according to primary reference voltages and determine a first range to which the analog output belongs among the coarse ranges, and
the first range corresponds to the upper part of the digital output.
3 . The MAC computation circuit of claim 2 , wherein
for the determining of the lower part, the second ADC circuit is configured to compare the analog output to fine ranges of the first range according to secondary reference voltages including the reference voltage and determine a second range to which the analog output belongs among the fine ranges, and
the second range corresponds to the lower part of the digital output.
4 . The MAC computation circuit of claim 2 , wherein the primary reference voltages are set based on the analog output.
5 . The MAC computation circuit of claim 1 , wherein
the digital output is 6 bits wide,
the upper part is 2 bits wide, and
the lower part is 4 bits wide.
6 . The MAC computation circuit of claim 1 , wherein
the bit-cell array comprises a plurality of input lines and a plurality of output lines, and
a first analog output of a first output line among the plurality of output lines corresponds to a result of a first MAC operation performed between voltage values applied to the first output line by the input signal and resistance values of bit-cells belonging to the first output line.
7 . The MAC computation circuit of claim 1 , wherein
the bit-cell array comprises a plurality of bit-cell groups that each comprise a plurality of output lines, and
first output lines belonging to a first bit-cell group among the plurality of bit-cell groups share a first ADC block and a first capacitor array assigned to the first bit-cell group.
8 . The MAC computation circuit of claim 7 , wherein the first capacitor array comprises:
a first capacitor configured to sample outputs of output lines of a first sub-group among the first output lines; and
a second capacitor configured to sample outputs of output lines of a second sub-group among the first output lines.
9 . The MAC computation circuit of claim 8 , wherein
the first ADC block comprises the first ADC circuit and the second ADC circuit, and
the first ADC circuit and the second ADC circuit operate based on a pipeline.
10 . The MAC computation circuit of claim 9 , wherein
the first ADC circuit is configured to determine a digital upper part of an output of an output line of the first sub-group for a first time interval,
the second ADC circuit is configured to determine a digital lower part of the output of the output line of the first sub-group for a second time interval,
the first ADC circuit is configured to determine a digital upper part of an output of an output line of the second sub-group for a third time interval, and
at least a portion of the third time interval overlaps the second time interval according to the pipeline.
11 . The MAC computation circuit of claim 8 , wherein an even-numbered output line among the first output lines belongs to the first sub-group, and an odd-numbered output line among the first output lines belongs to the second sub-group.
12 . The MAC computation circuit of claim 1 , further comprising a plurality of capacitor arrays comprising the capacitor array and configured to sample the analog output, a plurality of first ADC circuits comprising the first ADC circuit, and a plurality of second ADC circuits comprising the second ADC circuit.
13 . An electronic apparatus comprising the MAC computation circuit of claim 1 and a processor configured to generate a recognition result of an input corresponding to the input signal, based on the digital output.
14 . A multiply-accumulate (MAC) computation circuit comprising:
a bit-cell array configured to generate an analog output corresponding to a MAC operation result of an input signal, using a plurality of bit-cell groups, each comprising a plurality of output lines;
a plurality of capacitor arrays configured to sample the analog output by converting a current value of the analog output into a voltage value, the plurality of capacitor arrays being assigned to each of the plurality of bit-cell groups;
a plurality of first analog-to-digital conversion (ADC) circuits configured to receive the voltage value of the analog output and determine an upper part of a digital output corresponding to the analog output, the plurality of first ADC circuits being connected to the plurality of capacitor arrays; and
a plurality of second ADC circuits configured to receive the voltage value of the analog output and determine a lower part of the digital output based on a reference voltage corresponding to the upper part,
wherein the plurality of first ADC circuits are successive approximation register (SAR) ADC circuits, and the plurality of second ADC circuits are flash ADC circuits.
15 . The MAC computation circuit of claim 14 , further comprising:
a reference generator configured to generate reference voltage candidates; and
a reference selector configured to select the reference voltage corresponding to the upper part of the digital output among the reference voltage candidates.
16 . The MAC computation circuit of claim 15 , wherein, for the generating of the reference voltage candidates, the reference generator is configured to determine at least a portion of the reference voltage candidates based on the analog output.
17 . The MAC computation circuit of claim 14 , wherein first output lines belonging to a first bit-cell group among the plurality of bit-cell groups share a first capacitor array, a first ADC circuit, and a second ADC circuit that are assigned to the first bit-cell group among the plurality of capacitor arrays, the plurality of first ADC circuits, and the plurality of second ADC circuits.
18 . The MAC computation circuit of claim 17 , wherein
the first capacitor array comprises:
a first capacitor configured to sample outputs of output lines of a first sub-group among the first output lines; and
a second capacitor configured to sample outputs of output lines of a second sub-group among the first output lines, and
the first ADC circuit and the second ADC circuit operate based on a pipeline.
19 . The MAC computation circuit of claim 14 , wherein
for the determining of the upper part, the first ADC circuits are configured to compare the analog output to coarse ranges according to primary reference voltages and determine a first range to which the analog output belongs among the coarse ranges, and
for the determining of the lower part, the second ADC circuits are configured to compare the analog output to fine ranges of the first range according to secondary reference voltages including the reference voltage and determine a second range to which the analog output belongs among the fine ranges.
20 . A multiply-accumulate (MAC) computation method comprising:
generating an analog output corresponding to a MAC operation result of an input signal, using a bit-cell array;
converting a current value of the analog output into a voltage value, using a capacitor array;
receiving the voltage value of the analog output and determining an upper part of a digital output corresponding to the analog output, using a first analog-to-digital conversion (ADC) circuit;
setting a reference voltage of a second ADC circuit based on the upper part; and
receiving the voltage value of the analog output and determining a lower part of the digital output, using the second ADC circuit,
wherein the first ADC circuit is a successive approximation register (SAR) ADC circuit, and the second ADC circuit is a flash ADC circuit.
21 . An electronic apparatus comprising:
a bit-cell array configured to generate an analog output corresponding to a multiply-accumulate (MAC) operation result of an input signal;
a capacitor array configured to convert a current value of the analog output into a voltage value;
a first analog-to-digital conversion (ADC) circuit configured to receive the voltage value of the analog output and determine high-order bits of a digital output corresponding to the analog output by comparing the analog output to one or more primary reference values determined based on a maximum value of an input signal value range; and
a second ADC circuit configured to receive the voltage value of the analog output and determine low-order bits of the digital output by comparing the analog output to one or more secondary reference values determined based on the primary reference values,
wherein the first ADC circuit is a successive approximation register (SAR) ADC circuit, and the second ADC circuit is a flash ADC circuit.
22 . The apparatus of claim 21 , wherein, for the determining of the high-order bits, the first ADC circuit is configured to determine a highest-order bit of the high-order bits by comparing the analog output to a primary reference value of the one or more primary reference values that is half of the maximum value of the input signal value range.
23 . The apparatus of claim 21 , wherein a value of the analog output and the secondary reference values are within a range of a first primary reference value and a second primary reference value of the one or more primary reference values.
24 . The apparatus of claim 21 , further comprising a processor configured to generate a recognition result of an input corresponding to the input signal, based on the digital output.