Integer matrix multiplication based on mixed signal circuits
US-2022075596-A1 · Mar 10, 2022 · US
US11983507B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-11983507-B2 |
| Application number | US-202017139955-A |
| Country | US |
| Kind code | B2 |
| Filing date | Dec 31, 2020 |
| Priority date | Dec 31, 2020 |
| Publication date | May 14, 2024 |
| Grant date | May 14, 2024 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
A differential multiplier-accumulator accepts A and B digital inputs plus a sign bit and generates a dot product P by applying the bits of the A input and the bits of the B inputs to respective positive and negative unit elements comprised of groups of AND gates coupled to charge transfer lines through a capacitor Cu. One of the positive and negative unit element is enabled by the sign bit, the enabled unit element receives one bit of the B input applied to all of the AND gates of the unit element, and each positive and negative unit element having the bits of A applied to each associated AND gate input of each unit element, which charge to charge transfer lines, and the charge transfer lines are coupled to binary weighted charge summing capacitors and to an analog to digital converter to generate a digital output product.
Opening claim text (preview).
We claim: 1. A differential multiplier-accumulator operative to multiply an A digital input having a number of bits with a B digital input having a number of bits accompanied by a sign bit, the differential multiplier-accumulator comprising: a positive value unit element part enabled when the sign bit is positive and disabled when the sign bit is negative; a negative value unit element part enabled when the sign bit is negative and disabled when the sign bit is positive; a differential analog to digital converter (ADC) having a positive input coupled to an output of the positive value unit element part and a negative input coupled to an output of the negative value unit element part, the ADC forming an output; the positive value unit element part comprising: a plurality of AND-groups equal in number to the number of bits of the A digital input, each AND-group further comprising a plurality of AND gates equal in number to the number of bits of the A digital input, each AND gate of each AND-group receiving a unique one of the bits of the A digital input and one of the bits of the B digital input, the AND gate of each AND-group coupled to a particular charge transfer line through a charge transfer capacitor having a value C; a charge summing unit comprising a plurality of charge summing capacitors, each charge summing capacitor having a first terminal coupled to a particular one of the charge transfer lines, each charge summing capacitor having a second terminal coupled to the positive input of the ADC; the negative value unit element part comprising: a plurality of AND-groups equal in number to the number of bits of the A digital input, each AND-group further comprising a plurality of AND gates equal in number to the number of bits of the A digital input, each AND gate of each AND-group receiving a unique one of the bits of the A digital input and one of the bits of the B digital input, the AND gate of each AND-group coupled to a particular charge transfer line through a charge transfer capacitor having a value C; a charge summing unit comprising a plurality of charge summing capacitors, each charge summing capacitor having a first terminal coupled to a particular one of the charge transfer lines, each charge summing capacitor having a second terminal coupled to the negative input of the ADC. 2. The differential multiplier-accumulator of claim 1 where an analog charge line weight number for each of the charge transfer lines is determined by a sum of a bit number of the A digital input and a bit number of the B digital input associated with a particular AND gate, an analog charge line weight being a power of two to the analog charge line weight number for a respective charge transfer line. 3. The differential multiplier-accumulator of claim 2 where the charge transfer capacitors coupled to the AND gate have the same value and the analog charge line weight, the analog charge line weight determined by a size of a charge line capacitor associated with an analog charge line. 4. The differential multiplier-accumulator of claim 2 where analog charge lines with the same analog charge line weight number, wherein each analog charge line is a single analog charge line. 5. The differential multiplier-accumulator of claim 2 where analog charge lines with the same analog charge line weight number are separate analog charge lines. 6. The differential multiplier-accumulator of claim 1 where the A digital input or the B digital input has three bits or four bits. 7. The differential multiplier-accumulator of claim 1 where the sum of capacitances of charge transfer capacitors, corresponding to the positive and negative value unit element parts, coupled to a particular analog charge line has a value which is at least eight times greater than a value of an associated charge summing capacitor coupled to the particular analog charge line. 8. The differential multiplier-accumulator of claim 1 where each charge summing capacitor of the positive and negative value unit element parts has a value Cs*2 n where n is a weighting bit order of an associated analog charge line and for each positive and negative unit element part, the total capacitance of charge transfer capacitors coupled to a particular analog charge line is at least eight times larger than Cs. 9. A differential unit element (UE) for multiplying an A digital value having a number of bits and a B digital value having a number of bits accompanied by a sign bit and transferring a result to an analog charge bus, the differential unit element comprising: a positive analog charge bus comprising a plurality of positive analog charge lines, the positive analog charge lines coupled to a positive charge summing unit having a positive output; a negative analog charge bus comprising a plurality of negative analog charge lines, the negative analog charge lines coupled to a negative charge summing unit having a negative output; an analog to digital converter (ADC) having a positive input coupled to the positive output and a negative input coupled to the negative output, the ADC generating an output equal to the negative input subtracted from the positive input; a positive unit element part enabled when the sign bit is positive and disabled when the sign bit is negative, and a negative unit element part enabled when the sign bit is negative and disabled when the sign bit is positive; the positive unit element part comprising a plurality of AND-groups, each AND-group further comprising a plurality of AND gates having one input coupled to a unique one of the bits of the A digital value and an input coupled to one of the bits of the B digital value, each AND gate coupled through a charge transfer capacitor to a positive analog charge line; the negative unit element part comprising a plurality of AND-groups, each AND-group further comprising a plurality of AND gates having one input coupled to a unique one of the bits of the A digital value and an input coupled to one of the bits of the B digital value, each AND gate coupled through a charge transfer capacitor to a negative analog charge line; the positive analog charge lines and negative analog charge lines being coupled to a binary weighted positive charge summing unit and a binary weighted negative charge summing unit, respectively, each of the binary weighted positive charge summing unit and the binary weighted negative charge summing unit comprising a plurality of binary weighted capacitors according to a bit order of the analog charge line such that a charge transfer from each increased bit number of the sum of a bit number of the A digital value and a bit number of the B digital value is a factor of two greater than a charge transfer from a respective positive or negative analog charge line with a respective bit number lower by one. 10. The differential UE of claim 9 where each binary weighted charge summing capacitor associated with a charge transfer line with bit order n has an associated capacitance value Cs*2 n . 11. The differential UE of claim 10 where a capacitance equal to the sum of charge transfer capacitors coupled to a charge transfer line is at least eight times greater than a capacitance of an associated charge summing capacitor. 12. The differential UE of claim 9 where the negative and positive analog charge lines include capacitors for introducing a bias charge. 13. The differential UE of claim 9 where the negative and positive analog charge lines include switches for discharging analog transfer capacitors and charge summing capacitors.
Sum of products (for applications thereof, see the relevant places, e.g. G06F17/10, H03H17/00) · CPC title
for multiplication or division {(G06G7/19 and G06G7/24 take precedence; measuring electric power G01R21/00)} · CPC title
Neural networks · CPC title
Non-logic devices, e.g. operational amplifiers · CPC title
Related publications grouped by family.
Answers are generated from the same data shown on this page.