Charge domain mathematical engine and method
A multiplier has a pair of charge reservoirs. The pair of charge reservoirs are connected in series. A first charge movement device induces charge movement to or from the pair of charge reservoirs at a same rate. A second charge movement device induces charge movement to or from one of the pair of reservoirs, the rate of charge movement programmed to one of add or remove charges at a rate proportional to the first charge movement device. The first charge movement device loads a first charge into a first of the pair of charge reservoirs during a first cycle. The first charge movement device and the second charge movement device remove charges at a proportional rate from the pair of charge reservoirs during a second cycle until the first of the pair of charge reservoirs is depleted of the first charge. The second charge reservoir thereafter holding the multiplied result.
1 . A neural network comprising:
an analog multiplier comprising:
a pair of charge reservoirs, wherein the pair of charge reservoirs are connected in series;
a first charge movement device inducing charge movement to or from the pair of charge reservoirs at a same rate;
a second charge movement device inducing charge movement to or from one of the pair of reservoirs, the rate of charge movement programmed to one of add or remove charges at a rate proportional to the first charge movement device;
wherein a first charge into a first of the pair of charge reservoirs during a first cycle, the first charge movement device and the second charge movement device removing charges at a proportional rate from the pair of charge reservoirs during a second cycle until the first of the pair of charge reservoirs is depleted of the first charge; and
an input gathering device to store charge in the first of the pair of reservoirs in conformance with input information.
2 . The neural network in accordance with claim 1 , wherein the analog multiplier comprises a device to stop charge movement.
3 . The neural network in accordance with claim 1 , comprising an input photodiode formed on a same integrated circuit as the neural network.
4 . The neural network of claim 1 , wherein the input gathering device is a charge domain circuit and the input information is optical information.
5 . The neural network of claim 1 , comprising at least one Charge Coupled Device (CCD) shift register coupled to the input gathering device, the CCD shift register used as the first of the pair of charge reservoirs.
6 . The neural network of claim 1 , comprising at least one Charge Coupled Device (CCD) shift register coupled between the input gathering device and the first of the pair of charge reservoirs.
7 . The neural network of claim 6 , wherein the CCD shift register is a two dimensional shift register.
8 . The neural network of claim 6 , wherein the CCD shift register is a two dimensional coupled array capable of accepting information at cells around its periphery.
9 . The neural network of claim 6 , comprising a time weighted crossbar used to broadcast an input operand to the first of the pair of charge reservoirs, the at least one CCD shift register being used to couple a charge according to a systolic response.
10 . The neural network of claim 9 , where a second operand is, stored as charge in a second CCD shift register, wherein a plurality of CCD shift registers pass respective operands per a systolic algorithm.