Patent Application
20210073483
This patent application directs itself to physical analog computers. By this we mean computers that carry out computations by means of analog electrical circuitry that manipulates analog electrical signals, typically for the purpose of solving differential equations. Importantly, such computers are better suited in many ways than are digital computers for solving nonlinear differential equations. There was a time before digital computers became popular that analog computers were widely used for computation. When digital computers became popular, only a much smaller fraction of computation took place by means of analog computation. But in very recent times, perceptive investigators have come to appreciate that for certain types of real-life situations, it can be very helpful to make use of analog computation, often in a computational system that includes both a digital computer and an analog computer. This has, in very recent times, prompted perceptive investigators to try to think of ways to do analog computation better or faster or less expensively or in a smaller form factor or with greater dynamic range or with better bandwidth or with the ability to handle more complex mathematical computations. For tasks such as solution of a system of differential equations, the analog computation desirably consumes much less computational energy as compared with legacy digital computational approaches.
Such hybrid computation has proven to be particularly powerful for at least two categories of work: sophisticated simulation of systems, and sophisticated control and management of real-life systems.
A hasty reader might assume that what is being discussed is a digital simulation of analog circuitry, or a digital simulation of analog phenomena. Such is not the present discussion. What is being discussed is actual analog circuitry, such as integrators and amplifiers and other elements that make up analog computers, working alongside a digital computer. The challenges being described and, hopefully, solved are physical challenges of physical electrical voltages and currents, not mental steps.
Many analog computers in the past were made using expensive components, whose values remained within relatively narrow tolerances even in the presence of temperature changes and aging. This made such analog computers expensive. Some modern-day analog computer applications call for the designer to try to make the computer less expensive, and this may call for use of less expensive components. The decision to use less expensive components likely leads to a need for more frequent calibration and re-calibration so as to compensate for the effects of temperature changes (ambient and system) during operations. During such time as an analog computer is being calibrated it cannot, of course, be carrying out production calculations. In some applications (for example a student educational environment) it might not be a problem that sometimes the analog computer is not available for actual calculations. But in recent times some applications for analog computation, as mentioned above, are for modeling and control of real-life production systems. In such applications it is desired and indeed substantially necessary that the analog computation be carried out in an uninterrupted (continuous) fashion for long periods of time. If an analog computer were to be taken out of production service to permit recalibration, and if the consequence were an interruption in the ability to carry out the modeling or control of the real-life system, this would be a big problem. This is so because the need for such calibration operations interrupts overall system service. The magnitude of the problem is all the greater if the production system has humans in the loop.
The experienced user of analog computers will also appreciate that each time the analog computer is calibrated, it needs some time for transients to settle after the analog computer is reconnected to its inputs so as to be placed back into production service.
The aforementioned issues especially present themselves in many real-world applications, where downtime of a key analog-computing system may be highly detrimental to key systems within power plants, refineries, and other real-world applications.
To recap the overall challenge, what may be said is that some physical analog computers are used in a production environment, meaning that we want to carry out the analog computations more or less continuously. This lies in conflict with the fact that it is necessary to carry out calibration of particular elements of the analog computer from time to time. The need for calibration (or recalibration) might be due for example to localized temperature changes in a substrate. To carry out calibration, it is necessary to disconnect the analog computer from the production inputs of the system to which it would normally be connected during production service, and to connect it instead to calibration inputs. At such time the outputs of the analog computer, which in production service would be connected to production outputs of the system, would need to be connected to calibration connections.
What is needed is an improved analog-computing scheme that is initially calibrated and then during runtime, can be recalibrated as necessary, and yet somehow accomplishing the recalibration in a way that avoids interruption of the overall system operations.
The inventive disclosures described herein generally pertain to an improved runtime-calibratable analog-computing system. In many embodiments, the improved analog-computing system comprises at least two analog computers, wherein after initial calibration, the system is designed to stagger the runtime calibration modes of each of the at least two analog-computers such that at least one of the analog computers is always in service, thus preventing any downtime for the overall system. In other words, a system user sees one initial calibration, and computing by the overall system is never interrupted.
The foregoing Brief Summary is intended to merely provide a short, general overview of the inventive disclosure described throughout this patent application, and therefore, is not intended to limit the scope of the inventive disclosure contained throughout the balance of this patent application, including any appended claims and drawings.
The inventive disclosures described herein generally pertain to an improved runtime-calibratable analog-computing system. In many embodiments, the improved analog-computing system comprises at least two analog computers, wherein after initial calibration, the system is designed to stagger the runtime calibration modes of each of the at least two analog-computers such that at least one of the analog computers is always in service, thus preventing any downtime for the overall system. In other words, a system user sees one initial calibration, and computing by the overall system is never interrupted.
This Section II is directed generally to an improved runtime-calibratable analog-computing system. Refer to
[1] Achieving Transparent Runtime Calibration of an Analog-Computing System
In the prior art, to maintain the proper calibration of an analog-computing system, the system must be taken out of service, calibrated, then placed back in service. The experienced user appreciates that even the placing back into service is not instant, because there is some lag time before the system is back to optimum performance due to system transients as it comes back on line. See
In one embodiment, two analog computers (or two sub-analog-computers on the same chip) are used. Referring to
The alert reader will be aware that there are several ways to accomplish the computation that uses two or more analog computers as described herein, in coordination with digital computer resources.
The alert reader will further appreciate that if the number of circuit elements in the analog computer named Computer 1 is sufficiently large, then for a given set of differential equations that is to be solved, it may be possible to “construct” two analog computers within Computer 1, each of which is composed of whatever circuit elements are needed to carry out the desired computation. The two analog computers would then be contained within a single chip appearing to the human observer to be Computer 1. This gives rise to the notion of two sub-analog-computers on the same chip.
The teachings of the present invention offer their benefits regardless of whether the two computers that are linked up (with one being calibrated while the other is carrying out the externally observed computations) are on the one hand visible to a human user as separate chips, or are merely distinct analog computers constructed by means of suitable configuration of switching fabric and circuit elements all within a single chip.
See
In
It should be understood by the alert reader that in some preferred embodiments, some of the aforementioned switches 5, 10, 15, 20 should be of a “make-before-break” type to ensure that the transition between analog computers as one reenters its service mode and the other enters the calibration mode does not result in a dropped/lost signal (input or output). Others might be a “break before make” type as will be discussed below.
In more variations, the improved analog-computing system 1 includes a means to compare the outputs of the two analog computers 25, 30 to determine whether the analog computer 25, 30 outputs are within a predetermined differential with respect to each other in order to avoid a sudden and undesired system transient as a result of the swap of the in-service analog computers 25, 30. If the output differential between the oncoming analog computer 25, 30 exceeds some predetermined threshold, then in some embodiments, a means to provide a system alert/annunciation is provided for end users, and/or the handoff/transfer between the analog computers 25, 30 may be suspended to allow for end-user troubleshooting. The overall system 1 is configured to swap the operational modes of Computer-1 25 and Computer-2 30, back and forth, in a similar manner in order to maintain overall system 1 calibration while eliminating any externally observed downtime of system 1.
In other variations, more than two analog computers 25, 30 can be used for such swapping between service and calibration modes in order to increase overall system 1 reliability for especially high-risk practical applications. The aforementioned output-comparison means during the transition of in-service analog computers 25, 30, in which there are three or more analog computers involved, can be subjected to a coincidence-logic or “voting” system whereby if the output values for the majority of the analog computers are within a predetermined threshold of each other, then that majority value (which in variations can be an average of the majority values) will control whether or not to allow an analog computer that has been in calibration mode to come online and take-over the system operations.
The key in making this scheme work is the relative timing of the input and output multiplexers at the analog computers.
As was mentioned above in the discussion relating to
[2] Timing Constraints
Since one analog computer needs to be calibrated and since it is necessary to wait for its transients to die out, while the second analog computer is computing:
T
compute
>T
calibrate
+T
transients
[3] Application to Amplitude Scaling in Analog Computers
To eliminate the back-in-service transient, an embodiment an amplitude-scaling technique for filters can be used, in which the states (that is, capacitor voltages) are held while scaling changes are made, while transients are avoided. See, e.g., U.S. Pat. No. 5,541,600 to E. Blumenkrantz et al., for “Signal processing circuit including a variable gain input stage.” The teachings of U.S. Pat. No. 5,541,600 are incorporated by reference.
This amplitude-scaling technique can be extended to the case of two (or more) analog computers, in which the states (that is, the integrator outputs) are passed from one analog computer to the other. If the pause times can be eliminated or minimized/tolerated in a given application, then this state-information-exchange schema should be considered.
In additional embodiments, the approaches described above can be used for amplitude scaling without interruption. In fact, amplitude scaling can be considered as part of the calibration process. Some examples of circuit configurations that can be exploited to combine such system-input filtering with the calibration schemas presented above are disclosed in U.S. Pat. No. 7,274,760 to Palaskas et al., for “Circuit and method for dynamically modifiable signal processor,” and U.S. Pat. No. 7,274,760 is hereby incorporated by reference.
It may thus be helpful to review what has been described. The approach according to the invention is to provide first and second analog computers, each having respective inputs and outputs. At any given moment one computer or the other can be in production service, serving whatever the user goals are for the production system. This might be assisting in management of an electric vehicle or an electrical energy storage system but could be any environment in which analog computation is helpful. Typical environments where such computation is helpful are environments in which it is a goal to provide solutions or at least near solutions to systems of non-linear differential equations. The solution itself may be put to use directly, for example to determine some concrete result in management of some physical system. Or some near solution may be developed which is in turn provided to a digital system which takes the near solution as a starting point and arrives at a digitally computed result which is put to use directly to determine some concrete result in management of some physical system.
According to the invention, while one of the analog computers is in production service, its inputs are connected to the production inputs of the system and its outputs are connected to the production outputs of the system. This is typically carried out by means of an analog crosspoint switching fabric providing analog switching capability to each of the two analog computers, to calibration equipment, and to the production inputs and outputs.
While one of the analog computers is thus in production service, it is thus possible to carry out calibration of the other of the analog computers. Its inputs are connected to calibration inputs of the calibration equipment, and its outputs are connected to calibration output-receiving signal lines of the calibration equipment. Calibration is then carried out. Depending upon the analog computation elements involved, and their technology, the calibration might take milliseconds or seconds, but for some technologies the calibration might take on the order of minutes.
Because the production function is being supported by the other analog computer (the computer that is not being calibrated), it is not a problem if the calibration process takes seconds or even minutes.
When the calibration process has finished, then a handover takes place. The computer that just finished calibration gets its inputs disconnected from the calibration inputs and connected to the production inputs. This will typically be a “break before make” type of switching.
After a while this computer settles and its outputs can likely be trusted. Upon occurrence of a predetermined condition, the outputs of this computer get connected to the production outputs and the outputs of the other computer (that had previously been in production) get disconnected from the production outputs. This will typically be a “make before break” type of switching.
The alert reader will appreciate that in the figures, the many inputs to a particular analog computer are portrayed symbolically as a single signal line, the many outputs are portrayed symbolically as a single signal line, and so on, for simplicity of portrayal in the figures. The analog signal switching which is actually many analog switches is portrayed symbolically as a single-pole double-throw switch, again for simplicity of portrayal in the figures. It is nonetheless appreciated that in actual implementation there are a multiplicity of signals arriving at the inputs and at the outputs of any given analog computer or calibration apparatus, or at the inputs or at the outputs of the physical system for which the production apparatus is providing analog computation.
A predetermined condition is chosen by the system designer to determine when to connect the outputs of the just-calibrated analog computer to the production outputs. A simple predetermined condition which may be employed is simply to allow the passage of a predetermined time interval, based upon accumulated experience as to the settling time of the various analog computational elements, or based upon a modeled settling time thereof.
A slightly more sophisticated predetermined condition which may be employed is to provide a threshold device (by which is meant a system of threshold devices, one for each pair of signals) to compare output signals from the two analog computers, meaning the computer that has been in production and the computer that just finished calibration. And the threshold may be employed to arrive at a conclusion that settling has happened in the computer that just finished calibration. Given that settling has happened, then the condition is deemed satisfied, and that computer has its outputs connected to the production outputs. In plain language, that computer is put into production service.
The duty cycle for production and calibration is selected taking into account accumulated experience as to the likely pace and magnitude of drift in the analog computer computational elements, or a modeled drift. The goal of course is to carry out calibration often enough to minimize the accumulated drift at any particular time prior to the next calibration having taken place.
The alert reader will also appreciate one other possible benefit to provision of the threshold device mentioned above, namely that it can also be employed during production service to monitor drift in the analog computer that has had the longest time having passed since calibration. This monitoring can be helpful in any of several ways. First, it permits further accumulated experience as to drift, which permits more informed decisions in future as to generally speaking how often recalibration is likely to be needed. Second it could permit individual real-time decisions as to when to carry out calibration in that analog computer. Third, in the event of widely divergent outputs from the two analog computers, this could provide a warning that one of the analog computers may have suffered a failure of some analog computational element.
The threshold device is of course a two-input device. The alert reader will also appreciate that in production environments where high reliability is needed, it would be possible to provide (for example) three analog computers and a three-input voter containing suitable threshold devices and voting logic. In the event of one analog computer running amok, for example due to catastrophic failure of some analog computational element, the voter could work out which two analog computers to keep in production service and which one analog computer to disconnect from production service. The high-reliability system just mentioned can also carry out the calibration processes mentioned earlier, in which one computer provides production computations (or two computers provide production computations) while another computer is undergoing recalibration.
What is described, then, is a method for use with first and second physical analog computers in a production system having production inputs and outputs, each of the analog computers having respective inputs and outputs, and for use with a calibration apparatus having calibration signals to be provided for inputs and receiving signals from outputs, the method comprising the steps of:
connecting the inputs of the first analog computer to the production inputs, and connecting the outputs of the first analog computer to the production outputs, thereby putting the first analog computer into production service;
connecting the inputs of the second analog computer to the calibration input signals, and connecting the outputs of the second analog computer to the calibration output signals, thereby putting the second analog computer into calibration mode;
carrying out calibration of the second analog computer;
disconnecting the inputs of the second analog computer from the calibration inputs and disconnecting the outputs of the second analog computer from the calibration outputs, thereby taking the second analog computer out of calibration mode;
connecting the inputs of the second analog computer to the production inputs;
upon fulfillment of a predetermined condition, connecting the outputs of the second analog computer to the production outputs, thereby putting the second analog computer into production service;
disconnecting the outputs of the first analog computer from the production outputs, thereby taking the first analog computer out of production service;
disconnecting the inputs of the first analog computer from the production inputs;
and repeating this process over and over again. So for example at a subsequent time, the method involves once again connecting the inputs of the first analog computer to the calibration input signals, and connecting the outputs of the first analog computer to the calibration output signals, thereby putting the first analog computer into calibration mode;
carrying out calibration of the first analog computer;
disconnecting the inputs of the first analog computer from the calibration inputs and disconnecting the outputs of the first analog computer from the calibration outputs, thereby taking the first analog computer out of calibration mode;
connecting the inputs of the first analog computer to the production inputs;
upon fulfillment of the predetermined condition, connecting the outputs of the first analog computer to the production outputs, thereby putting the first analog computer into production service;
disconnecting the outputs of the second analog computer from the production outputs, thereby taking the second analog computer out of production service;
disconnecting the inputs of the second analog computer from the production inputs;
connecting the inputs of the second analog computer to the calibration input signals, and connecting the outputs of the second analog computer to the calibration output signals, thereby putting the second analog computer into calibration mode;
carrying out calibration of the second analog computer;
disconnecting the inputs of the second analog computer from the calibration inputs and disconnecting the outputs of the second analog computer from the calibration outputs, thereby taking the second analog computer out of calibration mode;
connecting the inputs of the second analog computer to the production inputs;
upon fulfillment of a predetermined condition, connecting the outputs of the second analog computer to the production outputs, thereby putting the second analog computer into production service;
disconnecting the outputs of the first analog computer from the production outputs, thereby taking the first analog computer out of production service;
disconnecting the inputs of the first analog computer from the production inputs;
connecting the inputs of the first analog computer to the calibration input signals, and connecting the outputs of the first analog computer to the calibration output signals, thereby putting the first analog computer into calibration mode;
carrying out calibration of the first analog computer;
disconnecting the inputs of the first analog computer from the calibration inputs and disconnecting the outputs of the first analog computer from the calibration outputs, thereby taking the first analog computer out of calibration mode;
connecting the inputs of the first analog computer to the production inputs;
upon fulfillment of the predetermined condition, connecting the outputs of the first analog computer to the production outputs, thereby putting the first analog computer into production service;
disconnecting the outputs of the second analog computer from the production outputs, thereby taking the second analog computer out of production service;
disconnecting the inputs of the second analog computer from the production inputs;
connecting the inputs of the second analog computer to the calibration input signals, and connecting the outputs of the second analog computer to the calibration output signals, thereby putting the second analog computer into calibration mode; and
carrying out calibration of the second analog computer.
As discussed earlier the predetermined condition for switching a just-calibrated analog computer back into production service may simply be the passage of a predetermined interval of time. Alternatively, the predetermined condition may be a determination that outputs of the just-calibrated computer have settled, measured by a threshold device.
The switching of inputs of an analog computer from calibration inputs to production inputs may be “break before make”. The switching of outputs of an analog computer that was just calibrated, to the production outputs, and the switching of the other analog computer's outputs away from the production outputs, may be “make before break”.
It will be reviewed that each analog computer will comprise several integrators, each integrator having an internal state defining the output thereof. Before an analog computer is placed into calibration mode, we may store the internal states of the integrators thereof. We then may transfer the stored internal state of integrators thereof to the corresponding integrators of the other analog computer having just been taken out of calibration mode.
A third physical analog computer may be provided in the production system, the third analog computer having respective inputs and outputs. What then can take place is that the inputs of the third analog computer are connected to the production inputs. Then, at a time when all three of the analog computers are in production service, the outputs of the three analog computers may be connected to respective inputs of the threshold device, and, in the event of an excursion of a production output of any one of the analog computers relative to the production outputs of the other two analog computers in excess of a predetermined threshold, the event may be annunciated by means of a communication external to the system.
We thus may have apparatus comprising first and second physical analog computers in a production system having production inputs and outputs, each of the analog computers having respective inputs and outputs, the apparatus further comprising a calibration apparatus having calibration signals to be provided for inputs and receiving signals from outputs, the apparatus further comprising a switching fabric disposed to selectively connect the inputs of the first analog computer to the production inputs or to the calibration inputs, and disposed to selectively connect the outputs of the first analog computer to the production outputs or to the calibration outputs, the switching fabric further disposed to selectively connect the inputs of the second analog computer to the production inputs or to the calibration inputs, and disposed to selectively connect the outputs of the second analog computer to the production outputs or to the calibration outputs.
The various embodiments and variations thereof described herein, including the descriptions in any appended Claims and/or illustrated in the accompanying Figures, are merely exemplary and are not meant to limit the scope of the inventive disclosure. It should be appreciated that numerous variations of the invention have been contemplated as would be obvious to alert readers with the benefit of this disclosure.
Hence, alert readers will have no difficulty devising myriad obvious variations and improvements to the invention, all of which are intended to be encompassed within the scope of the Description, Figures, and Claims herein.
This hereby incorporates by reference U.S. Patent Application No. 62/704,027 filed Oct. 26, 2018, and U.S. Patent Application No. 62/704,067 filed Oct. 7, 2019, and U.S. Patent Application No. 62/911,842 filed Oct. 7, 2019 for all purposes. Should any irreconcilable conflicts arise between this patent application and just-mentioned earlier patent applications for purposes of claim construction or interpretation, then this patent application's teachings shall govern.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/IB2019/058816 | 10/16/2019 | WO | 00 |
| Number | Date | Country | |
|---|---|---|---|
| 62704027 | Oct 2018 | US | |
| 62704067 | Oct 2019 | US | |
| 62911842 | Oct 2019 | US |
