Sequential Analog Computers?

AI Thread Summary
The discussion explores the concept of sequential analog computers, questioning whether any systems exist that process fully analog values in a discrete time-space. Classical analog computers operate in continuous time, while digital computers function in discrete time, leading to a debate on the advantages and applications of hybrid systems that combine both approaches. Participants note that while analog computers excel in real-time integration and differentiation, modern digital techniques can emulate these functions with greater speed and accuracy. There is interest in potential applications for analog systems in areas like AI and neuromorphic computing, where they could mimic biological processes. The conversation highlights the evolving landscape of computing technology and the ongoing exploration of hybrid systems.
Runsva
Messages
3
Reaction score
0
I was wondering whether there have been any documented uses of computer systems which work with fully analog values in a discrete time space.

Classical analog computers usually operate with fully analog values in a continous time-space (real-time execution), and normal everyday digital computers usually always operate with digital values in a discrete time-space (sequential execution).

Since there are methods of storing fully analog values in an electronic circuit for an extended period of time (usually stored as a time difference between two signals, 1, 2), I was wondering whether there are any working current or historical examples of a computer system that is able to process fully analog (real) numbers, while working in a discrete time-space, and if so, what their use cases and purposes are.

Thanks for reading my post, any guidance is appreciated.
 
Technology news on Phys.org
There certainly are hybrid systems that have periodic discrete inputs and time-sampled outputs.
Do you have an example application where a hybrid system would be advantageous? What advantages are you wishing to get out of it?
IMHO, for many applications, it would combine the worst of both worlds
 
Welcome to PF.

The advantage of an analog computer was that it could integrate and differentiate as a continuous time process, which could outperform a discrete time-stepping digital emulation. However, digital techniques can now emulate analog computers with a significant speed and accuracy advantage.

The bandwidth and accuracy of an analog circuit is less than a single precision digital processor. Anything that can be done in discrete analog steps can be more accurately emulated in a microcontroller chip today.

Analog "sample and hold" circuits were used to capture fast events for display in "sampling oscilloscopes". There are now switched-capacitor analog-to-digital sampling converters, that can outperform analog S&H circuits in speed and accuracy.

Some of the earliest digital calculators used sample and hold circuits as digital logic memory elements. Dynamic RAM still employs that technique.
 
Analog computers can implement models of differential equations with frequencies that digital computers can not match.
 
Several companies are investigating uses of analog computer chips. I am not an expert on this subject, but I believe that one intended application is in AI and pattern recognition (Neural networks?)
 
Runsva said:
I was wondering whether there have been any documented uses of computer systems which work with fully analog values in a discrete time space.

Classical analog computers usually operate with fully analog values in a continous time-space (real-time execution), and normal everyday digital computers usually always operate with digital values in a discrete time-space (sequential execution).

Since there are methods of storing fully analog values in an electronic circuit for an extended period of time (usually stored as a time difference between two signals, 1, 2), I was wondering whether there are any working current or historical examples of a computer system that is able to process fully analog (real) numbers, while working in a discrete time-space, and if so, what their use cases and purposes are.

Thanks for reading my post, any guidance is appreciated.
I think I would classify an analog television receiver/decoder as such a device. The discrete time space is dictated by the frame rate. Please indicate why this would not meet the criteria (I'm sure many folks are scratching their heads at this point.......).
 
  • Like
Likes Klystron and FactChecker
hutchphd said:
I think I would classify an analog television receiver/decoder as such a device. The discrete time space is dictated by the frame rate. Please indicate why this would not meet the criteria (I'm sure many folks are scratching their heads at this point.......).
Thanks for all your replies;

That's very interesting.
I think that my definition of a "computer" here would include the ability to perform some basic mathematical operations on the stored analog values, such as the linear operations of addition and subtraction, and store the results of those operations. Two analog values stored as time-mode variables (time difference between two signals), for instance, can be added or subtracted together.

I was simply wondering whether there have been any sequential-execution computer system designed in the past that have the ability to work with fully analog (real) numbers.

It appears that there exist a series of analog FPGA devices, known as FPAAs, which appear to be the closest fit to what I'm asking about here, albeit those devices appear to store their values in the classic Sample & Hold circuits istead of storing them as some sort of time-mode variables.
 
FactChecker said:
There certainly are hybrid systems that have periodic discrete inputs and time-sampled outputs.
Do you have an example application where a hybrid system would be advantageous? What advantages are you wishing to get out of it?
IMHO, for many applications, it would combine the worst of both worlds
I'm not sure what the applications of such devices would be, that's part of what I was curious about and why I asked about it here. I'd assume that such systems would be suited for applications in which calculations do not neccessarly need to be fully deterministic, but I was hoping I'd be able to get further ideas for the potential applications of such computer systems here.
 
  • #10
Runsva said:
I'm not sure what the applications of such devices would be, that's part of what I was curious about and why I asked about it here. I'd assume that such systems would be suited for applications in which calculations do not neccessarly need to be fully deterministic, but I was hoping I'd be able to get further ideas for the potential applications of such computer systems here.
The potential applications that I am aware of are massively parallel. See this. There may be other types of applications.
From the linked reference:
"the modern analog revolution is chip-based, with numerous companies delving into its potential, especially in neuromorphic computing. This approach seeks to emulate the human brain’s structure and function, using circuits to mimic neurons and synapses, offering a more efficient and parallel processing alternative to traditional digital methods."
 
  • #11
Hybrid computers are now digital, but with analog modules. The interface to the several analog modules is through D-A and A-D converters. Hybrid computers are stepped sequential, in that they store intermediate analog values in digital memory. Hybrid computers are optimised to run Monte Carlo techniques.
https://en.wikipedia.org/wiki/Hybrid_computer
https://en.wikipedia.org/wiki/Hybrid_computer#VLSI_hybrid_computer_chip

Biological networks gain accuracy through parallel duplication of inaccurate processes. Emulation of the biological network can be by digital or by analog elements. There is a tradeoff, fewer accurate digital nodes, against many inaccurate analog nodes. A digital download can clone an early education, without the wait.
 
  • Like
Likes FactChecker
Back
Top