I want to build a circuit that uses a single electrode that can act as both an input and an output.

I have attached diagram of my first guess of how to do this.

Do you think this would work? I'm a newbie, so be nice

Attached Images

splitcircuit.JPG (8.6 KB, 29 views)

 

Try telling us EXACTLY what you're trying to do, and not how you think it needs to be done.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

 

I want to build a neural network out of modules (i.e. blocks).

Each module has 6 sides with one electrode on each side. The modules are stacked together so that one electrode on one module touches one electrode on another module.

Each electrode acts as both an input and an output for the simple analog circuit inside each module. I need to split the input and output at each electrode so I don't get feedback within the module circuit.

 

What will the analog signals be used for? What is generating them? This is something for machine AI?

__________________
"Remember, you're special.....just like everyone else."

 

The goal of any neural network is ultimately machine AI, so I suppose mine is no different.

The analog signals could be any type of sensor input, for example the signal for one electrode could be from a single pixel of a CCD array.

 

What are you using as a voltage reference?


Torben

__________________
Curiosity was framed. Ignorance killed the cat.

 

ah, good question Torben.

Initially the modules would be be connected to and powered by positive and negative/earth rails.

Ultimately, I would hope to power the modules wirelessly or perhaps using an "Avramenko one wire circuit" (see this discussion for details).

 

I agree with the comments on that site: you must have a circuit in order for it to work. There are a couple of ideas toward the end involving using the air as a dielectric to capacitively couple the devices but that won't be trivial and I wouldn't think it worth the effort.

Remember, copper is expensive. If there were a reasonable way to skip running ground wires all over everything, then designers would do so. No offense, but I think the idea requires further baking. Maybe something related to RFID would be more feasible.

How large will the blocks be? What experience do you have with electronics?


Torben


Torben

__________________
Curiosity was framed. Ignorance killed the cat.

 

At the moment I want to focus on the single electrode input/output circuit using positive and earth rails rather than the pitfalls of wireless (or single wire) power transfer.

The blocks would be as small as possible. I was thinking the circuit would fit on a 5x5cm PCB, thus a module would be 5x5x5cm.

I took a course in electronics a while ago but never used these skills outside of the classroom, so I consider myself at an intermediate level.

 

What are you using as memory for the learning function? If you are using micros as the intelligence why not just use the UART? I must admit to not have read up about neural nets.

 

My idea is a fully analog neural network, so I don't want to use a UART or any other digital transmission device.

For memory I'm thinking of using a memristor (or probably an active equivalent of a memristor).

 

If it's fully analog then any connections depend entirely on exactly what it is.

Although it sounds a completely useless idea to be honest?.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

 

sorry, I don't understand what this means.

Basically this is a toy/tool for investigating analog neural networks. It may turn out to be completely useless, but I will not be convinced until either a) someone provides a compelling reason why it wouldn't work, or b) I test it out myself.

 

You need to provide the circuits of the nodes, overwise how can anyone suggest interconnections.

Build it and see than, but your question is much too vague to offer any suggestions - how complicated is each node going to be?, and how many were you thinging of?.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

 

What I want to know is:

can a single electrode be used for signals going into and coming out of an analog signal processor?

If so, what circuit could be used to achieve this? (the one constraint is: I don't want the output signal going straight back into the input).

If the answer is "yes, it's possible", I can move onto to considering other elements of the project. If the answer is "no, it's impossible", that's fine, I can reassess the feasibility of the project.

I don't understand how the "circuits of the nodes", "how complicated is each node going to be" and "how many were you thinking of" is significant to answering the question.

To help visualize the problem, let's say the voltage at the electrode, input and output are all within the range from 0 to 5V.