Let's Make Robots!

slowing time constant(discharge) for cap

so....I'm back again!...

I have a small question,

I have finished a perfect classical conditioning analogue circuit for an AI, now i am gonna be using 4 "lobes" for the brain, i.e, touch sensor, colour sensor, memory bank, and output load. I am gonna be using the 2 different sensors for the stimulus(reward and neutral) and the output load for the response...of course,

so, I am gonna be using some arrays of capacitors as the memory bank, but a capacitor dishcarges fast...1.how do i slow the discharge?

I have a solution that I don't know if it will work, here's an example circuit that I crudely drew :P ....btw, this is not the memory bank of course, just an example to slow down the discharge...2. will this work?:

btw, to operate the circuit above, first charge the caps using the charging switch, then turn it off, then turn the main swith of the transistor......(note: transistor is used as a switch here)will the time constant slow down because of the resistor and the sum of the parallel capacitor? and will the amount of the resistance and the capacitance will be directly proportional as the time? (i saw  a subject saying that t=r.c....(t=time constant in sec.)


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

tested with a 1 farad cap and a 69 ohm resistor...it worked! Max's link shows that it should last till 69 seconds...should be good :)

note: calculating the rc time constant is T=RC

One farad is a 'helluva' capacitor. Those 'super caps' can get expensive.


[ Not sure why it double-posted.  Ignore the redundancy. ]

the double post is ok, I think it's because of double clicks...

anyways, I've found a bunch of those 1 farad caps....so you are saying they are super caps? if so, then...wow!

I found 10 of them from a board that a friend gave who owns a computer shop....

this should be good >:)  <====evil grin :P

One farad is a 'helluva' capacitor. Those 'super caps' can get expensive.

The flipflop circuit I was thinking of is just two transistors plus 5 or 6 resistors & your capacitor(s). It could be made monostable.

With that or the circuit you have above, you must remember that the transistor must pass a certain amount of current to fully turn on and that is 1/10 the collector-emitter current in saturation mode. That will likely drain the capacitor more quickly than you are hoping.

Another option would be if you had a 555 timer chip on hand.

Question: how long of a time are you wanting the capacitor(s) to stay charged? milli-seconds, seconds, hours...? That will make a big difference in design.


ya...but I am using my own circuit of a timed memory...it's purely of logic and purely a patern of a psychological conditioning, plus I am using analogue since traditionaly, the human brain is analogue, flip-flop is digital...and as I have said, I am using only items I can find here...there will be no timer ic's etc.

and the time will probably only be about 10 seconds or so, a short term memory....the memory is basically a real memory of what the robot has done, in short, it really is* a memory. And since in classical conditioning, after the test subject had been tested with unconditioned stimulus, cwith unconditioned response...but if you send the neutral stimulus, it will be a conditioned with a conditioned response...but after for a while the conditioned response to the neutral stimulus will be gone.....

in short, my circuit, if this memory bank work(if I can slow down the time constant to 10 sec) will be purely soft AI, it will adapt base on it's memory.....

Ok, since you want it to last several seconds, I might suggest cascading a couple transistors so the drain on the cap is minimal. In a darlington configuration, the overall gain is equal to the gain of the first transistor times the gain of the second. In analog mode, a 2N3904 has a gain up to 300, but let's say you use values that will give 100 average, then the gain of two cascaded will be 10,000.


So in that case, were you to power this with 5 volts and change the collector resistor from 220 Ω to around 500 Ω (, then the resistor R could easily be as high as (500 x 10000) or around 5 megohms (5MΩ) and would drain off the charge on the capacitor(s) very slowly.  It would be a way to have a high input impedance without an FET.

[In actuality, the second transistor would be higher gain, possibly as high as 300, while the first one will be lower gain because it is passing a much lower collector current, but you should be able to get a 10,000 gain overall.]

I will experiment with that as well!

The ABC used a spinning capacitor bin for memory. I guess the questions here are a) precisely what you mean by "work" when you say "will it work?" (ie, when do you want the motor to run, to stall etc. in response to what series of switch closures?) and b) what values of capacitor, resistor, motor and transistor do you have or have access to?

a. will the resistors slow down the capacitor's discharge?

b. the values are unknown so far....all of the components are anything i can find...cn't buy anything online and there are no clear electronics store near here....