scalable hardware for neural network system

Similar projects worth following
scalable hardware for neural network system

  • ‚Äčall connections OK again

    3droberto2 days ago 0 comments

    Some resistors corrections

  • some problems

    3droberto3 days ago 0 comments

    some old printed PCB looks like a minefield (less varnish sprayed) and needs some fixes.

    I have made some correction on the tester too :)

  • selector for sensor type ADC or MPU

    3droberto7 days ago 0 comments

    I added some feature more in the new printed slave

    clock come OK too :)

  • simple pendulum

    3droberto09/14/2021 at 21:13 0 comments

    I'm going to test first learnings making the typical pendulum test. But I have only 1 slave now (1 sensor=MPU & 1 actuator=servo). I need other slaves to offer more sensors (without actuators for this configuration).

    I wan't overload slaves making many things and anyway I will need to add other slaves with different sensors (ADC, etc...).

  • more fast

    3droberto09/14/2021 at 14:26 0 comments

  • getting data, inference, sending action...

    3droberto09/12/2021 at 20:12 0 comments

    getting data, inference, sending action...  and now calculate reward though OpenCV and apply the backward.

    I'm going to place the Plot class (Graph Class based too) to show the error and I'm going to make a lot of OpenCV enhancement for the reward system, apart of much refactoring..

    All is based in GPU graphs, even the UI and the combinational logic simulator. Neural network was firsts but this use GPU adjacency matrix to perform more inputs quickly than the 8 input/outputs of the combinational logic simulator (without adjacency matrix)

  • changes OK

    3droberto09/12/2021 at 09:33 0 comments

    all changes OK. Working fine on breadboard and PCB.

    Voltage for servo is externally in slave now.

    I need make changes in schematics for better final placements for some components and connectors and order proffesional PCBs when finish but for the moment it's time to continue a bit with code now

  • LM1085 instead LM317 to get 3.3v

    3droberto09/11/2021 at 12:39 0 comments

    Some values changes & LM317 substitution

  • one slave connected

    3droberto09/03/2021 at 16:57 0 comments

    GPU neural network is not showing because I was disable it by the moment to test some performance

  • things

    3droberto08/30/2021 at 21:44 0 comments

    basic tester




    Playing with graffite to get a potenciometer I found one form to maintain graffite on paper, even in plastic.

    Adding Iodine ensure the hardeness of this

    At the left without Iodine and rubbed and at the right with Iodine and rubbed too

    with and without

    you can see iodine on the edges

    Maybe It can be used for 3d printings or air bearings









    NAND gates by using AC phases with no active elements? 
    Teorically delay by active components would not exist (combinational). 

    The half adder is the next to try :D

    ... ...

    not NAND at this but I will continue trying





    I'm going to make two instrumentation amplifiers with OPAMPs to have differential probes to get the diode curves that also can be usefully to be able do some test over my custom compounds and know if I have some something interesting :)


    TL084 as instrumentation amplifier and voltage reference


    Better results with symetrical source. Even with capacitors I have something like a diode curve. I have put a little more impedance to test this function generator but also I get nice sensibility without them for other purposes

    forward diode

    reverse diode





View all 38 project logs

Enjoy this project?



dearuserhron wrote 09/14/2021 at 21:17 point

The idea of using multiple small MCUs instead of big one always was on my mind. The hardest part is to get them talk to each other.

  Are you sure? yes | no

3droberto wrote 09/14/2021 at 21:40 point

I try to divide the work as much as possible so that the master only has to listen to the data of its associated sensor from the slaves and send it all via WIFI to the PC. And receive the action also via WIFI and indicate it to the slave that has that action associated with it.

Even one of the 3 MCUs of the slaves is just to coordinate the other two which are one MCU for the gyroscope and the other MCU for the servo. So that the information is quickly available and flows without much of a jam in making calculations when the master requests things from the slaves.
I have activated PLL to achieve 48MHz too

  Are you sure? yes | no

kavinjhon8 wrote 08/09/2021 at 06:19 point

Thanks for sharing this project details. I also work on different projects. Feel free to visit:

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates