TMD-3: Clever Hall Sensor Hack Leads To Better Turing Demo

We’ll beat everyone to the punch: yes, actually building a working Turing machine, especially one that uses a Raspberry Pi, is probably something that would have pushed [Alan Turing]’s buttons, and not in a good way. The Turing machine is, above all else, a thought experiment, an abstraction of how a mechanical computing machine could work. Building a working one seems to be missing the point.

Thankfully, [Michael Gardi] has ignored that message three times now, and with good reason: some people just grok abstract concepts better when they can lay their hands on something and manipulate it. His TMD-1 was based on 3D printed tiles with embedded magnets — arranging the tiles on a matrix containing Hall effect sensors programmed the finite state machine, with the “tape” concept represented by a strip of eight servo-controlled flip cards. While TMD-1 worked fine, it had some limitations, which [Mike] quickly remedied with TMD-2, a decidedly more complicated affair that used a Raspberry Pi, a camera, and OpenCV to read an expanded state machine with six symbols and six states, without breaking the budget on all the Hall sensors required.

TMD-3 refines the previous design, eschewing the machine vision approach and returning to the Hall effect roots of the original. But instead of using three sensors per tile, [Mike] determined that one sensor would suffice as long as he could mount the magnet at different depths within each tile. That way, the magnetic field for each symbol could be discerned by a single Hall sensor, greatly reducing complexity and expense. An LCD screen and a Raspberry Pi run a console app that shows the tape status, the state machine, and the state transitions.

[Mike] put a ton of work into this one — there are nineteen project logs — and he includes a lot of useful tips and tricks, like designing PCBs directly in KiCAD before even having a schematic. Of course, with a track record like his, we’d expect nothing less.

4 thoughts on “TMD-3: Clever Hall Sensor Hack Leads To Better Turing Demo

  1. Alternative way to do tile detection could be to have a LC resonator in each tile and a grid of PCB coils. Then check what frequency each tile location resonates at after poking it by transmitting a current pulse.

  2. I just bought 10 AS5600.

    Since they can detect rotation of a magnetic field, and while I expect their 4096/12bit resolution wouldn’t be functionally useful in meatworld, I would expect that targeting reasonable fractions of that would probably work fine…

    1 sensor per tile, magnet rotated approximately 10° segments.

    Careful mounting to get the angle right, but depth can be consistent at least.

    1. Cool idea. I can see this working well for a small numbers of sensors. My build required 96 sensors so for me the AS5600 would have been cost (I paid < 50 cents for each linear Hall effect sensor) and complexity (i2c vs single ADC pin for each sensor) prohibitive.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.