Just playing with it a bit, the default setup (a decaying value/voltage)…
Lower left, "-1/+1" are your source voltages representing the smallest allowable value (-1) and largest (+1).
Upper left "COEFF" are the coefficient multipliers—potentiometers, that take a voltage/value in (circle) and output (triangle) an attenuated value/voltage. You change the potentiometer values with the "COEFFICIENT POTENTIOMETERS" section on the right side of the split view.
The top section of the analog computer represent the all-important "INTEGRATORS" (5 of them). Circles are inputs (a few 1x and 10x each) and the output are the triangles.
A lime green wire connects from the +1 source voltage to the #1 coefficient multiplier, that attenuated voltage/value feeds "IC" on the first integrator. ("IC" represents the initial condition of the integrator—the value/voltage it defaults to when the machine is reset.)
That first integrator's output (triangle) is being fed into a second coefficient multiplier where the output from that is fed back into the input of said integrator. In short, the integrator is feeding back a portion of its output back into itself.
Another wire (purple) from that integrator goes first to an inverter in the "INVERTER" section (probably obvious what this operator does) and then connects to the "X" of the "OUTPUT" section. The oscilloscope is displaying "X" on "CH1" (channel 1). We see the decaying voltage/value of the integrator on the oscilloscope.
Playing with the coefficient multiplier sliders: P2 represents the amount of feedback, thus the speed with which it decays. If you turn on "AUTO-RUN ON EDITS" you can watch the oscilloscope change after moving a slider.
(Please let me know if I got something wrong above.)
EDIT: created a "hello world", mass+spring for the simulator—you can download the JSON file [1] and try uploading it to the simulator: in the "Design File" section you select the JSON file and "LOAD FILE"—set "OP-TIME" (underneath the oscilloscope display) to 100 ms or so and then "RUN".
You ought to see the position trace of a damped mass on a spring. Coefficient multiplier P1 determines spring constant, P2: damping, and P3 the mass of the spring.
Challenge: see if you can modify the demo to also plot the velocity of the mass+spring (integrator #1) on the oscilloscope (hint: use the "Y" output).
I took a deep dive for 12 months on how analog computers work(ed) and the result was a little kit and book that represent a kind of analog computer primer [1]. In short I took what I learned over the year and made something that would fast track others (you can probably go through the manual + kit in about two weeks and know as much as I do now).
I had originally wanted to get one of those Analog Things but was a little put off by the price (so of course I spent close to an order of magnitude more to develop my own analog computer).
(The ESP32 I added for the backend, to display the voltages/values [2], is, I think, the coolest thing I contributed to the hobbyist analog-computer niche.)
Truly one of the most elegant implementations of all the basics that one could ask for.
In case you haven't dug into this dear reader, the Lunar Lander example is fascinating! Takes a bit of patience to set up, but is truly fun.
I'm left wondering at what other kind of hijinks a real, life-size ANALOG THING could be pressed into executing .. in a general sense I've always had a personal fascination with analog computers in other shapes and forms - pinball machines, electronic music synthesizers, the flowerbed - but like most I suppose, have spent nearly all my life attached to a digital one, in some way or another.
As I see it, a gravity/water-based analog thing is doable, and .. one fine Sunday afternoon, might be the perfect thing to design/prototype with whatever ML model will be waking up for some pushing around.
If there were, say, an electromagnetic means of tracking the state of a pinball as it traverses the playfield, and of course if the drop targets and bumpers and switches and flippers and things were all appropriately wired up, I wonder what it would take to program an analog computer thing to play a mean pinball...
That (electronic) analog computers have integrators is, as I have read, the whole reason analog computers existed—stuck around. Integration is ideal for simulating forces like gravity, etc. over time.
Early experimental X-planes (like the X-15, for example) had flight simulators that were completely analog—modeling the forces of lift, etc. (In the 1970's, there was even a commercial analog flight simulator sold [1].)
Lower left, "-1/+1" are your source voltages representing the smallest allowable value (-1) and largest (+1).
Upper left "COEFF" are the coefficient multipliers—potentiometers, that take a voltage/value in (circle) and output (triangle) an attenuated value/voltage. You change the potentiometer values with the "COEFFICIENT POTENTIOMETERS" section on the right side of the split view.
The top section of the analog computer represent the all-important "INTEGRATORS" (5 of them). Circles are inputs (a few 1x and 10x each) and the output are the triangles.
A lime green wire connects from the +1 source voltage to the #1 coefficient multiplier, that attenuated voltage/value feeds "IC" on the first integrator. ("IC" represents the initial condition of the integrator—the value/voltage it defaults to when the machine is reset.)
That first integrator's output (triangle) is being fed into a second coefficient multiplier where the output from that is fed back into the input of said integrator. In short, the integrator is feeding back a portion of its output back into itself.
Another wire (purple) from that integrator goes first to an inverter in the "INVERTER" section (probably obvious what this operator does) and then connects to the "X" of the "OUTPUT" section. The oscilloscope is displaying "X" on "CH1" (channel 1). We see the decaying voltage/value of the integrator on the oscilloscope.
Playing with the coefficient multiplier sliders: P2 represents the amount of feedback, thus the speed with which it decays. If you turn on "AUTO-RUN ON EDITS" you can watch the oscilloscope change after moving a slider.
(Please let me know if I got something wrong above.)
EDIT: created a "hello world", mass+spring for the simulator—you can download the JSON file [1] and try uploading it to the simulator: in the "Design File" section you select the JSON file and "LOAD FILE"—set "OP-TIME" (underneath the oscilloscope display) to 100 ms or so and then "RUN".
You ought to see the position trace of a damped mass on a spring. Coefficient multiplier P1 determines spring constant, P2: damping, and P3 the mass of the spring.
Challenge: see if you can modify the demo to also plot the velocity of the mass+spring (integrator #1) on the oscilloscope (hint: use the "Y" output).
[1] https://drive.google.com/file/d/16d0MLmc9Tk8SSlqz8RtLDBxXeAz...)
I had originally wanted to get one of those Analog Things but was a little put off by the price (so of course I spent close to an order of magnitude more to develop my own analog computer).
(The ESP32 I added for the backend, to display the voltages/values [2], is, I think, the coolest thing I contributed to the hobbyist analog-computer niche.)
[1] https://www.tindie.com/products/jcalhoun/anna-analog-compute...
[2] https://github.com/EngineersNeedArt/Anna-Analog-Computer/tre...
In case you haven't dug into this dear reader, the Lunar Lander example is fascinating! Takes a bit of patience to set up, but is truly fun.
I'm left wondering at what other kind of hijinks a real, life-size ANALOG THING could be pressed into executing .. in a general sense I've always had a personal fascination with analog computers in other shapes and forms - pinball machines, electronic music synthesizers, the flowerbed - but like most I suppose, have spent nearly all my life attached to a digital one, in some way or another.
As I see it, a gravity/water-based analog thing is doable, and .. one fine Sunday afternoon, might be the perfect thing to design/prototype with whatever ML model will be waking up for some pushing around.
If there were, say, an electromagnetic means of tracking the state of a pinball as it traverses the playfield, and of course if the drop targets and bumpers and switches and flippers and things were all appropriately wired up, I wonder what it would take to program an analog computer thing to play a mean pinball...
Early experimental X-planes (like the X-15, for example) had flight simulators that were completely analog—modeling the forces of lift, etc. (In the 1970's, there was even a commercial analog flight simulator sold [1].)
[1] https://inspire.eaa.org/2022/02/16/eaas-attic-atc-510-person...