In my upper-division analog electronics class (the hard one), our lab project throughout the quarter was to build an analog computer that simulated the physics of a bouncing ball. Physical variables of the system were adjustable (gravity constant, coefficient of restitution, etc), and the ball was "released" by pressing a button. The output was viewed on an oscilloscope.
One of the hardest 10 weeks of my life, but also one of the most rewarding. Our team was one of the few that actually got it working in the end. I had to custom-make a gigantic breadboard to hold the entire circuit.
Today I still work in hardware, but mostly with digital circuits. While my analog knowledge has decayed over the last decade, that project and it's success gives me great confidence any time I have to deal with the domain.
My version of this was a 10-week discrete RF circuits course in graduate school. We had to build a fully functional GHz transceiver out of small FR4 PCBs (< quarter wavelength) and throw-away leaded BJT transistors. Neither were suitable for GHz circuits, so the course was hard by design. I learned so much and developed an intuition for electromagnetics that I still carry 20 years later.
Hey, I made something like this a couple months ago! (Except it's more like "Tennis for Two", so you also hit the ball in the X direction, and there's another button to hit it back in the other. I didn't have any space or potentiometers left to set the gravity, but it wouldn't be difficult.)
I also learned heaps! (Including after a few weeks when the circuit stopped working properly because one of the relays started to work just a little slower than another one, heh.) If anyone's interested, https://blog.qiqitori.com/2024/08/implementing-tennis-for-tw...
> Today I still work in hardware, but mostly with digital circuits. While my analog knowledge has decayed over the last decade, that project and it's success gives me great confidence any time I have to deal with the domain.
Do you think about the analog qualities of your traces when laying things out? If so then the course was well taken.
In my observations I've found that too many digital engineers assume a differential pair will save them without actually fixing the impedance and parasitic issues. Particularly as the timings of things become so much more precise analog is so important. People forget that a digital circuit is just an analog one under the covers.
Did the mathematical model being used have a differentiable heigh function? I’m imagining it would be the simplest if it didn’t but that could cause problems in the electronics.
Also what components did you have access to, just op amps?
There’s no better introduction to signals and systems than a modular synthesizer IMO - the combination of tactility and audibility for multi-sensory learning is so great at building intuition - and more importantly, excitement! - for signal processing.
This looks like a cool project in the same spirit!
I looked into it hook one up to my rack and ultimately decided that even with the high price of euro modules it's still not really sensible/useful for what it includes. There's a lot of more useful modules you can try the same vibe with for euro though.
For analog logic, Maths from Make Noise is the obvious. ANA from Mystic Circuits is pretty neat too.
For digital logic (Combining both types of logic is very fun IMHO) O_c (or uO_c) running hemisphers has a mode to construct complex logic gate changes in the module, which can be fun.
Also, consider that a LPF is effectively an integrator and a HPF a differentiator and you already have 90% of it already, probably.
You can do that but it's kind of overkill. With analog computing you think in the time domain with differential equations, and accuracy is more important than speed.
With synthesizers you think mostly in the frequency domain and speed is more important than accuracy. Integrators from AC become lowpass filters, adders become mixers*, multipliers become ring modulators, etc.
* In audio, a "mixer" is an adder. In RF a mixer is a multiplier.
As a different sort of analog computer, I have long been wondering about a “compiler” for fluidic logic that can output devices you could 3D print which would then operate on pneumatic or hydraulic signals. Probably entirely useless, but wouldn’t be affected by an EMP!
That idea was shamelessly inspired by the soft fluidic robot some years back.
A built-in scope display would be nice. Like this $10 module.[1]
Then you could use this standalone. They charge EUR 499 for the thing, after all.
The way you usually run an analog computer is to put it into fast repeat mode (which they call REPF), where it cycles between initial condition mode and run mode. Outputs go to a scope. Then you can twiddle the knobs and see the output respond immediately.
The other modes are used mostly during setup and debug.
I don't see it. Any non trivial analog computation involves a very large circuit, which has the problems of normal programming (bugs) and graphic programming (write-only), but with the extra pitfalls of electronics (resistance, delay, the resulting oscillations). And then you have to read all the outputs. That's going to be slow and expensive to build.
In what concrete problems do you (or Veritasium) think analog computing could beat a GPU?
In my upper-division analog electronics class (the hard one), our lab project throughout the quarter was to build an analog computer that simulated the physics of a bouncing ball. Physical variables of the system were adjustable (gravity constant, coefficient of restitution, etc), and the ball was "released" by pressing a button. The output was viewed on an oscilloscope.
One of the hardest 10 weeks of my life, but also one of the most rewarding. Our team was one of the few that actually got it working in the end. I had to custom-make a gigantic breadboard to hold the entire circuit.
Today I still work in hardware, but mostly with digital circuits. While my analog knowledge has decayed over the last decade, that project and it's success gives me great confidence any time I have to deal with the domain.
If you want to take a look, here's a pretty similar project: https://www.analogmuseum.org/english/examples/bouncing_ball_...
My version of this was a 10-week discrete RF circuits course in graduate school. We had to build a fully functional GHz transceiver out of small FR4 PCBs (< quarter wavelength) and throw-away leaded BJT transistors. Neither were suitable for GHz circuits, so the course was hard by design. I learned so much and developed an intuition for electromagnetics that I still carry 20 years later.
Hey, I made something like this a couple months ago! (Except it's more like "Tennis for Two", so you also hit the ball in the X direction, and there's another button to hit it back in the other. I didn't have any space or potentiometers left to set the gravity, but it wouldn't be difficult.)
I also learned heaps! (Including after a few weeks when the circuit stopped working properly because one of the relays started to work just a little slower than another one, heh.) If anyone's interested, https://blog.qiqitori.com/2024/08/implementing-tennis-for-tw...
Great writeup and thanks for including the CircuitJS simulations.
> Today I still work in hardware, but mostly with digital circuits. While my analog knowledge has decayed over the last decade, that project and it's success gives me great confidence any time I have to deal with the domain.
Do you think about the analog qualities of your traces when laying things out? If so then the course was well taken.
In my observations I've found that too many digital engineers assume a differential pair will save them without actually fixing the impedance and parasitic issues. Particularly as the timings of things become so much more precise analog is so important. People forget that a digital circuit is just an analog one under the covers.
The way my teacher put it: “High speed digital electronics is just analog with a rise time.”
Did the mathematical model being used have a differentiable heigh function? I’m imagining it would be the simplest if it didn’t but that could cause problems in the electronics.
Also what components did you have access to, just op amps?
Just op-amps and FETs for the active components. The design from my memory was:
- To get position, 2 integrators were applied to an adjustable voltage representing gravity.
- The FETs were used to set initial states of the integrators.
- A comparator used to detect the table (y=0), flip the velocity and apply a scaling factor for restitution
The math was actually quite simple given its just the standard velocity equations — the challenge was in handling state changes in the electronics.
I looked around a little more and this video is a very close replica of what we built: https://www.youtube.com/watch?v=qt6RVrmvh-o
There’s no better introduction to signals and systems than a modular synthesizer IMO - the combination of tactility and audibility for multi-sensory learning is so great at building intuition - and more importantly, excitement! - for signal processing.
This looks like a cool project in the same spirit!
I agree. I highly recommend [0] Moritz Klein's channel. amazing explanations and learning effect.
[0] https://youtube.com/@MoritzKlein0/videos
Just previewed the hi-hat from scratch vid. This channel looks great. Thanks for sharing!
As someone that has been in the synth diy world for a quarter of century, I couldn’t agree more!
I was going to say the same - does this follow 1V/octave standard, and is this available in Eurorack format? ;-)
https://the-analog-thing.org/wiki/FAQ_Synthesizers
TL;DR: No.
I looked into it hook one up to my rack and ultimately decided that even with the high price of euro modules it's still not really sensible/useful for what it includes. There's a lot of more useful modules you can try the same vibe with for euro though.
For analog logic, Maths from Make Noise is the obvious. ANA from Mystic Circuits is pretty neat too.
For digital logic (Combining both types of logic is very fun IMHO) O_c (or uO_c) running hemisphers has a mode to construct complex logic gate changes in the module, which can be fun.
Also, consider that a LPF is effectively an integrator and a HPF a differentiator and you already have 90% of it already, probably.
Cool, I was thinking about the other way around, using an analog computer to build synthesizers.
You can do that but it's kind of overkill. With analog computing you think in the time domain with differential equations, and accuracy is more important than speed.
With synthesizers you think mostly in the frequency domain and speed is more important than accuracy. Integrators from AC become lowpass filters, adders become mixers*, multipliers become ring modulators, etc.
* In audio, a "mixer" is an adder. In RF a mixer is a multiplier.
This fella is using it as part of his music making https://www.instagram.com/stephano.music/
As a different sort of analog computer, I have long been wondering about a “compiler” for fluidic logic that can output devices you could 3D print which would then operate on pneumatic or hydraulic signals. Probably entirely useless, but wouldn’t be affected by an EMP!
That idea was shamelessly inspired by the soft fluidic robot some years back.
> wouldn’t be affected by an EMP!
Even better, it would only be affected by relatively rare phenomena, such as vibration, temperature change, orientation and rotation.
and leaks.
Something like that is inside some automatic transmissions.
Sperry UNIVAC once built a 4-bit fluidic ALU as a demo, but it was useless.
> that can output devices you could 3D print
Make it analog all the way by hooking it up directly to a lathe or milling machine.
A built-in scope display would be nice. Like this $10 module.[1] Then you could use this standalone. They charge EUR 499 for the thing, after all.
The way you usually run an analog computer is to put it into fast repeat mode (which they call REPF), where it cycles between initial condition mode and run mode. Outputs go to a scope. Then you can twiddle the knobs and see the output respond immediately.
The other modes are used mostly during setup and debug.
Hours of fun. Ages 14 and up.
[1] https://www.alibaba.com/product-detail/YIXINTAI-DSO138-Digit...
Prior discussion:
https://news.ycombinator.com/item?id=36165513 (2023)
https://news.ycombinator.com/item?id=28614840 (2021)
It's a cute toy and a fun educational tool, but "computing for the future" seems like a bit of an overstatement.
It’s been making a comeback in CV and ML. It’s ridiculously low latency and cheap.
It’s unlikely to replace digital computers, but it might find a new specialized home in components.
I think they're saying analog computers could be the future of computing.
Veritasium explains it really well in general here (and demos the device) https://www.youtube.com/watch?v=GVsUOuSjvcg
I don't see it. Any non trivial analog computation involves a very large circuit, which has the problems of normal programming (bugs) and graphic programming (write-only), but with the extra pitfalls of electronics (resistance, delay, the resulting oscillations). And then you have to read all the outputs. That's going to be slow and expensive to build.
In what concrete problems do you (or Veritasium) think analog computing could beat a GPU?
> In what concrete problems do you (or Veritasium) think analog computing could beat a GPU?
Solving (systems of) ODEs without the issues that can arise from numerical solving like numerical instability. Oh and it does that instantaneously.
I've ordered one last holidays and haven't had the time to use it yet. Unfortunately it doesn't fit in the famous dev board drawer.
is it possible to buy this thing in the USA (no vat)