3 ms·

The greatest sin of undergraduate engineering education is sequestering signals and systems into electrical engineering curricula. I understand why it's done th

by duped 4y ago

The greatest sin of undergraduate engineering education is sequestering signals and systems into electrical engineering curricula. I understand why it's done that way (I even had to fight to take the course early, sidestepping some prereqs for reasons). But it's really so foundational to understanding concepts of stability, resonance, information/energy flow (from the conceptual perspective), and the simple analytical tools for building a solid conceptual base. It takes a semester to hammer home that step response matters, positive feedback bad, negative feedback usually good, and topologies are useful.

Agentlien 4y ago
I took some electrical engineering courses but mainly software engineering with a focus on low-level details and graphics. The more I specialize in graphics the more I realize I need much more knowledge about signal processing than we were taught at the university.
- AB1908 4y ago
  Could you demonstrate with an example?
  - Agentlien 4y ago
    I can give a few. In general, the theory behind rendering generally treats what's on the screen as a discretization of a continuous visual signal. Post processing especially is largely about signal processing. Anti-aliasing (AA) is a very clear example where the lack of it leads to moiré patterns and jaggies. Understanding a lot of AA techniques is simplified by seeing the frame not as a discrete set of pixels but as a continuous signal being sampled (and can be sampled at multiple sub-pixel points per pixel). A lot of other screen effects are essentially filters applied to the graphical signal (sobel, gaussian blur, ...) and understanding them from a signal processing view helps understanding how to modify and optimize them. A good example here is identifying whether your effect is a separable filter which can be split into a horizontal and vertical pass. Seeing the image as a continuous signal/field being sampled is also the theoretical basis for a lot of visual effects used in physically-based rendering and things like screen-space ambient occlusion. Finally, if you want to write your own ray-tracer it really helps to be able to take this view of things once you get past the basics.
codethief 4y ago
Can you recommend any good introductory books on the topics you mentioned for someone who studied (theoretical/mathematical) physics but never electrical engineering or signal processing? (Read this as: I am more or less familiar with Nyquist-Shannon's theorem but that's about it.)
- lagrange77 4y ago
  Not a book, but this series on Youtube is excellent. https://www.youtube.com/watch?v=Pi7l8mMjYVE&list=PLMrJAkhIeNNR20Mz-VpzgfQs5zrYi085m https://www.youtube.com/watch?v=Pi7l8mMjYVE&list=PLMrJAkhIeN... > familiar with Nyquist-Shannon's theorem BTW: In control theory there is also https://en.wikipedia.org/wiki/Nyquist_stability_criterion https://en.wikipedia.org/wiki/Nyquist_stability_criterion
- scrlk 4y ago
  Signals & Systems by Oppenheim is a classic introductory textbook.
- duped 4y ago
  ime the books are very loaded on theory and it's hard to connect that to practice and concepts. I don't know of any book that touches on what poles/zeros mean or why they're useful outside what are (seemingly) contrived properties of formulae like stability. It's one of those things that's just useful to go through in a university course, even just auditing it. The practical problems and lectures are extremely useful, particularly due to the overhead of notation (the practice is relatively young and derived from a different tradition of education, and a lot of contemporary notation can seem a little foreign, or is filled with shorthand). Applied Digital Signal Processing by Manolakis & Ingle is the book I always turn to for reference and the code examples don't suck. Oppenheim & Schafer is a classic but frankly only useful as a reference, that tome is a bit dated otherwise. The Scientists and Engineer's guide to DSP is also not bad as a practical text.
trenchgun 4y ago
I was already thinking of taking it this summer, but you convinced me to do it for sure. Thank you!
nbadg 4y ago
I went to school for mechanical engineering (though I now work in software). We were required to take signals and systems, but if I remember right it was a weed-out course for most MechEs (it certainly was a challenge for me, though I think that had more to do with the curriculum than the topic). Those lessons might have been hard-won on my part, but I definitely still use them. The general concepts (feedback loops etc) are applicable basically everywhere in life, and I still find uses for literal actual math (like using a convolution kernel to do rolling window sampling in numpy).
- thrwyoilarticle 4y ago
  I can imagine it's used for car suspension?
  - dboreham 4y ago
    Wasn't control theory invented to understand the dynamic behavior of steam engine governors?
  - nbadg 4y ago
    That's not considered signals and systems, at least not where I went to school (unless you're talking about fully-active suspensions, but those are very rare and highly specialized). Rather, that would be down the "dynamics" course progression -- which is bread and butter for MechEs in those lines of work. That's also extremely useful, but it's generally a different subject matter, at least until you get into graduate-level dynamics combined with upper-undergrad-to-graduate-level numerical methods.
    - jbay808 4y ago
      Vibration isolation can definitely be thought of within a systems framework, although it's perhaps overkill for passive filtering as you won't be doing much in the way of feedback.
koliber 4y ago
Does anyone know of any courses which could explain concepts of stability, resonance, information/energy flow and help build a solid conceptual base for managers or entrepreneurs? These concepts are crucial when making business decisions. I've been building up an understanding of this through experience. If there was a way to shortcut this and provide such education deliberately, it would allow people to become better decision makers quicker.
nabla9 4y ago
"signals and systems" - covering signals processing, systems and control theory is very generic subject. You need it in economics, biology, chemistry, physics, computer science, statistics, electrical engineering, robotics , automation, logistics, ... It should be taught like calculus or linear algebra, so that everyone gets gist of the basics before learning to apply it into their specific field.
Jach 4y ago
I always thought my college education was backwards, with the exception that differential equations (and laplace transforms, which can help lay the groundwork for other transforms) came early enough that I could get by -- though it'd have been better if they were earlier, like high school, and if I hadn't been able to skip two semesters of calculus thanks to my HS calculus then differential equations would have come even later in college. But as a CE student at a school nominally more about video game making education I ended up first taking as CS electives an audio processing course, and then an image processing course, before the CE side reached control systems, which I had to retake after taking the next digital signal processing course, after which control systems made a lot more sense. It was only after graduating that I felt like I had reached a level of sophistication to go back and really grok all the related theory and go deep into applications. Maybe that's how graduate students are meant to feel? But I just went into full time enterprise work and 'retired' after 6 years of that, so I've since forgotten a lot... Still, the concept of feedback has proved useful in systems analysis from time to time, and it's a framework that I think could yield many low hanging fruit in other disciplines. (The book Behavior: The control of perception applies control theory to psychology in a convincing way but it's understandably been neglected by psychologists who aren't often very sophisticated mathematically.)

by duped 4y ago

Agentlien 4y ago

I took some electrical engineering courses but mainly software engineering with a focus on low-level details and graphics. The more I specialize in graphics the more I realize I need much more knowledge about signal processing than we were taught at the university.

AB1908 4y ago
Could you demonstrate with an example?
- Agentlien 4y ago
  I can give a few. In general, the theory behind rendering generally treats what's on the screen as a discretization of a continuous visual signal. Post processing especially is largely about signal processing. Anti-aliasing (AA) is a very clear example where the lack of it leads to moiré patterns and jaggies. Understanding a lot of AA techniques is simplified by seeing the frame not as a discrete set of pixels but as a continuous signal being sampled (and can be sampled at multiple sub-pixel points per pixel). A lot of other screen effects are essentially filters applied to the graphical signal (sobel, gaussian blur, ...) and understanding them from a signal processing view helps understanding how to modify and optimize them. A good example here is identifying whether your effect is a separable filter which can be split into a horizontal and vertical pass. Seeing the image as a continuous signal/field being sampled is also the theoretical basis for a lot of visual effects used in physically-based rendering and things like screen-space ambient occlusion. Finally, if you want to write your own ray-tracer it really helps to be able to take this view of things once you get past the basics.

codethief 4y ago

Can you recommend any good introductory books on the topics you mentioned for someone who studied (theoretical/mathematical) physics but never electrical engineering or signal processing? (Read this as: I am more or less familiar with Nyquist-Shannon's theorem but that's about it.)

lagrange77 4y ago
Not a book, but this series on Youtube is excellent. https://www.youtube.com/watch?v=Pi7l8mMjYVE&list=PLMrJAkhIeNNR20Mz-VpzgfQs5zrYi085m https://www.youtube.com/watch?v=Pi7l8mMjYVE&list=PLMrJAkhIeN... > familiar with Nyquist-Shannon's theorem BTW: In control theory there is also https://en.wikipedia.org/wiki/Nyquist_stability_criterion https://en.wikipedia.org/wiki/Nyquist_stability_criterion
scrlk 4y ago
Signals & Systems by Oppenheim is a classic introductory textbook.
duped 4y ago
ime the books are very loaded on theory and it's hard to connect that to practice and concepts. I don't know of any book that touches on what poles/zeros mean or why they're useful outside what are (seemingly) contrived properties of formulae like stability. It's one of those things that's just useful to go through in a university course, even just auditing it. The practical problems and lectures are extremely useful, particularly due to the overhead of notation (the practice is relatively young and derived from a different tradition of education, and a lot of contemporary notation can seem a little foreign, or is filled with shorthand). Applied Digital Signal Processing by Manolakis & Ingle is the book I always turn to for reference and the code examples don't suck. Oppenheim & Schafer is a classic but frankly only useful as a reference, that tome is a bit dated otherwise. The Scientists and Engineer's guide to DSP is also not bad as a practical text.

trenchgun 4y ago

I was already thinking of taking it this summer, but you convinced me to do it for sure. Thank you!

nbadg 4y ago

I went to school for mechanical engineering (though I now work in software). We were required to take signals and systems, but if I remember right it was a weed-out course for most MechEs (it certainly was a challenge for me, though I think that had more to do with the curriculum than the topic). Those lessons might have been hard-won on my part, but I definitely still use them. The general concepts (feedback loops etc) are applicable basically everywhere in life, and I still find uses for literal actual math (like using a convolution kernel to do rolling window sampling in numpy).

thrwyoilarticle 4y ago
I can imagine it's used for car suspension?
- dboreham 4y ago
  Wasn't control theory invented to understand the dynamic behavior of steam engine governors?
- nbadg 4y ago
  That's not considered signals and systems, at least not where I went to school (unless you're talking about fully-active suspensions, but those are very rare and highly specialized). Rather, that would be down the "dynamics" course progression -- which is bread and butter for MechEs in those lines of work. That's also extremely useful, but it's generally a different subject matter, at least until you get into graduate-level dynamics combined with upper-undergrad-to-graduate-level numerical methods.
  - jbay808 4y ago
    Vibration isolation can definitely be thought of within a systems framework, although it's perhaps overkill for passive filtering as you won't be doing much in the way of feedback.

koliber 4y ago

Does anyone know of any courses which could explain concepts of stability, resonance, information/energy flow and help build a solid conceptual base for managers or entrepreneurs? These concepts are crucial when making business decisions. I've been building up an understanding of this through experience. If there was a way to shortcut this and provide such education deliberately, it would allow people to become better decision makers quicker.

nabla9 4y ago

"signals and systems" - covering signals processing, systems and control theory is very generic subject. You need it in economics, biology, chemistry, physics, computer science, statistics, electrical engineering, robotics , automation, logistics, ... It should be taught like calculus or linear algebra, so that everyone gets gist of the basics before learning to apply it into their specific field.

Jach 4y ago

I always thought my college education was backwards, with the exception that differential equations (and laplace transforms, which can help lay the groundwork for other transforms) came early enough that I could get by -- though it'd have been better if they were earlier, like high school, and if I hadn't been able to skip two semesters of calculus thanks to my HS calculus then differential equations would have come even later in college. But as a CE student at a school nominally more about video game making education I ended up first taking as CS electives an audio processing course, and then an image processing course, before the CE side reached control systems, which I had to retake after taking the next digital signal processing course, after which control systems made a lot more sense. It was only after graduating that I felt like I had reached a level of sophistication to go back and really grok all the related theory and go deep into applications. Maybe that's how graduate students are meant to feel? But I just went into full time enterprise work and 'retired' after 6 years of that, so I've since forgotten a lot... Still, the concept of feedback has proved useful in systems analysis from time to time, and it's a framework that I think could yield many low hanging fruit in other disciplines. (The book Behavior: The control of perception applies control theory to psychology in a convincing way but it's understandably been neglected by psychologists who aren't often very sophisticated mathematically.)