It is not a one trick pony if you know how to set up the equations and let the computer do the analysis. The danger of relying on the computer is when you do not understand the underlying theory.
I think I would take KA's opinion on (electronic) things over most other people. (You are high on the list to ... to be clear)
I don't mean that he isn't very good at his one trick, which he clearly is. When it comes to analysis, though, he just puts up straw men and knocks them down.
There's a lot more to instrument-building than circuits.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
Well, yeah I can't confirm the correct operation of a SMPS design by hand. Or most three-transistor feedback amplifier topologies for that matter.
I'm pretty fast with a handheld calculator these days! But of course I use software to design stuff for me all the time; I don't grunge through e.g. LC filter tables in a book and de-normalize them on paper we have software for that.
I don't know how to come up with plausible values for some random switching converter feedback loop that stabilizes it for a particular application just by looking at it I'm sorry. I can't read its mind. Someone who does that task day in and day out for the same converter for similar applications may be able to and I guess if that's the job that's the guy you'd hire.
It was enormously valuable to be forced to do the math in EE school. Circuit theory, units analysis, signals and systems, filtering and convolution, electromagnetics, comm theory, physics, even electrical machinery. (I cut the EE labs and faked the reports. My favorite course was Beginners Tumbling) But I don't do the heavy symbolic math any more... maybe a little algebra, and the occasional square root. Simulation is better, because the real world is mixed-signal nonlinear. Try sticking a quantizer into the middle of a differential equation.
Spice doesn't do noise well, but we usually have a lot of signal. Or a Dremel.
Equations are text, and I prefer pictures, waveforms and schematics. Some people, like you, can "see" the living equations, but I don't.
French and English almost flunked me out of high school. Words.
--
John Larkin Highland Technology, Inc
The best designs are necessarily accidental.
Today, that's all that is required. A good understanding of generalities and drive the computer. 60 years ago, one might well have had to know a lot more detail.
Spice does standard circuit noise such as shot noise, 1/f and thermal noise extremely well. Periodic steady state with phase noise extensions are also very good today. It allows the phase noise of any nonlinear system to be predicted quite accurately.
The thrust of this thread is calculate periodic switch on process, with the questioner making the common assumption that some drawn out mathematically process might be required. This question has been specifically addressed.
I have explained that solving equations analytically is essentially due to legacy historical lack of computers. Unfortunately there is often an emphasis on academic approaches to solving engineering problems. Its all bollocks. People need to move on.
I have addressed the elephant in the room, that is, despite ones ego in that some more detailed "understanding" of a problem would be appropriate, it just isn't true in reality.
I am making statements as to how it is in the real world with regarding to solving complex problems. Its all done in the virtual universe That is, trillions of chips resulting from millions of different products using 10s to billions of transistors in ASICs are all done entirely on computers by
100,000s of engineers in all the major companies.
Its not only in Spice. For example. I have done all the mathematical work for linearising varactors by constructing chebychev functions and inverse functions via chebychev functions entirely in Excel using its inbuilt non-linear solver. No detailed knowledge of approximation theory, non-linear solvers, orthogonal functions, or physics was required.
Thermal design of the packaging of the products I are am designing is all done in COMSOL. There is absolutely nothing to be gained by trying to say, manually modelling heat flow in a pin, then through to the board and the ASIC etc. Its simply impossible to get the understanding that one achieves by say, having a colour flow map of the heat distribution across the chip. In my case, an xtal needs a flat temperature profile to minimise stress that effects its tuning. Again, simply impossible to do this sort of analysis in any useful way without simulation.
Todays real world problems simply CANNOT be solved without computers. Thus, they should be used from the outset, essentially, exclusively. Its just delusional and pissing in the winds to pretend that they is some some value of an analytical approximation that spans 5 pages of A4.
Try running a bank without a computer. This is 2020, not 1930.
Sure, but not relevant to this thread. "Procedure for inverse Laplace transformation".
I've built my own time-domain noise generators in Spice. Is there a better way to do that? Where do you get Johnson and shot noise sources in time domain?
--
John Larkin Highland Technology, Inc
The best designs are necessarily accidental.
Looks like someone else posted a method in LTSpice. Standard spice doesn't directly support transient noise. One would have to connect a large pwl source to every noise terminal.
In my Cadence ASIC universe, transient noise is directly supported, however, PSS & PSS Noise is also supported, which is much more efficient.
A key part of Periodic Steady State capability is that it directly handles the fact that the "instantaneous" noise depends on the instantaneous operating point (current & voltage) as the large signal moves about.
Sure, I can build reasonable noise generators from lots of parts. But it isn't appealing to build 40 of them, one for every part on the screen. Not to mention the insides of models, where I can't reach.
Is a time domain simulation properly noisy everywhere?
How about shot noise? I gueees we use that selectively, so can make it ourselves occasionally.
--
John Larkin Highland Technology, Inc
The best designs are necessarily accidental.
Leaves even specialist numerical integration standing in terms of convenience. It even found faults in the symbolic algebra code generator for usage of continuation cards when N>9. The expressions are brutal and horrible but it is the goto method for precision ephemerides today.
You do need a way to tell that your simulations get the analytic simplest testable cases right though. Too many people blindly rely on simulations with no idea if the time step is appropriate or if it has really converged. Computer says yes/no/maybe delete as appropriate.
Monte carlo simulations certainly have their place. I quite like simulated annealing as a way to find a workable global optimum. Not necessarily the true global optimum but good enough for most purposes.
I don't disagree with you there. But there is still a lot to be said for the elegant analytic closed forms for certain types of network. Sometimes inelegant semi-analytic closed forms can work well too.
The right solution depends a lot on the problem you are trying to solve.
Not quite true. Fresnel got the prize for predicting against all the odds that there would be a bright diffraction spot on a screen exactly diametrically opposed to the source of the illumination.
One of the ten greatest theoretical predictions of all time according to Physics World. It was experimentally verified by Arago to settle the dispute. First nail in the coffin of Newton's corpuscular theory of light Poisson was not amused!
Title "The 10 greatest predictions in physics" in case the link fails.
A friend observing the passage of Titan in front of 28 Sagittarius observed the same diffraction effect except that the central spot was much brighter than theoretical predictions which was a very strong hint that Titan had a non-trivial atmosphere.
I know all about Poisson's spot. ('T weren't Fresnel that predicted it, it was Poisson trying unsuccessfully to rubbish him.) If the diameter of the shiny sphere is important, its scattering isn't isotropic.
The shiny-sphere theorem is a ray optics thing, and so valid in the limit k-> infinity.
Sure comes in handy sometimes in photon budgets. :)
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.