Transfer functions

PSIM used to have a (limited) demo available. Google/wayback may be able to find it?

There is a tool for doing this called Spice, which has been around for some fifty years. There's a free, and popular version of this - LTSpice

- which was first released in 1999 - which Edward Rawde has been using for years.

The "transfer function" tells you happens at an output when a stimulus is applied to an input.

Circuit simulation generalises this a bit. Spice records the state of every node in the circuit simulated for the whole length of the simulation (unless you tell it not to).

My post wasn't directed ay you Bill. It's only for people who know how to use a computer.

But it does obtain the transfer function from the schematic, which is what I wanted. And gets much faster when you've been throug it once.

This?

Can you tell me how to do something like this in LTSpice please Bill?

import sympy as sp

# Define symbolic variables s, t = sp.symbols('s t')

# Define your transfer function G(s) # Example: G(s) = 1 / (s^2 + 3s + 2)

G = 1 / (s**2 + 3*s + 2)

# Compute the inverse Laplace transform

g_t = sp.inverse_laplace_transform(G, s, t)

print(f"Time-domain function g(t): {g_t}")

Roberts has been publishing about SPICE for over 30 years. This latest text,which is free, is centered on LTSPICE. You are interested in Chapters 9 and 10.

Anything more complicated is probably not needed. You can always use Mason's Rule(s) to consolidate multi-multi- multi- signal flow graphs, but most of that doesn't even apply to practical circuits.

Entertaining qualification. What did you think you had to say to them?

Edward assesses his own skills correctly.

What might you mean by "sympy?

LTSpice has its own list of symbolic variables. You don't get to define time for yourself - you have to use their built-in variable.

The does look like a Laplace transform. You can pull them out of Williams and Taylor's "Electronic Filter Design Handbook" not to mention any number of more obscure textbooks.

That's what Williams and Taylor exists to let you find.

Getting trained in the relevant math is another option. Expecting a computer to do it for you is a trifle unrealistic.

Only those who have attended Hogwarts School of Witchcraft and Wizardry would know that.

Sympy is an adjective that means "associated with symps."

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Transfer functions are kind of old-fashioned. They are an invitation to do a lot of ugly algebra. And anything interesting is nonlinear.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Yes that's true, but when you do want to play with transfer functions and when they get more complex than anything you might find in a textbook, a computer algebra package makes it easy to do that which would otherwise be very error prone ugly algebra and take a long time.

So your choices would be to leave the algebra to Bill (since he has already memorised his Laplace transform table) or get a computer to do it. I know which I'd choose.

Lots of the perfectly linear stuff can be quite interesting, if unrealistic.

Far from memorising the Laplace transform tables, I took the advice of another engineer, who told me to buy a copy of Williams and Taylor who present a bunch of interesting Laplace transforms and document the consequences.

If you want frequency selective filters, that's a good place to get them.

None of the computer programs I've run into have been all that helpful. The mathematicians who put them together don't seem to know enough about circuit design to ask the users the right questions.

That's probably because you know nothing about python.

If you want software to do any kind of actual circuit design for you then good luck with that.

Here's the canonical reference:

Note that it will build the phase/magnitude plot by iterating through the simulation so it can take some time (depending on how finely you want the iteration to explore the frequency domain).

MatLAB is probably the go-to tool for this sort of thing. Octave being a FOSS competitor. But, the schematic requirement may complicate that.

Your SapWin tool may, in the long run, be the better approach. Maybe take on a more active "contributor" role to help them along?

There are a bunch of electronic-design AI startups here. We get their pitches at maker-space meetups; I expect more tonight.

I haven't been impressed so far. At least the beer and pizza are free.

Allspice Ai doesn't design boards; it just reviews them. That might be useful.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

I've used it before, but not do do anything python can't do now.

Also Mathcad, which still seems to still exist if you have $800/year to spare.

I don't know the extent to which python can do what Matlab can do but it must be starting to take some work away from Matlab.

In theory, any turing complete language can do anything that any other can. It's just a matter of style (and support, etc.) You'd have to look at the support libraries/frameworks available in each camp to decide where you have the best "starting point".

I am lucky (?) in that I made most of my software acquisitions under the "old" licensing style. Admittedly, older versions (how many truly significant features am I missing out on?) but I don't have to "rent" anything -- which would be REALLY annoying if I just needed to use something for a few weeks, or poke around with it periodically.

The issue I have with all FOSS "products" is that they are more "projects" than "products". No one takes ownership in the same way that a commercial vendor is required to. (PostgreSQL is a notable exception -- but, really for their Enterprise offering and for obvious reasons!)

With more than 30 years in the FOSS world (since ~1992), I've enough first-hand experience to know that "enough eyes" looking at a product does NOT ensure it is more robust or reliable than a closed product would be. Witness the number of bugs that have hidden in plain sight for years: "We also used KLEE as a bug finding tool, applying it to 452 applications (over 430K total lines of code), where it found 56 serious bugs, including three in COREUTILS that had been missed for over 15 years. Finally, we used KLEE to crosscheck purportedly identical BUSYBOX and COREUTILS utilities, finding functional correctness errors and a myriad of inconsistencies." A commercial product risks losing business if it fails to support their product(s) adequately.

If I opt to use a FOSS tool, then I make a commitment to supporting it, myself -- at least to the extent to which

I just had a look. The state-variable information from page 121 is interesting but it doesn't include some of the components in the circuit I posted. Probably because it would then start to get hard to fit the transfer function across one page.

It doesn't seem to include state-variable but this page is useful for fiter design