Add `smooth` sources in `Blocks` + Add `Square` and `Triangular` sources #59

ven-k · 2022-05-22T18:17:44Z

Adds smooth option for sources in Blocks. For Sine, Cosine, ExpSine, Step, Ramp
Adds Square and Triangular sources

baggepinnen · 2022-05-22T18:47:12Z

src/Blocks/sources.jl

+    equation = if smooth == false
+        offset + ifelse(t < start_time, 0, amplitude* sin(2*pi*frequency*(t - start_time) + phase))
+    else
+        δ = 1e-5


Should this parameter be exposed and documented?

I think it should

Perhaps yes. Although having δ = 1e-6 or such makes a miniscule change to the result, it can upset high precision tests

ValentinKaisermayer · 2022-05-22T18:58:12Z

src/Blocks/sources.jl

+    equation = if smooth == false
+        offset + ifelse(t < start_time, 0, amplitude* sin(2*pi*frequency*(t - start_time) + phase))
+    else
+        δ = 1e-5


I think it should

ValentinKaisermayer · 2022-05-22T18:58:40Z

src/Blocks/sources.jl

+    end
+
+    eqs = [
+        output.u ~ smooth_square(t, δ, frequency, amplitude, offset, start_time)


Is there no exact version?

ValentinKaisermayer · 2022-05-22T18:59:04Z

src/Blocks/sources.jl

    eqs = [
-        output.u ~ offset + ifelse(t < start_time, 0, amplitude * exp(-damping * (t - start_time)) * sin(2*pi*frequency*(t - start_time) + phase))
+        output.u ~ smooth_triangular(t, δ, frequency, amplitude, offset, start_time)


Is there no exact version?

When I had tried the solvers became unstable. Let me give another shot at it

I would leave it in. It is an upstream problem than and not of the library. Also add tests but mark it as broken.

ven-k · 2022-05-22T19:05:58Z

Triangular and Square sources now support start-time and offset.

So these are no longer the waveforms

If Triangular is going to be only smooth, I'm thinking, we should remove start_time and offset options

start_time shifts the wave but value isn't non-zero for t<st (Might have to resort to IfElse)
When offset=0, start_time=1, amplitude=1, f=5

Offset makes it very erratic
When offset=1, start_time=1, amplitude=1, f=5

ven-k · 2022-05-22T19:06:29Z

So this is no longer the waveform

Square Wave

ValentinKaisermayer · 2022-05-22T20:24:23Z

Is this an aliasing problem or is the implementation incorrect?

baggepinnen · 2022-05-22T21:14:57Z

One option that can be provided alongside the smoothing option is to use callbacks to make sure the solver steps exactly at the discontinuities (in value or derivative etc.) of the signal. The symbolic callbacks should be able to handle these signals, but discrete callbacks may be more efficient given that we know the time points of the events beforehand.

ven-k · 2022-05-29T12:36:55Z

For frequency=5, amplitude=1, offset = 2, start_time = 1

Smooth triangular:

Triangular

For frequency=1, amplitude=2, offset=1, start_time=2.5

Smooth Square:

Square:

ValentinKaisermayer · 2022-05-29T15:41:17Z

Cool!
The smooth square and exact square should have the same phase, for the same parameters, though.

ven-k · 2022-05-29T16:09:22Z

The smooth square and exact square should have the same phase

Yeah I had noted that. The last commit, I was working on, addresses exactly that. (I've updated the comment above with latest waveform)

ven-k · 2022-05-29T16:32:07Z

One option that can be provided alongside the smoothing option is to use callbacks

I will add this with a follow up PR

ven-k · 2022-05-30T09:05:06Z

The thermal.jl tests are no longer broken. See the logs here
([email protected] is used)

So can I go ahead & change @test_broken -> @test ?
cc @YingboMa

ValentinKaisermayer · 2022-05-30T10:51:41Z

If you do, thx. But in a separate PR. 😄

ven-k · 2022-05-30T12:33:20Z

Ok.
Tests of this PR pass. And this is ready

ChrisRackauckas · 2022-05-30T16:31:22Z

It looks like tests fail. Change test broken to test.

- For `Sine`, `Cosine`, `ExpSine`, `Step`, `Ramp` Add Square and Triangular sources Add a non-zero start_time to smooth Sine, Ramp, ExpSine tests & non-zero `offset` to smooth ExpSine Add "almost" exact versions of Triangular and Square waves The formulae from https://en.wikipedia.org/wiki/Square_wave & https://en.wikipedia.org/wiki/Triangle_wave are modified to accomodate any amplitude, frequency, offset and start_time translate t on x-axis with start_time for square and triangular waves trigger workflow Merge branch 'vk/smooth_blocks' of https://github.com/ven-k/ModelingToolkitStandardLibrary.jl into vk/smooth_blocks thermal tests are no longer broken; so `@test_broken` -> `@test`

codecov · 2022-05-30T17:36:57Z

Codecov Report

Merging #59 (883baa8) into main (a2d79c9) will increase coverage by 1.65%.
The diff coverage is 98.48%.

@@            Coverage Diff             @@
##             main      #59      +/-   ##
==========================================
+ Coverage   70.12%   71.77%   +1.65%     
==========================================
  Files          23       23              
  Lines         964     1024      +60     
==========================================
+ Hits          676      735      +59     
- Misses        288      289       +1

Impacted Files	Coverage Δ
src/Blocks/sources.jl	`98.94% <98.48%> (-1.06%)`	⬇️

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

ValentinKaisermayer · 2022-05-30T19:12:58Z

src/Blocks/sources.jl

+    offset=0.0, start_time=0.0, smooth=false)
+
+    if smooth
+        frequency > 25 && @warn "`frequency > 25` can lead to non-triangular wave pattern"


What's up with this?

Perhaps the smoothing constants need to be frequency dependent?

For f>25, the waveform is totally erratic

Perhaps the smoothing constants need to be frequency dependent?

I tried these:

amplitude * (1-2acos((1 - f* δ)sin(2π*(x - start_time)*f))/π) amplitude * (1-2acos((1 - δ/f)sin(2π*(x - start_time)*f))/π) amplitude * (1-2acos((1 - δ)sin(2π*(x - start_time)*f*δ))/π) amplitude * (1-2acos((1 - δ)sin(2π*(x - start_time)*f/δ))/π) amplitude * (1-2acos(sin(2π*(x - start_time)*f)/π)

None of these relations gave a better result.
I think it comes down to trigonometric functions used. Is there any source I can look at for this?

I would imagine that the delta parameter influences how high frequencies can be tolerated?

Removing that didn't make any difference (formula 5 above)

The above image is output of just smooth_triangular

This is still open which is why I didn't merge. What's the consensus here? I didn't look into it.

Aha, the real issue is the stepsize. For <0.001 the plot looks alright. And this is present in both smooth=true/false wave.

for t = 0:0.001:1

Should I mention a note on this in docstring?

(Will remove the @warn)

Adaptivity should handle it?

Somehow it didn't. For offset!=0 similar issue crops up. saveat=0.01 saves. (Here)

ChrisRackauckas · 2022-05-31T03:34:40Z

src/Blocks/sources.jl

+@register_symbolic smooth_cos(x, δ, f, amplitude, ϕ, offset, start_time)
+@register_symbolic smooth_damped_sin(x, δ, f, amplitude, damping, ϕ, offset, start_time)
+@register_symbolic smooth_ramp(x, δ, height, duration, offset, start_time)
+@register_symbolic smooth_sin(x, δ, f, amplitude, ϕ, offset, start_time)
+@register_symbolic smooth_square(x, δ, f, amplitude, offset, start_time)
+@register_symbolic smooth_step(x, δ, height, offset, start_time)
+@register_symbolic smooth_triangular(x, δ, f, amplitude, offset, start_time)
+@register_symbolic triangular(x, f, amplitude, offset, start_time)
+@register_symbolic square(x, f, amplitude, offset, start_time)


Why register these?

Earlier we didn't assign parameters while declaring them inside components. So @register was necessary.

But yeah, now its no longer required. Removed

baggepinnen reviewed May 22, 2022

View reviewed changes

ValentinKaisermayer reviewed May 22, 2022

View reviewed changes

ven-k marked this pull request as draft May 22, 2022 19:08

ven-k marked this pull request as ready for review May 29, 2022 16:08

ven-k mentioned this pull request May 29, 2022

Define Voltage, Current #51

Merged

ven-k force-pushed the vk/smooth_blocks branch from e651f5a to 0f2d737 Compare May 30, 2022 17:30

ValentinKaisermayer reviewed May 30, 2022

View reviewed changes

ChrisRackauckas reviewed May 31, 2022

View reviewed changes

Remove @register_symbolic

55668b1

ven-k requested a review from ChrisRackauckas June 3, 2022 07:32

Remove @warn in triangular source

883baa8

ChrisRackauckas merged commit 9a5d70d into SciML:main Jun 3, 2022

Uh oh!

Add smooth sources in Blocks + Add Square and Triangular sources #59

Add smooth sources in Blocks + Add Square and Triangular sources #59

Uh oh!

Conversation

ven-k commented May 22, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ven-k commented May 22, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ven-k commented May 22, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ValentinKaisermayer commented May 22, 2022

Uh oh!

baggepinnen commented May 22, 2022

Uh oh!

ven-k commented May 29, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ValentinKaisermayer commented May 29, 2022

Uh oh!

ven-k commented May 29, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ven-k commented May 29, 2022

Uh oh!

ven-k commented May 30, 2022

Uh oh!

ValentinKaisermayer commented May 30, 2022

Uh oh!

ven-k commented May 30, 2022

Uh oh!

ChrisRackauckas commented May 30, 2022

Uh oh!

codecov bot commented May 30, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ven-k Jun 3, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ven-k Jun 3, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Add `smooth` sources in `Blocks` + Add `Square` and `Triangular` sources #59

Add `smooth` sources in `Blocks` + Add `Square` and `Triangular` sources #59

ven-k commented May 22, 2022 •

edited

Loading

ven-k commented May 22, 2022 •

edited

Loading

ven-k commented May 29, 2022 •

edited

Loading

ven-k commented May 29, 2022 •

edited

Loading

codecov bot commented May 30, 2022 •

edited

Loading

ven-k Jun 3, 2022 •

edited

Loading

ven-k Jun 3, 2022 •

edited

Loading