Salt - tech preview / test demo

[eidenk]

Doesn't load in Cantabile 1.2 and I get an "entry point not found - not a vst plugin" message. All your other plugs load OK with this host. Loads fine in Cantabile 2 Lite and savihost however.

MM.. probably missing the older pre-2.4 VST entry point then. I really thought I solved that problem once and for all.. but apparently not.

I've checked it vs your other plugs in Dependency Walker and Salt doesn't have the "main" export function that the others have and that appears to be the issue since Salt will load fine in Cantabile 1.2 after replacing "MAIN" with "main" in the plugin with an hex editor.

Very interesting plug btw.

[Ivan_C]

Very interesting project

There is just one thing I don't understand here. What is the use of compiling code, instead of applying a MNA on matrices with the right sizes in a C/C++ program ? Is it to optimize the running performance of the algorithm each time a new circuit is generated, like a kind of C++ template ? I don't know that much JIT compilers, maybe I need to look for good information about them... I hadn't even know that you could develop a VST plug-in able to generate and compile code on the fly ! Anyway, I know a little more circuit simulation, and I will gladly look for the next features of your project

[mystran]

There is just one thing I don't understand here. What is the use of compiling code, instead of applying a MNA on matrices with the right sizes ?

That's what it does: applies MNA by stamping stuff to a matrix. It's not "naive" about the process, but at high level it does use quite traditional MNA really.

Maybe at some point in the long-term I'll even expose enough internals that you can stamp in whatever you want. There's very few technical reasons not to do it, and there's nothing circuit specific at lower levels either.

There are some practical limitations though.. if you want to run in real-time.

Is it to optimize the running performance of the algorithm each time a new circuit is generated, like a kind of C++ template ?

Something like that. The "REF" solver is essentially an optimized "traditional" solver, while the point of the "JIT" is to squeeze all the remaining uncertainty out of the whole logic, to try and get the solution speed closer to raw floating-point output (and you can compare how much it currently helps). The actual algorithm that they use is pretty much the same.

[Ivan_C]

Thanks for the explanations

[andy-cytomic]

That's what it does: applies MNA by stamping stuff to a matrix. It's not "naive" about the process, but at high level it does use quite traditional MNA really.

You're using MNA so solve audio circuits and you expect it to work?!?!?!?! Time to invoke what Theo told me about this on the music-dsp email list (hehe)!

...
most of what you're oresenting is boring old crap, that isn't worth working on unless you'd actually understand some of the theory and relevant tunings involved. Clearly you don't.... (You) insulted run of the mill academic EE insights from decades ago,
...
So, at the risk of simply proving that some people are serious of wanting the crime of betraying the good and replace and sell it by making bad products
...
Don't you feel that if I'm simply right, and refuse to subscribe to the sort of "crime of century", that the insult is really the way I described ? Unless you say I must suffer be higher edified, to give you the privilege to mock about in the digital domain, which is like saying it's ok for your friends to steal all the good info, rewrite history, and you don't want to hear a word about it. Rather an authority to presume.

Boy, have we been told! Ok, back to your great work Mystran, and I look forward to a Mac version some time

(ps: if you want to read Theo's full rant for fun then please check out the thread "Trapezoidal integrated optimised SVF v2" here: http://music.columbia.edu/pipermail/mus ... hread.html )

(pps: for the record these circuit simulation methods do work and they also certainly make approximations when compared to physical components because we are dealing with idealised components mostly for efficiency. If you start with idealised components then you end up with an idealised model, that doesn't invalidate any of the maths, it sounds great and all works just fine, last time I checked my models weren't particularly "evil" or "betraying" anything.)

[aciddose]

The real-world electronics are really the ones betraying the ideals, aren't they?

Mystran: I have to mention the PNP thing again because of the horrible circuits that would become possible once a SCR can be used!

For example I have a fractional divider circuit that would be great. Known as a "cup and bucket" circuit in some places.

http://xhip.net/temp/fractional_divider.png

A lot of my favorite oscillator circuits rely upon an SCR.

Also I have many mini discrete PSU circuits that could be used to generate realistic power rails including all the noise and ripple you'd expect. The most simple has only three transistors - two npn, one pnp.

Would make it possible to construct accurate models of LM13700, for example, to be used in filters. Could also provide for accurate models of SSM and related filters.



Can we look forward to a version including PNP any time in the near future?

[camsr]

Can we expect inductors, transformers, and coupled inductors in the future?

Haha, me and aciddose are on the same train!

[aciddose]

Inductors are a bit different!

Actually I wonder if inductors would ever make sense. They are as far as I'm aware horribly complex to model accurately.

The models you could do already work using a gyrator circuit.

I've already played with a few EQs using this method.

[andy-cytomic]

Inductors are a bit different!

Actually I wonder if inductors would ever make sense. They are as far as I'm aware horribly complex to model accurately.

The models you could do already work using a gyrator circuit.

I've already played with a few EQs using this method.

Linear inductors and coupled inductors are straight forward enough, but once you get into accurately modelling the saturation of the core then it's a lot more difficult, you need to model the physical dimensions of the core using something like finite differences and also take into account the position of the windings around the core.

You can use a gyrator to hack a solution to a non-linear inductor which is a good first approximation. You can read about it here in the section "Modelling nonlinear capacitors and inductors": http://qucs.sourceforge.net/docs/equations.pdf

[andy-cytomic]

Can we look forward to a version including PNP any time in the near future?

As Mystran has already pointed out, once you can solve NPN solving the PNP is trivial, you just flip the equations, so I'm sure he will add it when he has time.

[andy-cytomic]

The real-world electronics are really the ones betraying the ideals, aren't they?

Those naughty real-world components, why can't they be perfect like all of us?

Also I have many mini discrete PSU circuits that could be used to generate realistic power rails including all the noise and ripple you'd expect. The most simple has only three transistors - two npn, one pnp.

Cool! I'd love to see some if you are willing to share them I've not been able to look into modelling power supplies yet, but I really want to in the future for accurately being able to FSU with the PSU

[aciddose]

Cool! I'd love to see some if you are willing to share them I've not been able to look into modelling power supplies yet, but I really want to in the future for accurately being able to FSU with the PSU

http://xhip.net/temp/most_horrible.png
http://xhip.net/temp/less_horrible.png
http://xhip.net/temp/less_horrible_2.png

Now keep in mind nobody ever said these are "not" horrible

Nothing will come close to a zener regulated supply using something like LM7xxx but you do actually see quite a few power supplies like these using transistors in circuits from the 70s. If you ever wanted to actually build it, remember the main transistor/diode needs to be a power device, but the others as far as I'm aware should be within small-signal limits.

The simpler circuits come in handy sometimes when you want to quickly get a roughly regulated voltage from a rail that is too high. Although of course once you figure out which voltage/current you need you can find the LM7xxx regulators in to-92 format.

So mostly these are just for fun when you want to provide a supply with an intentionally very large amount of ripple or poor regulation over varying current.

"most horrible" is the simple "lets see if we can make the magic smoke come out" implementation that gives you piss-poor regulation but roughly gets the job done.

"less horrible" is a better implementation using the same system only with proper feedback. Much better regulation although still awful

"less horrible 2" is the same only with current limiting because we were running out of magic smoke.

[andy-cytomic]

Cool! I'd love to see some if you are willing to share them I've not been able to look into modelling power supplies yet, but I really want to in the future for accurately being able to FSU with the PSU

http://xhip.net/temp/most_horrible.png
http://xhip.net/temp/less_horrible.png
http://xhip.net/temp/less_horrible_2.png

Now keep in mind nobody ever said these are "not" horrible

Thanks very much for these I've got plenty of magic smoke, and my transistors can run at several million degrees no problem, I've not blown one yet

Can you please let me know which gui schematic program you're using? The Qucs symbols all look pretty naff and I'll like to be able to post nicer looking schematics without having to use latex stuff which takes forever.

[camsr]

Inductors are a bit different!

Actually I wonder if inductors would ever make sense. They are as far as I'm aware horribly complex to model accurately.

The models you could do already work using a gyrator circuit.

I've already played with a few EQs using this method.

I am using an active filter schematic for a lowcut, it's supposed to use a gyrator. More passive filters would be great though.

[aciddose]

Thanks very much for these I've got plenty of magic smoke, and my transistors can run at several million degrees no problem, I've not blown one yet

Can you please let me know which gui schematic program you're using? The Qucs symbols all look pretty naff and I'll like to be able to post nicer looking schematics without having to use latex stuff which takes forever.

I'm using a free app I grabbed years ago I guess around 2002.. "Circuit maker student edition"... pretty much the basic functions of the app with a 50 component limit or something like that. Pretty hard to ever run into that limit.

Horribly old, I just haven't ever had a reason to upgrade since most of the others I've tried are actually worse rather than better. I do sometimes write netlists by hand to run in better simulators, but for basic mucking around with schematics I find this works fairly well.

So unfortunately I'm definitely not the person to ask about where to find something good.