Realtime Spice?

[andy-cytomic]

This I think is one of the most detailed analog modelling plugin on the planet

I guess you're going to make a lot of friends with this sentence from people doing guitar amplifier simulations for a living for more than 9 years

Doing accurate tube amplifier models is very difficult, and I'm not trying to lessen their work at all, just pointing out that a little stomp box properly modelled is actually very difficult too!

A full tube amp emulation should only take an extra x1.5 to x2 cpu compared to The Scream. For example, using a Koren Tube model of a 6 tube (each triode counts as 2 tubes since it has two halves) amplifier like the Fender Deluxe will need around 2 log and 2 exp per tube, and either a sqrt or arctan for triode and pentodes respectively, so that's 6*4 = 24 calls to log/exp and 6 calls to either sqrt or arctan (which are a little easier to approximate and not as critical to the sound) and there are around 30 nodes (a sparse 30x30 matrix). In The Scream I'm doing 16 calls to log/exp and solving around 40 nodes (a sparse 40x40 matrix).

If anyone modelling guitar amplifiers takes affront at me saying The Scream is one of the most detailed analog modelling plugins on the planet then please speak up! To help you out in confirming such a statement below is the full schematic being solved in The Scream in realtime, along with the customised netlist I use (not exact spice syntax as I've added extra functionality, but it's pretty close). You can double check in trial version of The Scream that you can change the value of all the components listed, and also have high level control over the op-amp parameters like slew rate, open loop bandwidth etc. If anyone doing amp modelling cares to comment then please also provide the full schematic you are solving as well as its netlist so comparisons can be made.

https://cytomic.com/files/scream-schema ... etlist.pdf

(ps: did some minor typo correction as edits)
(pps: when I say 16 calls to log/exp I mean per iteration per sample, and there can be up to 12 iterations per sample in The Scream to converge to a decent solution. On average there are around 4 iterations per sample, but it depends on the incoming signal and knob settings)

[andy-cytomic]

Also by comparison LiveSpice solving a basic TS9 schematic using 3 exp, idealised op-amps and not even including the last buffer and around 20 nodes (a sparse 20x20 matrix) is reported to solve the circuit using x8 oversampling (352khz) takes around 53% cpu of an i5. The Scream at x8 oversampling (352khz) solving for x6 or more the level of complexity at least takes around 50% cpu of my i5 (including the up and down sampling cost).

[Urs]

This I think is one of the most detailed analog modelling plugin on the planet

I guess you're going to make a lot of friends with this sentence from people doing guitar amplifier simulations for a living for more than 9 years

In my experience, people who are upset about other people's breakthroughs are more than often upset because someone else did what they couldn't. Everyone else usually goes "welcome to the club!" when they see someone approach their level of knowledge.

[lalo]

https://github.com/dsharlet/LiveSPICE/issues/48

[Ivan_C]

I was just reacting to the vocabulary that you use to describe the technology in your products, and how you market it in general with comparisons to your competitors, not trying to lessen your work either, as I think you did a great job on the Screamer particularly.

Anyway, I'm looking forward to what you came up with for your next products related with analog pedals modeling. Specifically, I'd love to see a proper emulation of the Fuzz Factory pedal from Zvex, as it seems it challenges the laws of electronics and physics all by itself in the analog world

[andy-cytomic]

I was just reacting to the vocabulary that you use to describe the technology in your products, and how you market it in general with comparisons to your competitors, not trying to lessen your work either, as I think you did a great job on the Screamer particularly.

Anyway, I'm looking forward to what you came up with for your next products related with analog pedals modeling. Specifically, I'd love to see a proper emulation of the Fuzz Factory pedal from Zvex, as it seems it challenges the laws of electronics and physics all by itself in the analog world

I’ve got all the components for a fuzz face / factory and will be tackling that for my next stomp box plugin. Good news is it is a simpler circuit than The Scream, so will take less cpu

I’m very careful with what I say and I’m happy to back up and discuss in depth any statements I make. I do not indulge in hollow marketing hype, and I would actually like you to invite along some people that have been doing amp modelling for a long time to this thread so we can get down to business and discuss every detail of what is actually being modelled, as well as the strenghts and tradeoffs with various approaches.

[Vertion]

Like most folks reading this, I've invested a lot of money time and energy into software and music. Based upon Merit alone I would see that Andy is a very very polite and good person very humble and I would be the first to fight to say that. I think that instead of being envious or jealous of anybody skills the smartest people would actually seek to learn from those people instead of envying them being that knowledge can be accumulated that would be the smartest strategy for everybody here despite your Market situation. Andy is living the dream with great diligence and patience and leading in this market. Infinite kudos to Andy for his humility.

[Ichad.c]

I would like, at least, to discretize something as simple as a rectifier or something.

Assuming a solid state rectifier, it's pretty easy because a explicit solution exits (it's just some diode(s)).

https://en.wikipedia.org/wiki/Diode_modelling

For anything even remotely fun, you got to probably use Kirchhoff's Voltage and Current laws. And even those have their limitations.

https://en.wikipedia.org/wiki/Kirchhoff ... rcuit_laws