Real Time SPICE Simulation

[jazbo8]

An exciting take on SPICE simulation, more tube models will be added soon. It is already pretty good for modeling and hearing preamp distortion in real time. You can download the program here.

Please post your feedback to the developer here.

[Cliff Schecht]

NI MultiSim has been able to do this for a while. I've streamed audio into it, manipulated said audio (in this case was a bunch of tweakable filters) and listened to the results whilst viewing them with an virtual oscope. Of course being NI, it's all Labview based. One other positive thing to be said about MultiSim is their extensive library, literally the most extensive and usable component library I've ever dealt with.

Not to downplay what this fella has done, getting this stuff to converge with a non-optimized algorithm can be a nightmare! May have to give it a shot soon..

[jazbo8]

NI MultiSim has been able to do this for a while. I've streamed audio into it, manipulated said audio (in this case was a bunch of tweakable filters) and listened to the results whilst viewing them with an virtual oscope.

I was not aware that any simulation package can do that... Can you play the guitar in real time? i.e., not stream a pre-recorded audio track, run SPICE then play it back...

[capgreed]

hmm ..... just read the website ....potential very complex circuit and over simplified component mathematical models seem to be an invitation for "open bar" propagation of errors ..... especially when it comes to oscillations and signals in general..... just saying IF he solved this issue , i must say i'd be VERY curious to know how he dealt with it in various cases

exciting yes , well prospectively , let say in then years when the processors and gpus will allow the guy to trash his old less calculus intensive component models for more accurate ones.... His intention is good though

[jazbo8]

hmm ..... just read the website ....potential very complex circuit and over simplified component mathematical models seem to be an invitation for "open bar" propagation of errors .....

It all comes down the math models and CPU processing power, in fact, the existing HW & SW tube amp emulators already do a decent job, LIVESPICE is a geeky EE version without the nice UI, or the bells and whistles of the amp emulators.

[Cliff Schecht]

NI MultiSim has been able to do this for a while. I've streamed audio into it, manipulated said audio (in this case was a bunch of tweakable filters) and listened to the results whilst viewing them with an virtual oscope.

I was not aware that any simulation package can do that... Can you play the guitar in real time? i.e., not stream a pre-recorded audio track, run SPICE then play it back...

Yes, it can. Does a decent job too and is always improving. The problem with MultiSim is that the tube models as well as the engine itself are not optimized to handle the crazy non-linearities that occur when you distort a tube. The models work OK until you start trying to make them clip, then the whole mess can't/won't converge with the factory setup. It craps a brick with when you try to do switchers too as it really isn't setup to handle some types non-linear circuits very well. I didn't dig too far into how to optimize this but I bet that one could achieve better results with more accurate and thorough models (although this will be very computationally taxing if done in real-time with high order approximations!).

[capgreed]

NI MultiSim has been able to do this for a while. I've streamed audio into it, manipulated said audio (in this case was a bunch of tweakable filters) and listened to the results whilst viewing them with an virtual oscope.

I was not aware that any simulation package can do that... Can you play the guitar in real time? i.e., not stream a pre-recorded audio track, run SPICE then play it back...

Yes, it can. Does a decent job too and is always improving. The problem with MultiSim is that the tube models as well as the engine itself are not optimized to handle the crazy non-linearities that occur when you distort a tube. The models work OK until you start trying to make them clip, then the whole mess can't/won't converge with the factory setup. It craps a brick with when you try to do switchers too as it really isn't setup to handle some types non-linear circuits very well. I didn't dig too far into how to optimize this but I bet that one could achieve better results with more accurate and thorough models (although this will be very computationally taxing if done in real-time with high order approximations!).

there are mathematical tricks to keep the computational cost within range of nowadays stations while maintaining accuracy on non linear equations ... Like TVD or TVNI finite differences schemes in the case of hyperbolic conservation laws , you can keep it to the first or second order schemes including the appropriate flux functions , does wonders....Still the models ( equations ) DO have to be the right ones with all the dirty ( meaning accurate ) coefficients and high order partial derivatives ....you also have to keep in mind some highly non linear D.Es also have analytical solutions , which reduces the costs of dynamic resolutions , i think the big players cannot permit themselves to exclude the option to hire real mathematicians when they dip their fingers into this kind of signal mess ....

[VacuumVoodoo]

Switch states and discontinuous non-linearities are handled pretty well by Gear integration algorithm originally developed by Boeing for their Easy-5 dynamic simulation SW package. It's now under the MSC Software umbrella.
Gear integration algorithm is also used in SMPS simulation SW such as SIMetrix. Perhaps it could be applied here as well if it can run fast enough for RT.

[Pacosipulami]

Hi all

LiveSpice got updated recentely and I tested it on my crappy core 2 duo machine with one of my tube overdrive pedal designs (3 high voltage triodes cascaded) into a clean fender princeton type amplifier in real time (about 5ms RTL)

https://www.youtube.com/watch?v=TTOPCtSGTMw

I was so excited so about the results, so I bought a better computer (i7-10700K CPU / 10th gen intel 8 core with 3.8GHz base clock speed), Asus Z490Prime motherboard, 2933MHz DDR-4 ram and a ultra silent case/PSU, BlackRock Cooling system, all optimized for best audio performance to simulate circuits up to 5 triodes in series, which is running pretty stable so far. Latency is around 4 - 4.6ms roundtrip with a 3rd gen Focusrite Scarlett interface....

before and beside LiveSpice, I use Spectrum's Microcap 12 - which is freeware since July 2019, when the company decided to give it away for free because the developer got retired after 39 years.
If you don't know it yet, get it for free here: http://www.spectrum-soft.com/index.shtm