Download a Guitar Tube Amp Simulation Plug

[dscro]

My first attempt at a plug like this.

Thanks to metal_head_82 for the inspiration. And thanks to Norman Koren for his tube-models-for-pSpice page. This SB plug uses the math from his excellent model for a 12AX7 triode. The audio input in the plug supplies the grid voltage for the tube model. The plate voltage is 350 Volts which I think was pretty standard for these triodes.

The controls in the SB plug are self explanatory. But the first time you try it, play with the the Tube Pregain, the Output Saturation, and the Output gain before you change any of the filter values.

The Pregain boosts the amplitude of the signal to the tube's grid and this is where you'll hear the 12AX7 model doing its thing.

The output saturation boosts the level (after the tube circuit) going into a soft clipper (the usual hyperbolic tangent function everyone uses). I included this originally so that the output wouldn't ever go above digital 0 in case the tube model acted funky. Driving it hard is a very different sound than driving the tube circuit hard. Different combinations sound different.

After the tanh clipper there's an output gain so that you could, after all, boost the signal without driving the clipper. But BE CAREFUL. This output isn't limited.

Finally, the audio is 2X oversampled (hence all the cpu usage). All the math is done at 4x the sample rate and then then filtered just below 20k before returning to the original sample rate. This made a difference in the sound, I think because it removes the aliasing from frequencies above 20kHz that are generated by the math that mangles the waveform.

The link below is just to the raw SB plug. I exported it as an AU on my system and it worked great in Logic 8. But, as we all know, the land of digital audio is an uncertain place. Use this carefully. Although it's been fine for me, other SB plugs have occasionally given me loud clicks and pops when audio starts or stops. Just protect your monitors.

ps: if anyone has any impulse responses for speaker cabinets we could add a convolution section to the plug.


OK here's the link. Download the SB file "Tube1.sbc" when you get there.

http://dscrofani.junker.org

[metal_head_82]

Thx to you!
Sounds very good.
I'm currently working on a different approach.
I've found another formula to calculate the tube.
As soon as it works I'll post it here.

[zmix]

Very interesting... Norman Koren sold me some of his darkroom equipment a few years ago...We've had a number of chats about amplifiers too....

[dscro]

Is the world really that small?

[zmix]

Oh, yes... and it get's smaller... Norman Koren cites my friend Mithat Konar's work on computer tube models:

http://www.birotechnology.com/articles/VTspice.html

Mithat designed my main studio monitors...

.. and we have performed together in a hybrid guitar / electronic project called "House your Mother"

[zmix]

I've been playing with tis 12ax7 model a fair amount...

I wonder if the values should be optimized for a 32 bit floating point environment... the plate voltage is set to 350v but this is far in excess of the maximum plate voltage for an actual 12ax7. It does have affect the absolute clipping level of the model and it's important to avoid this region as the model behaves very un-tube-like there.

David, would you care to share some details about the model? The voltage scaling factors and signum function for example?

Thanks

[dscro]

Sure. I'm not certain there's much to share, though. The model literally uses Norman Koren's equations (4) from page

http://www.normankoren.com/Audio/Tubemo ... ticle.html

A little further down the page he gives some parameter values for common tubes including the 12AX7. I used those in the SB model as well for the voltage scaling factors and such.

Signum is there because if grid voltage becomes greater than plate voltage, electrons would stop flowing from grid to plate (no plate current), since the electrons are "flowing uphill" from lower to higher potential in the tube when plate is at lower potential than grid. So signum forces the plate current to be zero if relative potential goes negative.

The 350 volts was my own choice. The audio input signal to the tube model is pre-gained and fed into the equation as the grid voltage (Eg in Norman's equation 4). Is that where audio input to the real tubes went . . . grid potential?

I agree that it doesn't sound very tube-like when pushed hard. Hmmm

[zmix]

David,
I've been playing with Norman's equations and I have found some parameter settings that allow the simulation to respond in a much more 'musical' way. No hard clipping and a good assortment of odd and even harmonics. After tuning the circuit by ear, the values are a bit off from the 'textbook' 12ax7 but it's got unity gain and zero DC in the output!

[zmix]

Still not getting the amount of dynamic compression I am used to from my real 12ax7 circuits, but close...

[dscro]

That's good news. Keep me posted. I can't wait to hear it.

[Laribum]

Thanks dscro :!:

This is very cool! I am currently playing around with it. I downloaded all documents that you've stated as well as some more to read more about the topic.
I tried the triode model, that you defined with some audio signals and I like it quite a lot.
This will come in handy in designing some nice analog components

If you guys have news on some better definitions for the values, I would defnitely be interested in your results I will post my news regarding it here as well.

cheers,
laribum

[dscro]

Great. Yeah, it's a work in progress. But I think if we all chip away at it we may end up with some awesome stuff in the end. Keep me posted.

[butter]

Mac only...damn. Probably about the only time I'll ever say that.

[ew]

I wonder if the values should be optimized for a 32 bit floating point environment... the plate voltage is set to 350v but this is far in excess of the maximum plate voltage for an actual 12ax7. It does have affect the absolute clipping level of the model and it's important to avoid this region as the model behaves very un-tube-like there.

Not true in actuality, but unfortunately true in most 12AX7 circuits these days. A 12AX7 can take plate voltages up to about 450 volts in a properly designed circuit. Most circuits these days are designed with lower plate voltage ranges- they'll clip earlier and won't be as dynamic as they could be.

ew

[zmix]

I wonder if the values should be optimized for a 32 bit floating point environment... the plate voltage is set to 350v but this is far in excess of the maximum plate voltage for an actual 12ax7. It does have affect the absolute clipping level of the model and it's important to avoid this region as the model behaves very un-tube-like there.

Not true in actuality, but unfortunately true in most 12AX7 circuits these days. A 12AX7 can take plate voltages up to about 450 volts in a properly designed circuit. Most circuits these days are designed with lower plate voltage ranges- they'll clip earlier and won't be as dynamic as they could be.

ew

EW,

Can you please provide references for your claim that a 12ax7 can take 450v at the plate? Even if the current is controlled, it is quite possible that the tube element's coatings will become depleted under such conditions.