Experimenting with Amp Sim Algorithms

[ScootBaloo]

I am experimenting with Amp Sim algorithms based on the 12AX7 tube and have tried a few different algorithms so far. My approach is to simulate the pre amp and then send the simulated pre amp signal to the power amp of my 5150 and record the sound from my Mesa2x12 with a SM57.

So far the amp sim sounds have not been that great so I need to keep working on different ideas but I thought I might post a few of the better experiments that I did.

In this clip ( Ola's DI's ), the bass DI and the guitar DI signal were reamped through the amp sim and recorded using a SM57. The mix has too much bass and the drum sound is not great but you can hear the amp sim sound. It seems to my ear that the frequencies in the distortion are a little odd, does any one else think so ?

http://dl.dropbox.com/u/2785995/AmpSimTest-1-1.mp3

EDIT : In case it was not clear, yes I am writing my own amp sim.

 

[Ascendant]

Doesn't sound that bad, it's a good start if you're just getting into development, lacking a lot of gain for metal though. I think there's some aliasing going on or some noise in the high end, are you using any oversampling? Any info on how you built the sim? Gain stages tonestack etc.

The mix is mostly lacking power because of the drums and bass being so low but I suppose that was intentional to make the guitars more audible.

 

[ScootBaloo]

Thanks for having a listen and making a comment.

I am just starting development so I dont have a final design yet. At the moment, I have coded a circuit simulation
of a single 12AX7 stage. I have implemented the R's and C's using Z transforms ( a HPF and a LPF ) and I have
simulated the tube from the V-I characteristic curves published for the 12AX7. For each input sample I calculate the
output using an iteration that converges fairly quickly.
I pass the signal through the single tube stage multiple times in order to get a higher gain sound.

I have tried another experiment using more tube stages so it should be higher gain than the last one.

Does this clip have enough gain for a metal preamp ?

http://dl.dropbox.com/u/2785995/AmpSimTest-1-5.mp3

 

[FF666]

Just a suggestion, why don't you experiment with cab sim "algorithms"?? (or whatever it's called)
There are 3842348008342840239 amp sim plugins in the world.

 

[Ascendant]

That's not really enough distortion for metal and it sounds almost like a digital clipper at times, transients seem to get lost. Not sure if it's aliasing but the high end sounds a bit odd again. You should have a look at some relatively simple amp schematics because high gain amps don't really work the way you're doing it (signal -> tube -> tube -> tube and so on), there's a lot more going on than just tubes upon tubes.

Also this topic may or may not help you in the right direction: http://guitarampmodeling.com/viewtopic.php?f=51&t=7006

 

[ScootBaloo]

Just a suggestion, why don't you experiment with cab sim "algorithms"?? (or whatever it's called)
There are 3842348008342840239 amp sim plugins in the world.

Actually, modelling a tube preamp is already pretty difficult so I need to start somewhere. Perhaps people have enough amp sims now like you say but I
find it an interesting technical challenge to try and create one even though
I don't need to. Also, I don't find cab sims an interesting problem because
I think you need too much processing power to create a convincing cab simulation
so it will be cheaper to use a real cabinet for the foreseable future.

That's not really enough distortion for metal and it sounds almost like a digital clipper at times, transients seem to get lost. Not sure if it's aliasing but the high end sounds a bit odd again. You should have a look at some relatively simple amp schematics because high gain amps don't really work the way you're doing it (signal -> tube -> tube -> tube and so on), there's a lot more going on than just tubes upon tubes.

Also this topic may or may not help you in the right direction: http://guitarampmodeling.com/viewtopic.php?f=51&t=7006

The high end seems odd to me too, no doubt it will be a consequence of the model
that I am using. I had a read of the document from the great David T. Yeh who seems
to be a recognised guru in the area of amp modelling, great stuff.
Yes, my model is simple at this time, there are certain things that I don't model
yet. I do not model the grid current, I assume it is zero which means that cascaded
tube stages are often independent from each other. Assuming a zero grid current, my
circuit simulation is very similar to the first 4 12AX7 stages of a 5150 amp because
I do model the cathode resistor and cap and the anonde resistor and I do model the resistors and caps at the output of each stage.

This is a long term project for me, so I do appreciate the comments and advice
because it is not easy to create a model with a good preamp sound.

 

[ScootBaloo]

I have done another reamping test with my amp sim.

Can I ask some kind soul, does this clip have enough distortion for metal
and also, is the nature of the distortion pleasant or foul ?

http://dl.dropbox.com/u/2785995/mix-jul-05.mp3

 

[JakeAC5253]

I have done another reamping test with my amp sim.

Can I ask some kind soul, does this clip have enough distortion for metal
and also, is the nature of the distortion pleasant or foul ?

http://dl.dropbox.com/u/2785995/mix-jul-05.mp3

Saturation might be close to enough now, but the tone seems like it's HP at 1k. It needs more body and mids, but not so much that it sounds like your first few clips, because those needed less body. I'm a hobbyist circuit design engineer, if you would like to PM me a tentative circuit schematic for the preamp circuit you are designing, I could look it over and offer more pertinent advice. We could make a project out of it.

 

[userchanged1]

its definitely improving. But it does need more gain. Those DI's through any ampsim from lepou only needs a pretty low gain setting to sound full. Your sample is just barely getting enough gain.

Good micing a cab for recording is an art in itself, so if i were you, I'd remove that variable out of the equasion and use tried and tested IR's for power amp and cab simulation. So you can focus on the ampsim itself.

But I am impressed anyway. How do you start programming something like this?

 

[aviel]

i'm impressed too man. i just learned some deep stuff in C coding, and i wished to start coding stuff like that too.

 

[ScootBaloo]

Saturation might be close to enough now, but the tone seems like it's HP at 1k. It needs more body and mids, but not so much that it sounds like your first few clips, because those needed less body. I'm a hobbyist circuit design engineer, if you would like to PM me a tentative circuit schematic for the preamp circuit you are designing, I could look it over and offer more pertinent advice. We could make a project out of it.

Yep, the tone is really dark, its not HP at 1k but I can see why you said that. I am finding it difficult to get the right amount of gain and the right spectrum at the moment. I might try to increase the presence in the amp sim and then try to tame the savage high frequencies that result from that. As I mentioned before, I don't have a final circuit diagram for you to look at, but I can show you the circuit diagram that I have modelled so far. The circuit is shown on the top of page 2 of this pdf :

http://dl.dropbox.com/u/2785995/12ax7.pdf

I pass the signal multiple times through the tube stage shown above but I also apply EQ and multiband compression on the signal in between tube stages.
At the moment, I am using VST plugins for EQ and multiband compression. If I manage to get a good sound that I like then I will replace the VST plugins with code so that the amp sim is stand alone.

its definitely improving. But it does need more gain. Those DI's through any ampsim from lepou only needs a pretty low gain setting to sound full. Your sample is just barely getting enough gain.

Good micing a cab for recording is an art in itself, so if i were you, I'd remove that variable out of the equasion and use tried and tested IR's for power amp and cab simulation. So you can focus on the ampsim itself.

But I am impressed anyway. How do you start programming something like this?

Yes I agree that I still have not managed to get enough gain.
I also agree that I should probably use IR's like you say because micing up a cab throws in so many additional variables which I don't need right now, but I am determined to continue using the cab because my final goal is to implement my amp sim on one of these :

http://beagleboard.org/

Its a low cost single board computer with a 1GHz DSP processor on it.
For my next experiment, I will also post the raw preamp track and see if I can ask some one to try it out using IR's.
If you want to start programming an amp sim, some one posted a really good link in this thread above, and the following document is good :

http://dl.dropbox.com/u/2785995/comj.2009.33.2.pdf

I am not using a traditional approach to amp sim design myself but I have tried to describe above what I am doing.

i'm impressed too man. i just learned some deep stuff in C coding, and i wished to start coding stuff like that too.

I posted some info above about what approach I have taken , hope it helps. There are lots of books and resources you can buy that can help you with digital signal processing techniques, you need to be good at fourier transforms, laplace transforms and Z transforms. I have some really good books myself but I am going to have to do some revision because the maths if fairly hard.

 

[ScootBaloo]

I tried to get a little more gain from my amp sim and I also tried to get more "body" in the tone.
Do you think I managed to get more gain and/or body in the tone ?

http://dl.dropbox.com/u/2785995/mix-3-jul-09.mp3

 

[Plankis]

I tried to get a little more gain from my amp sim and I also tried to get more "body" in the tone.
Do you think I managed to get more gain and/or body in the tone ?

http://dl.dropbox.com/u/2785995/mix-3-jul-09.mp3

This is not a tone suitable for Ola. Although it might've worked with a Muse-styled song.