GuitarAmp computer simulation, Final University Project

[FranciscoPerez]

Hello everybody!

This is my last year in university. I studied Computer Engineering and specialised in Robotics,

Industrial Computering and Industrial Instrumentation.
We studied several subjects about electronics (both analog and digital but most of the second type).

Now it's time to work in my final project, and thought about mixing my passion for tubeamp building with

computering.

I think it would be a good idea to build an application that simulates an entire guitar amp with all its

sections (first stages, tone stack, power supply, phase inverter etc), so you can tweak and easily

change the design or component value from the different sections and watch how the signal changes

throughout the signal path.

On the other side I would have a real amp connected to the computer, and it would take the signal

information from the amp so you can compare the "real thing" with the simulated signal.

Additional considerations like making a study between the effects of different types of capacitors, resistors etc would probably be a good idea.

Well, what do you think about that? Impossible to do? Nonsense? Not useful at all? Would it be a good idea? It's been done before?


Any ideas about any computer application that could be useful for us amp builders would be strongly appreciated!!


Best Regards,
Francisco Perez.

[Bassmanster]

Sounds interesting.  Go for it!

Considerations:

1) I think it will be difficult to get enough horsepower in one machine to model an entire circuit in real time, unless it's a simple circuit.  That is, if you are modeling each discrete component individually with its resistance, capacitance, and inductance.  But you can always "render" a prerecorded input signal and then play it back, right?

2) I've always wondered what those digital "amp modeling" guys like Line 6 are doing.  I'll guarantee you they aren't modeling every component.  I think what they have is some kind of hack that puts some EQ in front of a power section model.  But hey, what do I know?  I'm not a DSP guy.

[EL34]

There are several high end amp modeling programs out there.

Guitar Rig4 and Amplitube are the most famous.

I have used Guitar rig since Guitar Rig 1 and I love it

These programs have tons of different amps, cabinets, mics, effects that you can put together any way you like.
They run very low latency via good sound cards that use ASIO

Not sure if that's what you were thinking about doing or not?

[PRR]

> simulates an entire guitar amp

Very ambitious. More like a career than a project.

It's been done. Many ways.

SPICE on a modern CPU will calculate a several-tube circuit in nearly-real-time, unless it goes into clipping which makes SPICE work very slowly. SPICE "can" emulate almost any circuit with its flaws (resistor non-linearity, effects of different types of capacitors, etc) but the "small flaws" which make a tube amp distort/clip differently than a transistor amplifier will take a lot of work to quantify, set up, and simulate.

Then there are the $99 pedals with 99 different Famous Models inside. I think for these they use a real amp, feed an input (impulse, tone, noise), capture the output, and then compute the -difference- (what changes happened inside). Once you compute the complete transfer function, it can be programmed into a $7 chip (somewhat truncated) and applied to signal in real time. And you do not have to know what is inside the box you are "emulating".

It may be a lot of work, and there are very few jobs in this exact field. Even though I like the idea, if I was your teacher, I would have to argue against it. I would not want to see you play with "old stuff" for a year, in a world which is hiring for cellphone and iPod and other small complex mass-produced products. Personally, I would let you talk me into it; most professors who worked as EEs 1970-2000 would not even understand what you want to do, and just say "No, pick another topic".

> Computer Engineering and specialised in Robotics, Industrial Computering and Industrial Instrumentation. We studied several subjects about electronics

Neil Young's cherished guitar amplifer still has knobs and pots, but (I hear) he has set up a servo system to instantly "turn to" all the combinations of settings he uses in performance.

A lot of good old machines are the same way. They were built to be adjusted by hand, but computer control has many advantages.

Printing presses, sewage processing plans, electric generating stations.....

Sometimes the answer is to rip-out the entire manual control system and replace with a box. But sometimes the box does not duplicate every function, and now there is no way to turn a knob and make it work right.

Can you build a knob-turning robot, adaptable to many different and often unique manual control boards and panels, which allows computer control while retaining full direct manual control? Can the computer "learn" how the operators normally adjust the knobs, and which gauges and thermometers guide their hands, and "learn" their routine and common operations?

[FranciscoPerez]

Not sure if that's what you were thinking about doing or not?

Well, no that's not. I would simulate the signal going through the amp circuit and change in realtime component values and design of any stage to see what changes in the signal.

I've always wondered what those digital "amp modeling" guys like Line 6 are doing.

I agree with PRR that basicly they put an input signal to the real amp and store the output. Then they build the algorithm that calculates the transfer function (output/input), and finally load the program into some Embedded system that applies realtime this transfer function to the input signal.

I also agree that probably the professors (most of them about 30 years old) have never worked with tubes and won't understand what I intend to do. That's maybe the main problem for me, because he must first accept my project to be able to do it.

You're right that nowadays most of the projects related to what I studied consist in designing applications for iPhones, iPods, web/database modelling, medical surgery applications... nothing to do with what I was thinking about doing.

I would like to see that servo controlled system that Neil Young uses, it's a strange way to achieve that!
It's true that hand control is being replaced by computers very fast in those days,sometimes maybe for economical reasons having as a result a lessquality product.

I don't think hand control is better than computer control or viceversa, just it should be chosen correctly in each situation.
Machines depending on the application should have the flexibility to be able to be controlled in both ways, and actually they are. For example in the car industry, robots are often controlled by hand to do some kind of operation, or to teach the computer the exact movements the robot must perform when they are too difficult to set in the computer.

With no doubt computer controlled machines are most of the time better in every aspect compared with human controlled machines in the industrial area. Nevertheless in some specific applications, hand control is the best solution.

[Bassmanster]

They must not be using enough higher terms in that transfer function.  Well maybe that means with more embedded horsepower one day comes better models.  And a finer discretization will be needed as well.

I think some digital EQ into a power amp model would be interesting.  Not having to model every component would leave you a lot of time to work on your power amp model.  You can add the effect of higher order derivatives (what SPICE ignores) of the input signal as you go.  (You would have to combine theory and empiricism).

General purpose programming should be a piece of cake compared to this kind of stuff.  Unfortunately the employers may be too dumb to realize that.  I'm a solid geometric modeling kernel programmer and no one else knows what that means.  Or that, for instance, GUI or data processing programming is kind of a joke to me.

[tubesornothing]

My older brother is the dean of elec engineering at a local university.  He was baffled by why I was using tubes.  I soon as I explained that for guitars we like to overdrive the tube and the non-linearity of the tube is different than solid state(different harmonics), he understood.  You are simply trying to model this with simulation.  You professors should be able to understand this. 

If a whole amp is too much, why not just two stages, one to get the signal up and the next to overdrive the tube? That will definitely show the characteristics of the tube when overdriven and you can easily compare that to the real circuit.  Spice would be my choice.

[PRR]

> higher order derivatives (what SPICE ignores) of the input signal

SPICE can be fed a STIMulus file which is basically a WAV file of any arbitrary signal. (The file format predates WAV, so you'd have to fiddle the header and maybe swap bytes.)

But here is the thing. The cheapest/best simulation of a tube guitar amplifier IS a tube guitar amplifier. The cheapest-best way to compare capacitors is to heat up a soldering iron and change capacitors.

You simulate when it is a LOT of work/cost to build something which is not available, or when there is major risk and danger, AND it can be simulated quicker or safer than building and testing.

When you interconnect electric power generators, lines, loads, in a large power network, unexpected things happen. Leading VAR here, lagging VAR there, one generator responds in 2 seconds and another takes 15 seconds to change output. A fault in Michigan can ripple all the way to New York City and cause shut-downs. My father worked on simulating such problems on a vacuum tube analog computer, which cost thousands of dollars, but a real system is many millions of dollars. And when the computer "burns up", it is just a red over-limit light, not a fire on the poles or a city in the dark.

A tube amp is $500-$1000. The PC to run your simulation is at least this same price. Yes, in one way the computer is "free", because you already need it for school-papers, MySpace, etc. But it needs to be a better computer than a basic word processor, with better audio I/O. And then if you want to compare capacitors, you must measure some caps to VERY high resolution, translate that data into cap-models. It really is quicker to pick up a soldering iron.

[Bassmanster]

I don't think practicality is the object.  It wouldn't be for me on this project.

Everest.

[tubesornothing]

Additional considerations like making a study between the effects of different types of capacitors, resistors etc would probably be a good idea.

Another option is to looked at the claimed sonic differences of capacitors.  Lots of people claim to hear the difference between different brands of capacitors, it would be interesting to see if this can be measured.

[FranciscoPerez]

Well yes, it doesn't necessary need to be practical at all, it would be enough to be interesting.

Tubesnorthing, I was also thinking about capacitors, paper in oil vs polypropilene etc...That study could be a part of the project, not the main purpose of it.

Any more ideas would be really helpful for me!!

Thank you all so much!
Best regards,

FranciscoPerez