Axe-FX II

[roodyrocker]

I own several tube amps as well as the Axe Fx Ultra. I like them all and I don't feel in the least bit slighted by the release of the Axe Fx II. I plan to keep my Ultra and have no plans on dumping it. Its a great sounding unit. Just because a newer version comes out doesn't mean the previous one is no longer any good or "outdated". The amps as well as the effects in it sound stellar! not only would it cost you a ton of money to get all the amps and effects in it, but as Tim pointed out, you couldn't fit all that stuff on stage anyway. The other issue is that most tube rigs with a few pedals in front of them or with rack processors through their loops buzz, hum, and make all kinds of funny noises. I've even seen some guitar greats live with horribly noisy rigs that they literally fought with all night long. On the Axe Fx you can get clean, distortion, with or without as many effects as you want at any volume you want and its a quiet rig! I've gigged my tube amps and continue to do so but I've also gigged my Axe Fx in a two guitar band situation. All that talk about it doesn't have the cut of a tube amp or doesn't feel or move you like a cranked tube amp is all BS. Nothing wrong with tube amps but so far all the reasons I've seen on here not to get an Axe Fx are pretty misguided. And by the way, I also gigged a Line 6 Pod Pro rack unit way back when...Once! I hated it. The Line 6 stuff can not even begin to compare to the Axe Fx stuff.

[Cliff]

Thanks for the paper, Javert. Not enough info to actually go and implement something like this, but still cool. I see the Axe FX II has a 600Mhz DSP dedicated to the amp modelling. I wonder if a modern desktop CPU could compete? I presume the DSP has much faster memory access? Anyway, I wonder if something like this could be thrown together as either a standalone app or some sort of plug-in for AmpliTube or even Garageband?

Here's a thought: if such an app wasn't actually fast enough to be real-time on PC, could it still actually be any use? I mean, what if you had something that gave half-decent results in real-time, but once you had your track down you could post-process it off-line to get the slower, better modelled amp tone? Does the lack of real-time make it a complete nonstarter?

[Cliff]

So, yeah, the monumental supidity of that last suggestion is only just beginning to dawn on me. Please ignore .

[Hellbat]

Actually, a GPU would be better than a CPU for this. They should see if they can do a plug in version that uses NVIDIA or ATI graphics chips for the math bits. Since they are massively parallel and have huge memory busses they could probably pull it off.

[Endrik]

New technology is cool but no matter how much it improves digital modeling is still zeroes and ones.

Generally people in the music business accept that old primitive analog equipment sounds the best. Theoretically that equipment doesn't look good on a paper, but in reality it sounds great. All its "theoretical" disadvantages like noise and distortion are actually things which create "the magic". The imperfectness of simple analog technology perfectly complements the imperfectness of human beings. Unlike digital modeling, it's in direct contact with the musicians, and if we go into more metaphysical world we can say that it carries on the energy of the musicians, and boosts in it's own way, like it all becomes one.

Analog tube amps, good vintage mikes, good vintage pre-amps, analog consoles, 2" inch tape and into vinyl record -> the whole chain has so much noise and imperfectness but it's really amazing how much energy of the musical performance it retains.

Modern digital world realizes it, it acknowledges that this old primitive set-up sounds the best and therefore tries to replicate it. Digital recording technology is definitely more advanced than guitar modeling technology. There's so much advances made. It becomes better and better every year. Why? Because the folks who are working on it are breaking down the code of all the good analog technology more and more. The better something sounds, the more it's coded to work like old equipment.

The reason why most use digital technology is because it's way cheaper and easier to use than the old technology. But everyone still loves the "old sound" no matter what music they make. It can be overproduced, really modern pop crap, that doesn't mean the producer doesn't like to use some vintage compression emulators... he just over uses it.

There's a lot of great stuff made purely in digital environment, no question about that but no matter how much the technology improves the music makers themselves still want to hear something better. There's still that "something" that's missing. It's pretty common among those who work with digital equipment to purchase analog mic pre-amps and converters too. And not only that... ProTools and whatnot can make as many updates as they want but the songs are still coded. Zeroes and Ones. You record something and in the mixdown all those recorded parts are taken apart by and coded together with 0101010111000 or something like that. With analog technology the recorded parts aren't decoded and recoded. In the mix all the original parts co-exist together. That's why many also buy analog gear for mixdown. It simply sounds better.

The general audience may not care but the people who actually create music want to make everything sound as good as they can, they want to be happy about what they can achieve. So those with really picky taste and high demands go back to analog gear.

It's the same thing with guitar modeling. It can sound very good but many still feel that this is not the absolute best. The audience may not notice the difference but the musicians themselves want to make themselves happy too.

So it basically comes down to compromises. For the ultimate tone heaven you have to be ready to sacrifice a lot of money and your nerves caused by the nature of old technology or you can get good results very easily but that doesn't mean you secretly aren't disappointed in yourself because you are not giving your 100%.

[Trem]

The other issue is that most tube rigs with a few pedals in front of them or with rack processors through their loops buzz, hum, and make all kinds of funny noises. I've even seen some guitar greats live with horribly noisy rigs that they literally fought with all night long. On the Axe Fx you can get clean, distortion, with or without as many effects as you want at any volume you want and its a quiet rig!

The biggest reason i like the unit. It trumps any benefit if there is any gained by using "real" tube amps at high gain/volume to me. The gate in the axe is superb, like it's not even there.
It's just not the same with a tube amp and noise suppression or a gate, the feel is off. Not so with the axe, it's like playing cranked tube head with no noise and no choke whatsoever.

[javert]

Thanks for the paper, Javert. Not enough info to actually go and implement something like this, but still cool. I see the Axe FX II has a 600Mhz DSP dedicated to the amp modelling. I wonder if a modern desktop CPU could compete? I presume the DSP has much faster memory access? Anyway, I wonder if something like this could be thrown together as either a standalone app or some sort of plug-in for AmpliTube or even Garageband?

Yeah, I know, but there's a whole bunch of references to more techinical papers for those interested. I figured that if I referred to a paper full of math, nobody would read it anyway. DSP are processors optimized for doing signal processing (which usually involves reading data, multiplying, adding and then writing data) and they can do additions and multiplications in a single clock-cycle. Moreover, they've got dedicated instructions for handling circular buffers and are usually built based on a Harvard architecture, which means that they have separate program and data memory and buses, so both can be read or written at the same time. They're usually not competitive with CPUs in pure clock frequency, but are preferable due to their optimized architecture and instruction set and low cost.

Hellbat is right that a GPU is good at the same kind of stuff (although they are of course aimed specifically at computer graphics, which is not exactly scientific computing), and this is why much signal processing research is now devoted to figuring out how to use GPUs for things like this.

[Cliff]

I found this, which looks like almost enough info to make an implementation:
http://dafx10.iem.at/proceedings/papers/MacakSchimmel_DAFx10_P12.pdf

They give some audio samples of what their simulator does here:

http://www.utko.feec.vutbr.cz/~macak/DAFx10/

I'm certainly no expert, so I'd be interested to know if you guys think those sound plausible.

It's interesting to note that the authors mention several compromises and assumptions they make to get the thing to work in real time. I guess there are always going to be compromises, though.

Javert, Hellbat - I wondered about the GPU idea also, but I'm not so sure. It only pays off if there's a lot of parallel work to be done. We can either take an individual sample and maybe apply some parallel ops to solve the differential equations, though I haven't looked deeply enough to see if this would be the right way to do the solve. Or instead process a whole batch of samples simultaneously in parallel. I don't think this second approach buys you anything. Suppose for the sake of argument you have 1ms between each sample. If you process serially, you can take up to 1ms on a sample before coming up with the result and moving on to the next. Suppose instead you process 1000 samples in parallel, and you have 1000 multiprocs in your GPU. At first it looks like you've got a second to process all of those in parallel. It's true that there's 1 second between sample 1 and sample 1001. But you've still only got 1ms between sample 1000 in the first batch and sample 1001 in the second batch. Since the GPU is SIMD, we have to apply the same ops to all 1000 samples, so we still have the same time constraint. I think . Add to that the fact that the GPU runs slower than a CPU, has no cache (except the very latest Nvidia ones), and you have to take time to schlep the data to and from the GPU, I'm not sure it would be a win. Unless some sort of parallel solve would be of benefit on a per sample basis.

[javert]

Thanks for the link, Cliff. I'll check it out. I've been meaning to atttend DAFx for some time now. It used to be really, really bad quality, but it seems like more and more good stuff turns up there. I'm not so much into the whole GPU thing in signal processing (perhaps I should be), so I can't say for sure, but I think you're right that it needs to be parallelizable. The problem is that if there's memory in the system (i.e., if what happens at time n depends on the data from n-1 being processed already), then you can't do it. I think you're also right that the delay incurred by bufferering may be a problem.

[usajackson]

I'm working on a new "special magic box" and computer program that will perfectly replicate not only any tube amp in the world, but also any guitar/pickup/string gauge combo one could think of. It will be able to be used by non-instrument playing "musicians" to "play" music. All you non-tube amp users will be able to take your iPad to a gig and plug "the magic box) into the house PA system. Each "band" member will just need their own iPad (you would not even have to be on stage; you could sit at the bar and jam with wifi or 3g!). Still have some bugs to work out for the drum and bass guitar software, but I'll figure it out. 1's and 0's rule the world.

[javert]

1's and 0's rule the world.

You got that part right.

[Chief_CrazyTalk]

LOL. Which is why I'll be sticking with my humble, little JCM800 ca. 1981, thanks.

[VitaminG]

you guys still waffling on about this Axe thingo?

Since I got my JamVox, I've sold all my gear. It's not pretend like the Axe box - it uses real analog technology inside my computer, so there are no transistors in the distortions.

I've found that using the Uber Metal amp model with a Xtreme distortion in front of it, and some epic reverb, chorus & delay for solos, I can exactly replicate Metallica's Our Justice For Ever tone when I use my Eet Fuk signature Hondo Explorer with EMG 81s. I know not everyone can afford gear like this, but you have to at least have some high quality monitors for your PC. I'm talking the $70 kind, not the $30 shit from Radio Shack.

[bibz]

Can you take an impulse of those monitors for me G? My $30 shack monitors are still on layby. I could just sell my jsx, but its doing a great job keeping my quadbox in place. Why did they put wheels on these things, they're too easy to move out of the bedroom!!

[Shawn Lutz]

LOL Gary...