Has the Axe-FX II Achieved 99% amp modeling accuracy?

[Clockwork Creep]

What if an IR is made from a mic that is far away from the cab?

 

[guitarnerdswe]

What if an IR is made from a mic that is far away from the cab?

We already have that. They're called far field IRs.

 

[Speculum Speculorum]

There's also ways an amp and can interact that haven't been measured yet. There are interactions going on that we probably don't even know exist yet. Tubes do some crazy things in physics and sound does even crazier things.

Hell even quantum was the result of discovering a parasitic aspect of power amp interaction Cliff didn't know about.

The axe is a powerful and great sounding piece of kit, but it's still a model. We aren't there yet and I'm not sure we ever will be.

 

[Ed DeGenaro]

I agree, and here's why.

It doesn't actually respond to a physical amp in the same way that a speaker would in a cabinet.

There's a special relationship that between a speakers in a speaker cabinet and the output transformer of a tube amp. That back and forth relationship between the output stage of the amp and the speaker is measurable (Impedance curve).

The impedance curve considerably affects the way that an amp sounds and responds. Depending on what type of cabinet you have that speaker loaded in, will actually change the impedance curve. The impedance curve affects the tone of your amp and the way that it responds.

So, if every speaker and cabinet combination has it's own special impedance curve, how do you account for all of those and put them into the Axe-Fx's speaker page?

You can't.

Cliff probably modeled an impedance curve for each amp model and put that as the default into that model's speaker page.

So, those resonant peaks might not be lined up with the IRs that we use.

However, it's close.

But until some breakthrough in technology comes along, this is where we're at with the Axe-Fx.

That while sales pitch for reactive loads is all fine and good with one problem.... That impedance isn't a constant static curve it changes depending on frequency....

 

[Lionheart]

That while sales pitch for reactive loads is all fine and good with one problem.... That impedance isn't a constant static curve it changes depending on frequency....

I'm not sure if you're agreeing with me or not.

 

[Brownmatthall]

I vaguely remember cliff saying the impedence curve for the models were measured with the amps matching (or the closest thing to it) cabs, so I don't think it's this completely arbitrary number cliff slaps on but it's only going to correspond correctly to the cab.

 

[Ed DeGenaro]

I'm not sure if you're agreeing with me or not.

I do mostly, the one thing that I don't subscribe to is the two dimensional nature of an impedance curve, I mean part of a speakers nature is to go between reactive and resistive above and beyond its resonance frequency. It doesn't figure in sanding etc... Which a waterfall plot would. It's kinda the same when you look at the frequency response curve of a ported studio monitor vs. Sealed box, as oppossed to a waterfall plot of the same.

 

[Lionheart]

I don't think it's this completely arbitrary number cliff slaps on but it's only going to correspond correctly to the cab.

No, I didn't think it was just something he pulled out of thin air. I'm sure he put his technical knowledge to use. No doubt about that.

 

[Lionheart]

I do mostly, the one thing that I don't subscribe to is the two dimensional nature of an impedance curve, I mean part of a speakers nature is to go between reactive and resistive above and beyond its resonance frequency. It doesn't figure in sanding etc... Which a waterfall plot would. It's kinda the same when you look at the frequency response curve of a ported studio monitor vs. Sealed box, as oppossed to a waterfall plot of the same.

Okay, thanks for the clarification. I can get on board with that.

 

[jesussaddle]

While I am confident we are experiencing a high rate of accuracy, it is fair to quibble with qualitative assertionalities. Instead of using 99% as a measurement, I would suggest dividing PI by the sample rate of 48k, as multiplied in terms of Egyptian cubits of the IR. This would put the Axe FX on par with a primitive (and radically slave-dependent for its day) form of ancient architecture.

It is said that the otherwise ridiculous and primitive Egyptians (just watch "Walk Like an Egyptian" to see the ridiculous way they walked, without any help from the Bangles) managed to get a pyramid that approximates Phi and Pi.

( maybe, at least if this is true:
http://www.goldennumber.net/phi-pi-great-pyramid-egypt/
)

Therefore, an appropriate qualitative constant would actually be useful if possible.

To determine something accurately is no joke. To determine it both qualitatively and quantitatively, that's rare. I'm actually glad Pythagoras figured out how simultaneous the quantitativeness and qualitativeness was, and created an actual method of less primitively deviating from the Equal Tempered Western Tuning, finally!

Now if I could just get the frets of my guitar to be 99% accurate to the musicality envisioned by the Egyptians, which used geometry evidently (from which we equally divide circles into a single value, 360, which can then be evenly divided into 6, 5 and 4 parts and so on, rather than boring base ten which, unlike those funny walking primitives, decimates everything). The fact that Cliff outsmarted the rest of the musical technology world, that's on par with Pythagoras, but did he outsmart the Egyptians too?

Wait, I think I almost decimated a perfectly useful topic.

Sorry.

 

[lqdsnddist]

Arguments over the finer sonic nuances of tube amp design always seem a bit academic to me when the reality is that most likely any signal (ie; music) is going to be heard/enjoyed via crappy earbuds and low bit rate mp3/streaming audio, or via a house PA system to a room full of drunks.....

Its kind of like "debating" the merits of specific vintage of wines and then throwing them in a wine spritzer, regardless of how fancy of vintage it started out

 

[Lionheart]

Arguments over the finer sonic nuances of tube amp design always seem a bit academic to me when the reality is that most likely any signal (ie; music) is going to be heard/enjoyed via crappy earbuds and low bit rate mp3/streaming audio, or via a house PA system to a room full of drunks.....

Its kind of like "debating" the merits of specific vintage of wines and then throwing them in a wine spritzer, regardless of how fancy of vintage it started out

Well, hell. I'm selling my Axe-Fx for a Line 6 Spider IV 30, then.

 

[∞Fractals]

I simply enjoy playing for hours on end; accurate enough for that.

With a knob, quick twist and bam ... mo' betta inspiration/tonez with finding a new amp/config never gets old. (I do play out to "real" stereo cabs 95%.) Fiddling with all the various FX and parameters is just as fun.

Maybe I don't know enough, don't have the talent or my ears are shot - but I don't really care (or even feel the need to analyze), because I enjoy every single day that I get to play.

 

[stm113]

It depends on what you're playing. This thing is pretty astounding for clean tones, but once you start getting into breakup and high gain territory there's still room for improvement. This is just opinion of course.

It's much debated, but the high and low end on distorted tone still suffers, in my opinion, from a digital sound in top end, and a whompy low end. Also, the way the breakup sounds is still too smooth and undefined. It's gotten a lot better since the last few firmware iterations,to a point where many don't seem to be bothered by it. But I've yet to hear something as saturated sounding as, say "Silvera" by Gojira, "Psychosocial" by Slipknot, or "Vicarious" by Tool. Sure, the tone might be spot on, but the sputtering crackle just isn't there.

I think you're right about this. The biggest thing I wrestle with in dialing in tones with my fractal is the top end bite. It just doesn't sizzle, its either too smooth or just way too fizzy but not what I'm trying to get.

Another good example is Slipknot's "Before I Forget".

 

[Axeman]

Hi everyone.

I have been a modeling amp user for many years, starting with the Fender Cyber Deluxe, then moving to the AXE standard, and now I love my Axe-FX II. Over the years, I've owned only a few tube amps, so I don't have the depth of experience that many others on this forum have with all the various amps that have been built over that last 50 years or so. So this question is for the seasoned amp users out there that have a very deep well of knowledge and experience with many different amps.......

With the quantum generation firmware updates, has the accuracy of the AXE-FX II achieved the 99% accuracy level? In other words, is there anything significant left to develop with the amp models to further improve tone/feel/touch sensitivity/etc? the latest quantum firmware is so incredible, it's hard for a less experienced amp user like me to imagine it can get any better.

Love to hear all the expert's thoughts.......

How do you define "99% accuracy level?" This the first question before you can answer your question.

My personal opinion is that any tone can be replicated by the Axe at "99% accuracy." But here is what I mean by this. First, one needs to build the Axe model (and cab if using an FRFR system) in the exact same room with the same guitar, cable etc. It likely will take a number of hours to make all the right adjustments. You may or may not start with the exact amp model in the Axe, but obviously it would be best to start with the model that represents the amp you are using. Then an IR would get created using the cab that is used in the room (with the exact mic that is on the cab).

After spending a few hours on this with at least two or three other people listening and providing input to the changes, the group would step away for a day and come back the next day and see if they hear the same thing. They would adjust again and repeat until they believe they have the exact tone replicated. The more people brought into the process the better.

Then you select a few hundred people at random and put them through a double blind study and played the exact same chords and/or notes (record dry guitar and play exact signal through each system). You play the signal through each system a number of times in some determined order (same for each listener). You then ask the if they thought all the clips sounded the same (yes or no). There may be a set of other questions to make sure the answers are consistent from each listener. In a double blind study BTW, people who didn't set up the preset run the study, and hundreds of people are brought in to hear the comparison but don't see the equipment or know if there are one, two or more systems.) The key is always to get the bias out of the study. You can't tell a listener that there are two systems because then they are already inherently biased and trying to listen for differences.

I'm pretty certain 99 out of 100 will say the tones all sounded the same. If I'm wrong, of course one could argue not enough time was spent building the preset or in fact the Axe does not model the real world. But my point here is we will never know how accurate the Axe is until we have a double blind study with a statistically significant number of participants.

No disrespect to anyone here, but I think all the opinions on this topic on this board over the years should be questioned until someone runs such a study. There are so many biases behind so many of the opinions. We all have different hearing, like different types of music, have different gear, play at different volumes and have inherent biases or are easily biased by the information we are given.

Even my own comparison video on YouTube (comparing real AC30 and and Axe-Fx) is far from perfect because I tell everyone there are two systems (which I should have never done). This alone taints the comparison because then listeners assume there must be a difference and the mind can play tricks on you (just create a preset over a few hours and try it the next day and see if it sounds the same). I should have posted a bunch of clips and asked if there were one, two or more different systems. Then I could have run another sample and changed the number of systems (maybe only just one - the Axe playing the same clip twice) and asked the same question. This takes out the bias of knowing there are two and focuses the listener on whether they are hearing the same sound or not, independent of any other info. Ultimately what I'm getting at is if I were to post the same clip from the same system twice, some people will say they hear a difference and in some cases when I post the same clip through two different systems, some will say they don't hear any difference. It's important not to let the listener know anything about the setup and that the sample data is large enough to be statistically significant.

I've posted this video before and I think it supports what I'm trying to explain above.

 

[jesussaddle]

I think you're right about this. The biggest thing I wrestle with in dialing in tones with my fractal is the top end bite. It just doesn't sizzle, its either too smooth or just way too fizzy but not what I'm trying to get.

Another good example is Slipknot's "Before I Forget".

This could be due not to the amp models, but the IR.

While an IR is possibly a perfectly effective solution to be combined with the Fractal amp models, still I can't help but wonder if Cliff doesn't have something up his sleeve to one day improve on that, since its obvious that the power amp, direct, is going to sound pretty harsh no matter how good Cliff's algorithms are.

I also notice that most everyone argues in favor of the lowering of EQ ranges, rather than the increasing of them. But I'm finding that harmonic EQ'ing is more interesting that viewing EQ in linear bands. But EQ technology seems to be in stark contrast to the sounds from natural media that oscillate in complex timbres.

My (nonexistent) understanding is that IR's are similar to a linear filter. If so, even when slicing ranges according to a static map, surely this departs from the flexible nature of physical mediums like a speaker, whose characteristics must vary as it changes shape. Maybe it is unnatural to have any higher frequency ranges maintained for any length of time.

The nature of eyesight is that of desaturation and resaturation of pigment, and I'm wondering if our ears are just not designed to resonate with such constant bands of higher frequencies (maybe this applies to anywhere above 1/2 K, but is more unnatural the higher one gets, like at 5 K and above). Very high and very low frequencies (arguably two separate forms of energy intensity) may each produce some form of fatigue if continued unnaturally. I like to picture most natural mediums used for musical instruments as falling somewhere on the rigidity scale less than totally solid, and totally inflexible.

Just saying. Ooops - didn't mean to open up another unrelated topic.

 

[rrogers]

That would mean, that they are different... Which means, the models in the Axe are not yet accurate...

Here's another question: Even if amp modelling is 99% there, what about cab modelling? This whole manual thing about the speaker page does create inconsistencies with reality... Also, are IR's really accurate to a real mic'ed up cab in a room? I'm skeptical.

Being different doesn't mean a model is "less accurate," particularly when dealing with older tube amps. I don't have two Marshalls that sound the same. Sure, two of the same heads if I A/B them with a single cabinet are in the same "ballpark" in terms of tone, but there are still subtle differences from one head to another. For example, I have one JCM800 that sounds much better (to me at least) than my other two. Older amps, such as the JTM45s I have, seem to vary even more.

I judge the tone achieved with using IRs on the sound of an amp mic'ed on stage. If I walk out in front of the PA mains, what I hear when using the Axe sounds better than the tones I have achieved when micing an amp.

 

[CodeStation]

I honestly don't care if it has.
I am getting the most... realized... sounding recordings on my slowly crawling project that I've ever achieved.

Frankly, I think it's better than the real thing from the standpoint of usability for someone without a 500,000 $ studio.

 

[dpeterson]

I'm fine too with what I'm getting. Benefits far outweigh the crap nobody in an audience can hear. I'm sure Cliff has some things he's heard that he cannot replicate, history being a guide. He'll eventually figure out something that has been stumping him and put in the firmware.

Some people resist change and don't want to touch anything but tubes. It's a generational thing, growing up, non tube amps were horrible, hard to break out of that mindset for some people.

 

[Roland]

For me the accuracy question is no longer relevant. It's a cost effective tool for recording and for mic'd up performances. There is something lacking in the un-mic'd live sound, but that's about the simulation of un-mic'd cabinets, not about amp modelling. Further, I don't think there's a modelling solution because it's primarily about physical speaker and cabinet characteristics.