Guitar Lessons by Chip McDonald - "... Two EEs Walk Into a Bar, and..."

Wednesday, November 13, 2019

"... Two EEs Walk Into a Bar, and..."

 I've just wasted time reading a Gearslutz thread-war regarding IRs (impulse response).   

 Briefly, if you don't know what an IR is: an impulse response is a data file of  the "captured" transform characteristics of a signal through a device.

 In other words, the math behind what happens to a signal when it is changed by running it through a device.  In the guitar idioms, through an amp or more specifically, a speaker cabinet+microphone+microphone preamp.  You do a process involving running a test tone through the speaker, recording it on the computer, converting it to an IR file.  You then apply it in different software to replicate what the speaker does to the output signal of a guitar amp to replicate its sound without having to play at a loud volume.

 Most "amp modelers" use this technology called "convolution" math.  Some use it to replicate different points of the circuit of an amp as well as the speaker, others just the aggregate result.

  On Gearslutz it's the usual argument, "can you tell a difference?", in another thread "are they ok or not?", a classic battle in the vein of digital vs. analog, records vs. CD, MP3 vs. lossless.

 The problem is that it seems nobody is paying any attention to the application differences, while also ignoring reality.

 Reality is that an IR does nearly perfectly capture a snapshot in time of *one state* of what an amplifier or speaker does to a signal. Without a doubt, if one bashes an open A chord through a guitar rig and it's captured with an IR, if you play the open A chord the same way through the IR it will be indistinguishable. 

Stray from playing that open A the same way, you'll also stray from the identical results.  How much will depend on how stressed the speaker was initially, and how much air loading was happening inside the cabinet.  This is also discounting the electrical non-linearities of the amplifier to speaker combination.

  What that means is this:

 If you're playing guitar with a constant, steady-state signal a speaker IR works great.  What means is that if you're playing modern metal with a very compressed (non-dynamic) amp sound, AND you don't want the sound of a speaker and cabinet being killed with volume, it can work great. 

 Or if you're playing non-dynamic, clean guitar sound based music - like modern pop-country music.  It works maybe a little less well in this case, because often there are passages where the instantaneous level deviates greatly on the "twangy" attack character of some sounds.  It's still plenty convincing for live use.

 Where IRs start to fail is when the input signal varies drastically.  This happens in blues-based music.

 Blues based music is rife with multi-levels of dynamics, where chords or notes are struck differently, to emphasis a change in tonal character. This is typically done with the awareness of the historical context of "classic loud guitar sounds" - which involves a fair amount of speakers behaving erratically and non-linearly relative to the input.

 Furthermore, a lot of that historical context involves recordings with a fair amount of ROOM AMBIENCE.

 What it comes down to is that if you try to play "Since I've Been Loving You" by Zeppelin, with it's hyper-delicate and soft, guitar-volume turned down and light touch morphing to a hard-hitting loud sound through an IR, it doesn't sound identical.

 Because the IR only perfectly captures the speaker in one state, the volume at which the IR was recorded at. 



 Impulse responses only takes a snapshot of one state of the speaker, one state of the room.  If you match that perfectly you get perfect results.  Otherwise, you're in an Uncanny Valley scenario.

 People that choose to enter into this "argument" while ignoring that the application, the GUITAR SOUND has everything to do with the results.  For some things it's fine.  For others it's a degree of "almost there". 

 In the thread I'm thinking about there are 2 guys, one with a million dollar studio, both with electrical engineering degrees.  They're going back and forth on the following 2 truths:

A) In perfect steady state situations, IRs create identical results as the captured hardware creates.

B) In non-steady states situations, the results are not perfect.

 It's alienating to me that 2 people that should be educated well enough to realize there ISN'T a dichotomy there, will present A or B as a mutually exclusive premise. 

"... but listen, you can't hear the difference!" (in a perfect clinical example with identical input signals)

"... but listen, it doesn't sound anything like a real speaker cabinet! (under non-clinical, multi-variate state input signals)

 This is a fundamentally basic thing to grasp, but apparently nobody does, and everyone is willing to expend enormous amounts of energy arguing that A or B negates the other.  This shouldn't be a high i.q. thing, but simple observation of reality.  Yet I haven't seen, ever, a cogent and self aware evaluation of the upsides and downsides to the digital vs. analog guitar amp argument.

 In certain specific scenarios digital works well, and in some cases it's a much better choice than analog.  However, it doesn't cover all scenarios that an analog arrangement can occupy.  

 So please, world: stop "demonstrating" amp modelers by just bashing out chords with a lot of distortion through a close-miced speaker IR.  You need to demonstrate different dynamics on the input, with different registers: low chords, high chords, low notes, high notes, complex chords all at different volumes.

 You're not going to get the non-linear mechanical effects of the speaker cone distorting, or a cabinet that is being pummeled by pressure into extreme resonant ringing, or the spectral shifts of low to high volume. It's not going to feedback the same. An IR CANNOT DO THAT. 

 But you probably don't *need* that unless you're trying to replicate a Hendrix, Page, Jeff Beck or VanHalen live recording. 

 Yes, modeling amps will do a great job matching certain sounds at a specific input structure.  They'll fail on other things - but it may not matter based on what music you play.

 It also is a better option if you have to make compromises because of volume or lack of the best selection of gear in the analog domain.  

 It's funny to hear people knocking digital amps when they only have a small, ersatz variation on some sort of Big Classic Tube Amplifier than they can't play at the crazy volumes their favorite recordings were made at.  Or conversely, people who play very steady state compressed metal insist digital amps are perfect, who never do anything that requires a finessed and dynamic touch that creates a signal of many varying tonal inflections at different levels.

STOP MAKING NON-SEQUITER ARGUMENTS!  Creating a dichotomy out of presenting a single hypothesis doesn't mean you're finished with your thesis.  That's not being scientific.  It's just being ignorant, and this kind of presumed exclusionary single hypothesis basis for everything is wrecking modern society.  Seldom will a problem consist of a single parameter.  Digital amps vs. analog is not a single parameter topic.

No comments:

Post a Comment