Modelling preferences and opinions

[Orvillain]

I'm really getting bored of guitarists insisting that XYZ is totally fine, because it suits their preferences.

When we talk about modelling quality, capture quality, accuracy, we are simply NOT having any kind of discussion about preference. We are having a discussion about fidelity and accuracy to a reference system.

Preference is subjective. Accuracy is not.

If someone prefers a darker tone, a more compressed feel, a different texture of distortion, that's completely valid. But it's a different conversation.

When we talk about modelling quality we’re talking about things like:

• Frequency response accuracy
• Dynamic response
• Harmonic structure of distortion
• Transient behaviour
• Compression characteristics
• Phase response
• Aliasing artefacts
• Non-linear interaction between stages

Those things can be measured, analysed, and compared to a reference amp or circuit.

You can absolutely prefer something that is less accurate. Plenty of people do. Some players prefer early digital modellers, some prefer Kemper profiles that colour the sound, some prefer darker IRs, etc. But preference doesn’t magically make something accurate.

If someone says X isn't like Y because of this evidence here, and your response is "well I enjoy it anyway" you're simply not adding anything to the conversation.

 

[MirrorProfiles]

 

[ian_dissonance]

I’m loving the MK2 thread for the “I don’t care about good but it is the best for me” showing up in the middle of a technical conversation every couple of pages. You’re out of your element, Donny.

 

[KHAAAAAAANNNNNNN!!!!!!!!]

This belongs in the rants forum.

But I enjoyed it anyway.

 

[Boudoir Guitar]

I'm really getting bored of guitarists insisting that XYZ is totally fine, because it suits their preferences.

When we talk about modelling quality, capture quality, accuracy, we are simply NOT having any kind of discussion about preference. We are having a discussion about fidelity and accuracy to a reference system.

Preference is subjective. Accuracy is not.

If someone prefers a darker tone, a more compressed feel, a different texture of distortion, that's completely valid. But it's a different conversation.

When we talk about modelling quality we’re talking about things like:

• Frequency response accuracy
• Dynamic response
• Harmonic structure of distortion
• Transient behaviour
• Compression characteristics
• Phase response
• Aliasing artefacts
• Non-linear interaction between stages

Those things can be measured, analysed, and compared to a reference amp or circuit.

You can absolutely prefer something that is less accurate. Plenty of people do. Some players prefer early digital modellers, some prefer Kemper profiles that colour the sound, some prefer darker IRs, etc. But preference doesn’t magically make something accurate.

If someone says X isn't like Y because of this evidence here, and your response is "well I enjoy it anyway" you're simply not adding anything to the conversation.

I mean, sorta. The number of people that have the experience to meaningfully discuss accuracy is pretty slim. So a discussion that is only about accuracy by folks with ability to contribute meaningfully to that discussion - is gonna be not that long or involve that many people.

And “I owned an X Amp from ‘19-‘24 so know how it’s supposed to sound/feel” is just as lame a contribution to a discussion about accuracy as subjective opinion is.

 

[paisleywookiee]

I mean, sorta. The number of people that have the experience to meaningfully discuss accuracy is pretty slim. So a discussion that is only about accuracy by folks with ability to contribute meaningfully to that discussion - is gonna be not that long or involve that many people.

And “I owned an X Amp from ‘19-‘24 so know how it’s supposed to sound/feel” is just as lame a contribution to a discussion about accuracy as subjective opinion is.

So, Orv arguing with Jay for six pages.

 

[Frodebro]

Accuracy to the original amp, or accurate to the sound in my head that is distorted by a vague, nostalgic memory of a sound I heard forty years ago and have been chasing ever since?

 

[Mongillo19]

So, Orv arguing with Jay for six pages.

 

[GTR37]

So, Orv arguing with Jay for six pages.

Possibly Sascha by page 3

 

[ChordInversion]

It's a valid point of very specific applicability. When I talk about whether or not a modeler is good, accuracy isn't my primary consideration. That's why I like my THR100HD, Katana Artists, Nextone Special, Blues Cubes, and so on. They're not aiming for a specific physical amp target, so they can't be wrong, and I don't let the thought enter my mind.

And as an example of the point made above, I haven't the foggiest idea whether the Fractal models of Orange or Dr Z or vintage Fenders are accurate. My experience with any of those is long in my past. I can only determine whether I like them.

But yes, accurate and "sounds great to me" are absolutely different things.

 

[Orvillain]

So, Orv arguing with Jay for six pages.

hahaha. Jay would run circles around me on most of this stuff. Cliff too.

 

[Orvillain]

Accuracy to the original amp, or accurate to the sound in my head that is distorted by a vague, nostalgic memory of a sound I heard forty years ago and have been chasing ever since?

Certainly accuracy to the realworld model of that specific type of amp. When I make a claim about an amp I own versus a model, that's a different thing to relying on a memory from 2008 of an amp I no longer own. Surely we can agree on that?

 

[Frodebro]

Certainly accuracy to the realworld model of that specific type of amp. When I make a claim about an amp I own versus a model, that's a different thing to relying on a memory from 2008 of an amp I no longer own. Surely we can agree on that?

Absolutely. But “accuracy” in itself should also take into consideration the tolerance specs for the components used in the amp being modeled. Mass produced amps often use components with a 10 or 20% tolerance, so three amps with sequential serial numbers, built on the same day, would all sound different from one another. Not radically, but noticeably.

Tube amp fans are aware of this and it’s no big deal to them, yet those same people will trash talk the shit out of a modeler if it isn’t nuts on to their specific amp. Funny how that selectivity works…

 

[HotRats]

If someone says X isn't like Y because of this evidence here, and your response is "well I enjoy it anyway" you're simply not adding anything to the conversation.

Well I disagree. They are adding a relevant opinion about a tool they use to do what they love to do.

Many players are simply much more interested in making music and having fun, and every comment they make about the tools they use comes from a pragmatic point of view.

The good old “if it sounds good, it is good” has its merits, in other words, accurancy is intererseting but making music intersts more.

While this is a gear forum, it's not a developer forum. It’s unrealistic to expect that users more interested in music than technology will approach the discussion differently or follow your personal ruels about what qualifies as a relevant answer.

 

[Orvillain]

Absolutely. But “accuracy” in itself should also take into consideration the tolerance specs for the components used in the amp being modeled. Mass produced amps often use components with a 10 or 20% tolerance, so three amps with sequential serial numbers, built on the same day, would all sound different from one another. Not radically, but noticeably.

Tube amp fans are aware of this and it’s no big deal to them, yet those same people will trash talk the shit out of a modeler if it isn’t nuts on to their specific amp. Funny how that selectivity works…

I agree with you, but also it is very hard for anyone without the insider knowledge on the tech, to truly know whether a difference is due to tolerance, or bad modelling.

I'm involved in a conversation elsewhere, where someone is insisting a model is spot on. But I have the amp. Tolerances simply would not account for the differences I am hearing.

 

[laxu]

It's a valid point of very specific applicability. When I talk about whether or not a modeler is good, accuracy isn't my primary consideration. That's why I like my THR100HD, Katana Artists, Nextone Special, Blues Cubes, and so on. They're not aiming for a specific physical amp target, so they can't be wrong, and I don't let the thought enter my mind.

And as an example of the point made above, I haven't the foggiest idea whether the Fractal models of Orange or Dr Z or vintage Fenders are accurate. My experience with any of those is long in my past. I can only determine whether I like them.

But yes, accurate and "sounds great to me" are absolutely different things.

This is how I see it. That's why I say "accuracy is a great goal for the people making the device". But it doesn't really matter that much to the user - if they are getting the tone/feel they want out of the device.

I have tried only a subset of the 100+ unique amps on a Fractal. I haven't owned most of them to have intimate knowledge of them. I've just tried them in stores, as backline amps, or a friend owned one. So what do I know how exactly that Plexi 1970 model is supposed to sound? At best I have a "this is what I think some Plexi sounds like" type reference for it. Then it fits into my vague idea of that sound.

A lot of the time I find modelers fun and practical, but ultimately unnecessary for me if I have to be honest. I have pretty much only one sound that I truly need. It's the sound of a higher gain, Marshall style amp. Some oomph in the lows, some sizzle and bite in the mids and highs. The kind of thing that is perfect for those 1980s rock/metal riffs. Goose it up with some effects and wail.

I can get that thing out of so many things. Either of my BluGuitars. Fractal's various models. Tonex. Yes, even my Mesa Mark V's ch3 does that. They're all just different flavors of that thing.

It doesn't mean I don't find other sounds cool too, but it doesn't move the needle for me much if a modeler has tons of models. If even one of the models does my core tone the way I like, I'm good. I had a lot of fun with the Fractal Trainwreck Express model the other day. That does the thing too.

 

[MirrorProfiles]

Tube amp fans are aware of this and it’s no big deal to them, yet those same people will trash talk the shit out of a modeler if it isn’t nuts on to their specific amp. Funny how that selectivity works…

I think it’s equally wrong to reduce any differences between modelling and real amps to just tolerance variation. There are differences that can be attributed to variation of components, or age etc, but not all differences fall under that.

A case in point would be capturing tech which removes most of those variables, and yet there are still differences (even on the highest accuracy capturing and training methods available).

I feel like there are characteristics across ALL digital models that distinguish them from real amps. Some are hard to verbalise, or measure in a consistent way. One very obvious area is real amps of stereo tracked guitars always sound wider than models.

I think there's something still in the way the low end behaves, and also harmonically. These aren't always drastic, but there is something that real amp intrinsically do.

 

[Jarick]

There is no common language to properly measure and compare the accuracy of modeling right now, so there's no point in having those discussions purely from an analytical lens. Until someone makes a Geekbench for digital modeling/capturing, it's all just apples and oranges.

What's most interesting to me is people's experiences, specifically those that are unbiased. If you're a tube amp lover but super geeked out on a certain modeler, that's interesting. If you like one unit for amp modeling but pair it with another unit because you prefer the modulation and then you like to add this pedal because you can dial a knob in on the fly and it's really inspiring, that's interesting to me.

Unfortunately there's so much financial incentive to either lie about loving something, or hating something, or even just to stir up controversy, that so much of these discussions have lost all meaning. So it defaults back to personal experience for me.

 

[dean701]

I am not arguing anything stated so far with respect to what you like vs accuracy as I agree but I do think that the measurement strategies are up for debate.

We've had similar conversations about null tests or debunking them and the like so the tools aren't just cut and dry.

If it were so easy to copy every element then new firmware updates wouldn't be so active. Cliff and his endless quest is a shining example. New calculations new algorithms...v2 Tonex, v2 QC are other examples.

PopTart had a thread on TOP that got deleted about null tests. They can close threads but why delete?
Another one started after that and it may still be there.
Regardless, Digital Igloo had posted some info about Proxy and it got deleted...Time domain/Freq domain/hybrid info that was starting to interest me.

 

[Orvillain]

There is no common language to properly measure and compare the accuracy of modeling right now, so there's no point in having those discussions purely from an analytical lens. Until someone makes a Geekbench for digital modeling/capturing, it's all just apples and oranges.

What's most interesting to me is people's experiences, specifically those that are unbiased. If you're a tube amp lover but super geeked out on a certain modeler, that's interesting. If you like one unit for amp modeling but pair it with another unit because you prefer the modulation and then you like to add this pedal because you can dial a knob in on the fly and it's really inspiring, that's interesting to me.

Unfortunately there's so much financial incentive to either lie about loving something, or hating something, or even just to stir up controversy, that so much of these discussions have lost all meaning. So it defaults back to personal experience for me.

Historically, the amps I own, Fractal has always been the closest. But quite a lot of the time, I've preferred the sound of a given NeuralDSP plugin. So I'm definitely open minded in terms of what to use.

But there are some common approaches and frameworks to properly measure. PluginDoctor itself can give you a 7 order Hammerstein plot across the spectrum, and it also accepts audio I/O as well as a plugin. So you actually CAN measure a plugin Marshall JCM800, against a real Marshall JCM800, and come up with some decent results that way. And that's just one thing off the top of my noggin.