2023年6月23日 星期五

Modeler vs Profiler ?! |NUX AMP ACADEMY vs IK TONE X vs Kemper Profiler Player

Modeler vs Profiler ?! 


Stomp-box Modeler

之前我們在2022年3月份,寫過一篇橫向對比的Stomp-box Modeler,介紹了市面上常見的產品,如NUX Amp Academy、Strymon Iridium、Walrus ACS1、Atomic Ampli-Firebox、BOSS IR-200等等。

Previously, in March 2022, we wrote an article on the Stomp-box Modeler comparison, introducing common products on the market, such as NUX Amp Academy, Strymon Iridium, Walrus ACS1, Atomic Ampli-Firebox, BOSS IR-200, etc.

而主打Plug-In軟體的IK Multimedia也在AmpliTube的基礎上,推出Stomp-box Modeler: TONE X。 TONE X採用的算法是類似QC的AI建模,源自跟KEMPER系出同門的『黑盒算法』。

IK Multimedia, which mainly focuses on Plug-In software, also launched the Stomp-box Modeler: TONE X on the basis of AmpliTube. The algorithm adopted by TONE X is an AI modeling similar to QC, which is derived from the "Black Box Algorithm" of the same family as KEMPER.

註:關於Amp Modeling的技術與IR的觀念介紹,請見:

Note: For the introduction of Amp Modeling technology and IR concepts, please see:

IK TONE X vs NUX Amp Academy


I understand that everyone has no patience to read a bunch of text descriptions, so we go straight to the main course and compare the sound.
Here we select a few references with the same Amp Models for comparison. (Equipped with the same IRs.)

1. IK ToneX vs NUX Amp Academy | JCM800 

2. IK ToneX vs NUX Amp Academy | Super Reverb

3. IK ToneX vs NUX Amp Academy | Twin Reverb

Amp Modeling & IR

在說明Modeler與Profiler之前,我們先來講述一下Amp Modeling跟IR之間的區別,我發現有許多人對於這兩者之間的定義有所混淆。

我們在中文說到的『音箱模擬』指的就是『Amp Modeling』,也就是針對『音箱』來做建模的動作。後面我們會講述到Modeling的方法。

Before explaining Modeler and Profiler, let's talk about the difference between Amp Modeling and IR. I found that many people are confused about the definition of the two.

The "Amp Simulation" we mentioned in Mandarin refers to "Amp Modeling", which is the action of modeling for an "Amplifier". We will talk about the Modeling method later.

From the picture above, it can be seen that what IR restores are the sound of "Micing the Cabinet (speaker)".)

而IR其實是近幾年用來實現Cab Sim的一種方法,在之前的文章有提到過,近年因為DSP運算力上來了,所以早期採用濾波器做Cab Sim的方法就改由IR卷積取代了。 所以IR指的是『箱體模擬』,並非『音箱模擬』。

IR is actually a method used to realize Cab Sim in recent years. As mentioned in the previous article, in recent years, due to the increase in DSP computing power, the early method of using filters to do Cab Sim has been replaced by IR convolution. So IR refers to "Cabinet Simulation", not "Amp Simulation".

Amp Modeling: Modeler & Profiler


Amp Simulation, as the name suggests, is to "simulate" amplifiers, so we have to take a look at what kind of amps we want to simulate.

首先我們理解到我們要模擬的對象『Electric Guitar Amps』有很重要的特性,那就是帶有『失真』信號。即使是我們認為的Cleanish類別音箱,其實相較於Acoustic Guitar Amp 或者 PA系統,還是帶有失真特性。

First of all, we understand that the object we want to simulate, "Electric Guitar Amps", has a very important characteristic, that is, it has a "distortion" signal. Even what we think of as a cleanish category amp has distortion characteristics compared to an Acoustic Guitar Amp or PA system.

而我們所說的Electric Guitar Amp往往指的就是Tube Amp。 因為吉他手們喜愛真空管音箱那“溫暖且肥厚的音色”以及“充滿生命力的失真”。

And what we call Electric Guitar Amp often refers to Tube Amp. Because guitarists love the "warm, fat tone" and "organic distortion" of tube amps.

而電吉他音箱主要形成音色的部分在前端的PRE-AMP,這部分也是主要失真特性的來源。 所以在Digitech 、Zoom、Line 6這三個經典綜效老牌(2005~2010主力綜效品牌)在當時半導體行業發展下,採取了『Waveshaper』的建模方式。 這種建模方式的好處是,工程師不用理解真空管電路的工作原理。(小編為1983年生,在大學時期的Smith電子學聖經本早就沒有“真空管”了,當時電子行業也早已經是半導體為主流了。)

The main part of the electric guitar amplifier to form the tone is the PRE-AMP, which is also the source of the main distortion characteristics. Therefore, Digitech, Zoom, and Line 6, the three classic brands (main brands from 2005 to 2010), adopted the "Waveshaper" modeling method during the development of the semiconductor industry at that time. The advantage of this approach to modeling is that engineers don't have to understand how vacuum tube circuits work. (I was born in 1983. When I studied at university, the Smith Electronics textbook had no "vacuum tubes". At that time, the electronics industry was already dominated by semiconductors.)

這是Preamp的最後一級,旨在驅動效果環路,也是第一個出現失真的Preamp級。 該設計類似於 Mesa Boogie MK I 中的Preamp失真級。

陰極電阻器被完全旁路,以防止負反饋降低三極管電路的豐富非線性行為。 R1 和 R2 形成分壓器,以將大約 100V 峰值的最大板電壓擺幅降低到大約半伏峰值的最大影響發送擺幅。

根據 12AX7 計算器,柵極偏壓在非常溫暖的 -1V。

This is the final stage of a preamp designed to drive an effects loop and is the first preamp stage to break into distortion. The design is similar to the preamp distortion stage in the Mesa Boogie MK I.
The cathode resistor is fully bypassed to prevent negative feedback from reducing the triode circuit's lush nonlinear behavior. R1 and R2 form a voltage divider to reduce a maximum plate voltage swing of about 100V peak to a maximum effect send swing of about half a volt peak.
According to the 12AX7 Calculator, the grid is biased at a very warm -1V.

隨著輸入信號的增加,上游電阻網絡上的電網電流的影響會壓縮正向電網電壓擺幅。 輸入信號的負擺幅不受限制。 嚴重的不對稱性會產生大量的二次諧波失真,這種失真會提前達到穩定狀態並持續到負電網電壓擺幅達到大約 -3V。 此時,峰峰值板電壓擺幅幾乎達到極限。

當負柵極電壓擺幅超過-3.5V時,正極板電壓擺幅從溫和壓縮轉變為急劇削波。 輸出波形變得更加對稱,三次諧波失真開始占主導地位。 超過該點,輸出信號達到最大值 - 它無法響應輸入信號幅度的進一步增加。

As the input signal level increases, positive grid voltage swings are compressed by the effects of grid current on the upstream resistor network. Negative swings of the input signal run free. The severe asymmetry produces bucketloads of 2nd-harmonic distortion that plateaus early and lasts until negative grid voltage swings reach approximately -3V. At that point, the peak-to-peak plate voltage swing has almost reached its limits.

When negative grid voltage swings exceed -3.5V, positive plate voltage swings transition from mild compression to sharp clipping. The output waveform becomes more symmetrical and 3rd-harmonic distortion begins to dominate. Beyond that point the output signal is at its maximum - it cannot respond to further increases in input signal amplitude.

Traditional Waveshapers (Waveshaper Modeler)

傳統上,電吉他音箱的非線性行為是通過『Waveshaper』來建模的,它將輸出描述為輸入的函數。 例如,這是 Line 6 專利中包含的『波形整形器』(Waveshaper)。

Electric guitar amp nonlinear behavior has traditionally been modeled by a 『Waveshaper』, which describes the output as a function of the input. Here, for example, is a waveshaper included in a Line 6 patent.

對於線性電路,傳遞函數是一條直線。 Line 6 Waveshaper在正負輸入擺幅的極端情況下顯示壓縮。 它還描繪了不對稱失真,因為正電壓擺幅的曲線不是負擺幅曲線的鏡像。

For a linear circuit, the transfer function is a straight line. The Line 6 Waveshaper shows compression at the extremes of positive and negative input swings. It also depicts asymmetric distortion, because the curve for positive voltage swings is not the mirror image of the curve for negative swings.

聊到這裡,估計非工科的朋友已經暈倒了,所以我們就此打住,關於Waveshaper建模,後期有更多改進方法如“採用高通濾波器來捕捉耦合電容器 CG 的低頻行為“。
以及 “Cubic Spline Interpolation of the Transfer Function“、” A Digital High-Pass Filter to Emulate the Coupling Capacitor“ 等作法。

It is estimated that non-engineering friends have fainted, so let's stop here. Regarding Waveshaper modeling, there were more improvement methods in the later stage, such as " A high-pass filter that captures the low-frequency behavior of coupling capacitor CG".
And “Cubic Spline Interpolation of the Transfer Function“, ”A Digital High-Pass Filter to Emulate the Coupling Capacitor“, …etc.

The origin of Modeler

在2000年時,Line 6發表PDO,由於主打Amp Modeling的概念,從此綜合效果器就有了新的名詞定義~Modeler。

In 2000, Line 6 released PDO. Because of the concept of Amp Modeling, multi-effects has a new definition of the term ~Modeler.

Black Box & White Box 

2007年,Fractal Audio發表Axe-Fx,採用ADI的SHARC,運算力足夠實現『IR卷積』以及『白盒算法』。

In 2007, Fractal Audio released Axe-Fx, using ADI's SHARC, the computing power is enough to realize "IR convolution" and "white-box algorithm".
In 2010, Kemper released Profiler, which uses the "black-box algorithm" to do modeling. (Let's call it Profiling.)


一般來說,工程師不需要程式代碼應用、內部結構和編程知識的具體知識。 測試人員知道軟體應該做什麼,但不知道它是如何做的。 例如,測試人員知道特定的輸入返回特定的、不變的輸出,但不知道軟體首先如何產生輸出。

Specific knowledge of the application's code, internal structure, and programming knowledge. In general, is not required. The tester is aware of what the software is supposed to do but is not aware of how it does it. For instance, the tester is aware that a particular input returns a certain, invariable output but is not aware of how the software produces the output in the first place.



The black-box algorithm is to guess the working status of the whole machine by virtue of the signal input/output relationship without knowing the internal structure of the physical amplifier or effector. When the correct rate of guessing reaches a certain level (for example, higher than 98%) That is to say, the sound of this amplifier is realized.

The advantage of the black-box algorithm is that it has a simple structure and does not require high computing power. Generally speaking, the black-box algorithm uses a simple general-purpose structure to realize the simulation of sound. Under the same computing power, black-box algorithms can usually obtain good accuracy. His main shortcoming is that it cannot cope with the real-time change of tone when the internal mechanism of the system changes. For example, amplifiers and effectors have adjustment knobs such as volume, high, mid, bass, gain, ...etc. When profiling a model with the black-box algorithm, the above knobs need to be fixed at one position, and then the input and output characteristics of the signal are tested to obtain the parameters of the model. Once the position of the knob is changed, the relationship of the model to the new system does not hold. Therefore, this method is suitable as a method of taking pictures, and the sound of the current amplifier is fixed to use.



白盒測試的基本過程要求測試人員對被測試的源代碼有深入的了解。 程式員必須對應用程式有深入的了解,才能知道要創建什麼類型的測試用例,以便運用每一條可見的路徑進行測試。 一旦理解了源代碼,就可以對其進行分析以創建測試用例。

The white-box algorithm is to open the shell of a device to see what components are used inside and how the components are connected and then uses a set of mathematical equations to realize the working state of the entire machine.

White-box testing's basic procedures require the tester to have an in-depth knowledge of the source code being tested. The programmer must have a deep understanding of the application to know what kinds of test cases to create so that every visible path is exercised for testing. Once the source code is understood then it can be analyzed for test cases to be created.


When choosing to build a virtual device, you must first determine what physical system to use as the basic conditional constraint. For electronic networks, you only need to follow Newton's law and Kirchhoff's voltage and current laws:


The above laws can satisfy the expression of the general resistance-capacitance circuit network. If it is necessary to introduce magnetic field components such as transformers and inductance coils, Faraday’s electromagnetic law can also be introduced:


Arbitrarily complex circuit topologies can be described through the above four laws. Since this way of expressing the system does not involve any artificial influence, a very natural signal response can be obtained, which is the charm of the white box algorithm.
Below we use two examples to explain how to solve the above system. The figure below is a classic circuit for mic preamp mid-EQ. The circuit includes three linear components: resistors, capacitors, and inductors:



1. 要為每一個類比元件建立單獨的數學模型,例如:電子管,晶體管,運算放大器,電阻,電容,電感,變壓器等元件,以及類比元件的電路是如何連接的,因此工作量十分巨大。
2. 需要設計者俱有非常深層次的類比電路設計理論知識才能重現原型機的音色。 
3. 需要大量的DSP算力來達到模型的高精確度。

I know it's hard to go on here, let's just talk about the characteristics of the white-box algorithm.

The advantage of the white-box algorithm is that it can completely reflect all the characteristics of the simulated equipment, including the adjustment knob of the equipment, the influence of parameters such as power supply and temperature on the sound, etc. It also achieves the same dynamic response (what players call playability) and highly realistic tone as simulated equipment. 

The main disadvantages are:
1. It is necessary to establish a separate mathematical model for each analog device, such as tubes, transistors, operational amplifiers, resistors, capacitors, inductors, transformers, and other components, as well as how the circuits of the analog devices are connected, so the workload is huge.
2. It is necessary for the designer to have a very deep theoretical knowledge of analog and electrical design in order to reproduce the sound of the prototype.
3. A large amount of DSP computing power is required to achieve high accuracy of the model.


The picture above is the program mesh structure modeled by the NUX brand using the "white-box algorithm".

The definition of Modeler & Profiler

由前面的介紹,我們可以了解Amp Modeler可分為 Modeler與Profiler。 這邊我們再補充一個新的做法,那就是『AI類神經網路』。

2020年,Neural DSP發表Quad Cortex,採用的就是AI算法,AI算法是基於『黑盒算法』的延伸,一樣無需知道系統內容,量測輸入、輸出信號的關係。 AI在於,不需要測試人員進行測試,而是自主測試,並且根據假設的系統方程進行訓練來獲得最逼近reference輸出結果的方程式。

From the previous introduction, we can understand that Amp Modeler can be divided into Modeler and Profiler. Here we add a new approach, which is "AI neural network".

In 2020, Neural DSP released Quad Cortex, which uses the AI algorithm. The AI algorithm is an extension based on the "black-box algorithm". It also does not need to know the system content to measure the relationship between input and output signals. AI lies in the fact that testers are not required to test, but are tested independently, and trained according to the assumed system equations to obtain the equations that are closest to the reference output results.


The TONE X recently launched by IK is also based on AI. (IK named it MACHINE MODELING.)

總結上述,Amp Modeler分為Modeler與Profiler,其中Modeler的技術又分成Waveshaper與White-box。 Profiler的技術分為Black-box與AI。

To sum up the above, Amp Modeler is divided into Modeler and Profiler, and the technology of the Modeler is divided into Waveshaper and White-box. Profiler technology is divided into Black-box and AI.

1. Waveshaper Modeler: 如LINE6、 ZOOM、 Strymon Iridium、 Walrus ACS-1、 BOSS IR-200, …etc.

採用Waveshaper的考量是,工程師不需要具備真空管電路數學模型的知識背景與工作原理,建立的Amp Model數學模型較簡單,不耗費DSP運算力。 
但缺點就是無法還原真空管音箱的負反饋與動態響應,進而影響手感。 最明顯的就是當你外接『過載效果器』時,聲音會被壓縮,縮掉。無法像真實音箱承受大信號。

The reason for using Waveshaper is that engineers do not need to have the knowledge background and working principle of the vacuum tube circuit mathematical model. The Amp Model mathematical model established is relatively simple and does not consume DSP computing power. 
But the disadvantage is that it cannot restore the negative feedback and dynamic response of vacuum tube amplifiers, which will affect the playability. The most obvious thing is that when you connect an "Overdrive pedal" externally, the sound will be compressed and shrunk. It cannot withstand large signals like real tube amplifiers.

2. White-box Modeler: 如Fractal、 NUX。

如前面所提到,白盒算法考驗著工程師要對系統(也就是真空管音箱電路)有高度的知識背景,且Programmer也要理解系統物理建模的數學模型。 再者每一層級的電路都需要進行量測校正,所以需要拆解原型機來做驗證,建模的過程相當耗時。

As mentioned earlier, the white-box algorithm tests that the engineer must have a high degree of knowledge background on the system (that is, the tube amplifier circuit), and the Programmer must also understand the mathematical model of the physical modeling of the system. In addition, the circuits at each level need to be measured and corrected, so the reference needs to be disassembled for verification, and the modeling process is quite time-consuming.
But for musicians, the most important thing is that White-box is the closest to the "playability feedback" and "overdrive signal tolerance" of tube amplifiers, as well as the "chaos phenomenon" when turning the EQ knob.

3. Black-box Profiler: 如Kemper、 Mooer。

建模速度快,所以品牌方或者第三方,都可以進行Profiling,並且分享Profile檔案。(Amp Model) 但Profile缺點就是靜態參數,當你轉動EQ時,會跟reference有所差異,且你會發現雖然整體頻響跟原型機很像,但是彈奏手感都是定型的,即使你換了不同Amp Model。 當你外接過載效果器時,聲音也會產生不自然的壓縮。

The modeling process is fast, so the brand or the third party can perform Profiling and share the Profile file often. (Amp Model) 
But the disadvantage of Profile is the static parameters. When you turn the EQ, it will be different from the reference, and you will find that although the overall frequency response is similar to the reference, the playability is fixed, even if you change to different amp models. When you plug in an overdrive, the sound will be unnaturally compressed.

4. AI Profiler: 如Neural QC、 IK ToneX。

建模速度最快,其內建通用模型,AI自主訓練擬合曲線方程,Profile的過程較傳統黑盒子簡單。 整體頻率響應可以比傳統Profiler精準,但目前的AI Profiling時間還是相當長,如Neural QC約要40分鐘,一般用戶還是很少使用這個功能。 另外跟黑盒子算法一樣的缺點:彈奏手感都是定型的,即使你換了不同Amp Model。 當你外接過載效果器時,聲音也會產生不自然的壓縮。

The modeling speed is the fastest, its built-in general model, AI self-training fitting curve equation, and the process of Profile is simpler than a traditional black box. The overall frequency response can be more accurate than the traditional profiler. 
But the current AI profiling time is still quite long, for example, Neural QC takes about 40 minutes, and ordinary users still rarely use this function. 
In addition, the same disadvantage as the black box algorithm: the playability is fixed, even if you change to a different Amp Model. When you plug in an overdrive, the sound will also be unnaturally compressed.

Conclusion & My 2 cents

現代吉他手選擇Modeler的機會越來越高,因為演出時遇到Amp-less(無音箱)的場景越來越多。 喜歡單顆效果器組盤的吉他手會選擇Stomp-box Modeler類型來搭配自己喜愛的單顆效果器,尤其是『Overdrive』與『Distortion』。
The chances of modern guitarists choosing Modeler are getting higher and higher because there are more and more Amp-less (no speaker) scenes during performances. Guitarists who like effect pedals will choose the Stomp-box Modeler type to match their favorite effectors, especially "Overdrive" and "Distortion".

所以,如果你的考量是以live gigs為主,那麼挑選『白盒算法』為基底的NUX Amp Academy 將會是一個好選擇。 如前所述,『白盒算法』擁有近似真空管音箱的『Negative Feedback』、『良好的彈奏手感反饋』、『優異的動態範圍』、『過載效果信號承受性』。 
此外產品本身帶有可獨立切換IR開關的XLR輸出及1/4”(6.35mm)輸出,搭配OTG可以在現場使用手機進行深度調參,Send/Return外接post efx。 
但缺點是沒有傳統MIDI口或者3.5mm MIDI口,如果要使用MIDI踏板的話還需要使用帶有USB HOST的MIDI控制器。

If your consideration is based on live gigs, then choosing NUX Amp Academy based on the "White-box Algorithm" will be a good choice. As mentioned above, the "White-box Algorithm" has "Negative Feedback" similar to vacuum tube amplifiers, "good playability feedback", "excellent dynamic range", and "overdrive signal tolerance". 
In addition, the product itself has an XLR output and a 1/4” (6.35mm) output that can independently switch the IR on/off. With OTG, you can use your mobile phone for in-depth parameter adjustment. Send/Return is connected to an external post effector. 
But the disadvantage is that there is no traditional MIDI port or 3.5mm MIDI port, if you want to use MIDI pedals, you need to use a MIDI controller with USB HOST.

反之,如果你以家裡宅錄為主,希望有比較多的Amp Model,或者第三方做好的音色預設,那麼Profiler會是你的選擇。 畢竟宅錄,對於手感上的要求度會低一些,且錄進去後使用DAW外掛其他效果來修飾音軌是很常見的使用場景,就無須考慮『過載效果信號承受性』。且Profiler聽覺上更像『成音』的音色。(像是唱片裡錄製好的吉他音色。) 

而最近市場上討論度最高的平價Profiler,當屬IK ToneX。 小編的使用體驗上,覺得可惜的部分就是編輯預設的介面跟邏輯跟一般的綜合效果器差異很大。 機器本身只有LED屏,顯示的信息量少,現場調參不友善。 此外沒有Send/Return來外接post efx。 如果你要Capture其他reference的話,還需要另購ToneX Capture。

註記:如果你已經買了IK ToneX,對於編輯音色不太熟悉,可以參考這個教學影片。

If you mainly do home recording and hope to have more Amp Models or presets(patches) made by third parties, then Profiler will be your choice. Home recording has lower requirements on playability and it is a very common usage scenario to use DAW plug-in effects to modify the audio track after recording, so there is no need to consider the "overdrive signal tolerance". And the Profiler sounds more like a "processed sound". (Like the guitar sound recorded on a record.) 

The most discussed affordable Profiler on the market recently is IK ToneX. In the editor's experience, the regrettable part is that the interface and logic of editing presets are very different from those of general multi-efx devices. The machine itself only has an LED screen, which displays little information, and the parameter adjustment is not friendly. In addition, there is no Send/Return to external post-efx. If you want to capture other references, you need to purchase ToneX Capture separately.

Note: If you have bought IK ToneX and are not familiar with editing tones, you can refer to this tutorial video.


Recall, what I call "Overdrive Signal Compatibility".

IK ToneX vs NUX Amp Academy | Twin Reverb

IK ToneX vs NUX Amp Academy | Super Reverb

IK ToneX vs NUX Amp Academy | JCM800

Neural Quad Cortex vs NUX Amp Academy

NUX Amp Academy vs LINE6 HX STOMP | Fender Twin Rvb & Friedman BE-100

Strymon Iridium VS NUX Amp Academy | Drive Pedal Headroom & Response Test

相信聽完以上的對比後,大家應該更理解我所謂的『彈奏手感 Playability』、『動態範圍 Dynamic Range』、『過載信號兼容性 Overdrive Signal Compatibility』。

青菜蘿蔔各有所好,愛其所擇,擇其所愛~ 錢沒有不見,只是變成你喜歡的樣子!

I believe that after listening to the above comparison, you should understand what I call "Playability", "Dynamic Range", and "Overdrive Signal Compatibility".

Vegetables and radishes have their own preferences, love what you choose, choose what you love ~ the money is not gone, it just becomes what you like!

結尾附上小彩蛋~ IK TONEX與NUX AMP ACADEMY個別的耗電量:

Amp Academy Firmware V4

(2024/04/04 更新blog)
最近NUX又更新了Amp Academy,我相信大家最關注的不外乎有哪些新的Amp Model?

1. Fender Deluxe Reverb (Deluxe Rvb搭配 IR "FDR112 BLU")

 2. VOX AC30 (Classic A30 搭配 IR “A30 212 BLEND”)

 3. Fender Princeton (Princetone搭配 IR “JEN P10R")

4. Mesa Boogie Lonestar (LSTAR 搭配 IR ”MES 212 V30”)

5. Marshall JCM2000 (Brit 2000 搭配 IR “M1960TV”)

6. Peavey 5150 (VIVO 搭配 IR ”PV412")

除了新的Amp Model以外,也釋出旋鈕鎖定功能,這個對於現場演出避免誤操作很實用!

Comparison with Kemper Profiler Player

這邊都是使用相同的第三方IR檔,錄音的方式都是採用USB RE-AMP。 盡量做到最小的差異性,音色也盡量調到接近。 不過兩者的Gain跟EQ屬性不太相同,不容易調到完全一至。 至於誰好誰好,就由看官們自行評論囉~

1. Mesa Boogie Dual Rectifier

2. Friedman HBE-OD

3. VOX AC30

4. Fender Deluxe Reverb

5. Marshall JCM800