They do use IRs. They use 256 point IRs augmented with parametric EQ. This is easily demonstrated by examining the data structure of the profile and by measuring the device (or by using their IR converter which converts long IRs to 256 point IR plus EQ). What they say is "we use something more advanced" or something like that. By "more advanced" they mean they augment the IR with parametric EQ. However they must augment the IR with parametric EQ because the IR is too short to reproduce low frequencies accurately. The frequency resolution of an IR is proportional to its length: the longer the IR the better the resolution. To accurately reproduce a guitar cabinet requires at least 1000 points from my research. (Note: I'm using the term "parametric EQ" as a substitute for IIR filter since most people don't know what an IIR filter is.)
This technique has been around since the early days of modelers (I believe the Pod 2.0 was the first to do this). The impetus for this is that it uses much less processing power since the amount of processing power required is directly proportional to the length of the IR. A 256 point IR plus, for example, an 8-band EQ can be equivalent to a 512 point IR but uses less processing power. The downside is the phase response isn't the same but that is usually inaudible.
In their specific case the amplifier output frequency response and cabinet frequency response are combined into a single IR plus EQ since they aren't measured separately. See my MIMIC whitepaper for more information on frequency response, etc. You can make assumptions about the amplifier output frequency response in order to separate it from the cabinet response. In their case I believe they assume there is a 6 dB resonance at 125 Hz plus a 6 dB highshelf (incidentally this is the same power amp frequency response model that the Marshall JMP-1 used albeit using analog filters). For many amplifiers this is a reasonable approximation.
The Axe-Fx II IRs are 2048 sample (eight times the resolution) and don't require augmentation.
Wow, great explanation. I had a feeling this is what was going on. I wonder why with this method some people favor the Kemper tones to Axe FX. For example, you will see in many threads people saying that the 'core amp tones' are better in the Kemper than the Axe. I am going to go out on a limb and guess most people that say that either A) Have never used an Axe FX or B) It is all in their heads. It is so funny to me when users or the Kemper correct people when they refer to the Kemper as a modeler, lol. Most times they say , Kemper doesn't user Impulse responses like all other modelers, it uses profiles and isn't a modeler. It is also funny on the Kemper website where they say the Kemper is able to recreate certain high gain sounds where impulse responses fail miserably. So glad I decided on the Axe FX II over the Kemper. I will be honest, after reading all the threads where people say the Kemper is better than the Axe FX at pure amp tones, I have been GASing a little bit for a Kemper. Now after reading Cliff's post, I have no desire whatsoever to get one. Seems like the Axe does everything the Kemper does and more. Thanks again for that post Cliff.
I also wonder why he says it took him several years to come up with the profiling if it is something that has been around for awhile. Sounds like a bunch of marketing hype to me. Seems like they took a old method and had some great marketing.
Wow, great explanation. I had a feeling this is what was going on. I wonder why with this method some people favor the Kemper tones to Axe FX. For example, you will see in many threads people saying that the 'core amp tones' are better in the Kemper than the Axe. I am going to go out on a limb and guess most people that say that either A) Have never used an Axe FX or B) It is all in their heads. It is so funny to me when users or the Kemper correct people when they refer to the Kemper as a modeler, lol. Most times they say , Kemper doesn't user Impulse responses like all other modelers, it uses profiles and isn't a modeler. It is also funny on the Kemper website where they say the Kemper is able to recreate certain high gain sounds where impulse responses fail miserably. So glad I decided on the Axe FX II over the Kemper. I will be honest, after reading all the threads where people say the Kemper is better than the Axe FX at pure amp tones, I have been GASing a little bit for a Kemper. Now after reading Cliff's post, I have no desire whatsoever to get one. Seems like the Axe does everything the Kemper does and more. Thanks again for that post Cliff.
The core amp tones in V10 are superior IMO. Our modeling captures both preamp and power amp distortion. Their modeling lumps it all into one and attempts to fit a (k-x)/(k + x) waveshaper using difference minimization. See the papers by Klippel for more information on model fitting. V10 uses the same technique but fits true triode models to the preamp measurements and true push-pull tetrode/pentode models to the power amp measurements. If you listen to their models you'll never hear any crossover distortion because they don't model it. If you don't like crossover distortion then this is irrelevant, however many modern, aggressive amp designs exploit crossover distortion. This is probably why the Axe-Fx II has a better reputation for high-gain tones.
I wonder why with this method some people favor the Kemper tones to Axe FX.
To me, this is a good example of ears over eyes. Regardless of math and algebraic formulas, the Kemper sounds phenomenal. Whatever is happening in terms of IR's in the KPA, it certainly hasn't had a negative impact on the tone.
And this is not a knock on the Axe II, because we all know that it sounds phenomenal too. But don't let others influence your decision simply based on numbers and theories. Give them both a try if you can and keep whichever suits you best. I have both in my rack (along with an Eleven Rack) and find things I love about each. The Kemper may have flaws, but great tone is not one of them.
I don't have any engineering knowledge, but I'm candidly curious about one thing : a profile, which contains the complete rig settings including the effects and the IR, has a file size of only 4 KB. A standard 1024 points IR is 7 KB. Question : is there a way to "compress" an IR, like a .mp3 is to a .wav ? IRs are audio files, after all...They do use IRs. They use 256 point IRs augmented with parametric EQ. This is easily demonstrated by examining the data structure of the profile and by measuring the device (or by using their IR converter which converts long IRs to 256 point IR plus EQ). What they say is "we use something more advanced" or something like that. By "more advanced" they mean they augment the IR with parametric EQ. However they must augment the IR with parametric EQ because the IR is too short to reproduce low frequencies accurately. The frequency resolution of an IR is proportional to its length: the longer the IR the better the resolution. To accurately reproduce a guitar cabinet requires at least 1000 points from my research. (Note: I'm using the term "parametric EQ" as a substitute for IIR filter since most people don't know what an IIR filter is.)