Digital Modelers: Methodology comparison

    While I would take exception to some parts I agree with MM that it did take a lot of work and includes some good insights on both units and is worth a read. The two areas I would question are criticizing frequent firmware updates for the AE II and criticizing FA for lack of firmware updates on the older units and calling the UR IR's snake oil. I do think that most AE II users can't tell the difference but one user posted that he wasn't using AE for its IR's as the sound wasn't as good as a plugin runnig the IR. He took a lot of flack from the AE fanbois but Cliff did some tests and confirmed that it was a programming error. Even though just one person caught it, there is a measurable difference and so I would disagree with the snake oil description.

    Thanks for the comments. With regards to the Fractal Ultra-res IRs. I only call it snake oil because if indeed Fractal thinks it makes a difference for the IRs to be in length beyond 22ms, then he needs to be using the true IR representation past that length. But according to published reports, the IRs are manipulated beyond a certain frequency and decimated and downsampled. So any accurate representation is compromised because of the need to reduce the computation needed by convolution for such lengths due to hardware limitations in the Axe-FX.


    There is a perfect explanation for why a DAW plugin running a longer length IR may sound "better" than one truncated and running in the Axe-FX. The longer one is clearly capturing aspects not only of the cab but the environment and room. This may subjectively sound "better." However, take the same truncated IR and use the Axe-FX or some other plugin, then post-process with a high quality suitable reverb (allowing room reflections, early reflections, etc.) and you might be surprised at the result.


    You can use IRs to capture the cab, the room, or both. I believe cab IRs don't need to be as long as Fractal says to capture the cab sound contribution. Fractal do argue that using longer IRs allow better low frequency resolution for stuff like 80 Hz and below. The low E string is at 82 Hz on a guitar. However, it's common when recording to employ a high pass filter at such frequencies on electric guitars to clean up muddiness. Even if there's some energy at that those frequencies, I don't think they contribute significantly to the positive aspect of the sound.


    All this is debatable I believe (sound often is!), and I have not seen any definitive analysis on the subject. The key point is that if you are going to use IRs that go out to 170 ms like Fractal says, then they should be the true representation. For Fractal, they aren't.


    Personally for me, using 20ms IRs for the cab and other post processing SW for the reverb work fine and sound great. Just listen to the Kemper! :) Or listen to a Lexicon reverb. There are plenty of reverbs that actually sound better than convolution reverb. Lexicon reverbs are proof of that. to And in the end -- it's the sound that matters -- not necessarily recreating the sound exactly like it would sound in some specific acoustic space, which is what convolution reverbs do best.

    Agreed on all counts, 'junkie.


    The Lex algorithms were constructed over many years, decades even, based upon what sounds good, whereas real-world-based IRs are snapshots of physical spaces where at best the ability to tune is limited.


    If anything IMHO, we're better off with the short cab IRs 'cause the less the environment is baked into the "cab sound", the more flexible our options as far as placing the rig in any space we'd like with IRs, algorithmic 'verb or early reflections are concerned, become.

    Hey Ampjunkie. Most of your explanation is well above me! I probably did not make myself clear about the IR's. The AE hardware now matches what you would hear with a plugin with the IR. A programming error was playing the UR IR's as standard IR in AE (I forget the finer details). Only one AE user could hear the difference between the AE with the programming error vs the plugin but there was enough of a difference that he could hear it which makes me believe that the UR do carry more information that is useful to some users. It may not be useful for you personally but that doesn't make it "snake oil". I re-read your article and really enjoyed your (well informed) speculation as to how each company is going about doing what they do. Interesting and well thought out stuff. It is only a matter of time before it gets picked up on the AE side and will be fun to read those comments! Especially CK's comments regarding Cliff's comments. And of course the fanboi's who will disagree with your conclusions. They are a bit wrapped up in a debate over the Helix vs AX8 right now but give it some time.

    Great article.
    I think CabMaker is a tool to convert IR to IIR filter coefficients.Some time ago i was using Matlab do do such transformation. Thats why the Cab files are so small. IIR filters require less computational power and less parameters then IR. But this is only my guess.

    Quote from DamianGreda

    Great article.
    I think CabMaker is a tool to convert IR to IIR filter coefficients.Some time ago i was using Matlab do do such transformation. Thats why the Cab files are so small. IIR filters require less computational power and less parameters then IR. But this is only my guess.


    Yes, it isn't clear from Kemper's statements or from the available data if Kemper is using IR convolution for the cabinet block or some other method.


    The irony is that so many people are caught up with reproduction accuracy, etc. of amps, cabs, etc. I would say that this is not hi-fi reproduction of a live symphony! If the sound can come very close to the original amp and cab -- or perhaps even eclipse it -- then who cares if it's an exact reproduction?


    Fractal's mantra is to continouusly strive for more accurate modeling. But if you think about it, this should not always equate to better sound. I can accurately model an amp and it may still sound crappy. Not every Fender, Marshall, VOX, or Mesa amp sounds good ...


    Kemper's goal is to reproduce a great sound from an amp and cab that exists and you can hear it. This goal to me is more relevant than more accurate circuit models.

    Hey Ampjunkie


    I had a post with Cliff about this once on the Fractal forum and at least then, his stated goal was not to continuously strive for more accurate modeling but rather, again at the time, he said that he removed elements in the modeling that were present in the actual amp that made it sound bad. I use the same screen name there so you can look it up if you are interested. He also has created a number of virtual amps that don't exist in the real world. I would actually say that Kemper would be the one to more faithfully capture an amp at those settings warts and all while Fractal could crest an idealized version based on an amp they like.

    Just saw this ...


    http://forum.fractalaudio.com/…on/105021-ax-8-price.html


    At this price point, 8-10 effects and one amp/cab path -- this is competitively priced $100 less than the Helix. It may come down to how close Line6 can get their sounds to the Axe-FX. If it's close, than the Helix with it's better GUI, expression pedal, and routing may be worth it. If the sounds are mediocre, then the AX8 would be the better choice.


    Axe-FX II XL+ $2200 USD
    Kemper $2000 USD
    Floor models:
    AX8: $1399 USD
    Line6 Helix: $1499 USD


    The digital modeling landscape is certainly heating up -- and the consumers win!


    In those prices, remmber that Fractal is an American company, and Kemper is a German one; which means that in Europe, the price difference is much larger.

    Quote from SwAn1

    Hey Ampjunkie


    I had a post with Cliff about this once on the Fractal forum and at least then, his stated goal was not to continuously strive for more accurate modeling but rather, again at the time, he said that he removed elements in the modeling that…


    I stand corrected, though I recall in the beginning it was all about accurate modeling. It isn't surprising that this has shifted. Fractal need to introduce elements and variations in the components in order for the tone to sound better. But in doing so, sometimes old patches have to be recreated.


    Please note that I am not saying that the Axe-FX is not capable of great sounds. It clearly has great sounds and effects. But if you read my analysis, I believe Kemper has the edge in flexibility in tone generation because of how he achieves it through black-box modeling as opposed to modeling via circuit simulation and component modeling. There are far more amps that are able to be profiled than can be modeled in FW. And because it's also done at a higher level, I also think Kemper can recreate sounds more efficiently. This is speculation on my part, however, but the HW on the Kemper vs the Fractal seems to bear this out.


    Thus, I am not debating the sounds or tone; rather, I am only saying I prefer the Kemper methodology as well as GUI and layout. I also like all the free profiles on the RigExchange, as well as the commercially available profiles from many sources. These seem to be a great bargain compared to buying Ultra-Res IRs from only a single source. :)

    No need to explain your findings Ampjunkie. It is an analitical comparison based on facts and speaks for itself.


    Slowly but steady the consumers will see the difference in approach and the difference in business conduct.


    So keep doing the good work!