This is just a comparison of amp models. No speakers or mics involved. It's a measure of how accurate the frequency response of the amp model alone is. Captures/profiles will likely be more accurate same as a Tone Match will be more accurate.That would depend on how the amp sounds were recorded. Microphone choice and position, or direct in. These graphs also don't take into account people's personal captures of the competing product which is its major selling point. I don't think they are trying to compete with Fractal as much as they are trying to bump Kemper out of the way. I had a side by side setup with the new against my kemper. Sold the kemper. Had the intention of keeping both my Fractal and the new one but after finding how far behind the effects were, I dumped it. Fractal alone now.
In fairness the accuracy above 10 kHz isn't terribly important as the cab filters out most of that anyways. The critical band is 50 Hz - 10 kHz.That cliff-drop (pun not intended!) in the blue curve looks to me like a low-pass filter being engaged somewhere in the amp model.
In fairness the accuracy above 10 kHz isn't terribly important as the cab filters out most of that anyways. The critical band is 50 Hz - 10 kHz.
Not my experience with the PS-100 at all. I find it has a very marginal effect if I run my real tube amps through it vs running direct to a cab at the same volume. The line out to me sounds no different from my Bluetone Loadbox which like the PS-100 has a very marginal effect on the behavior of the real amp. When used with a modeler to the PS-100 line in I find that turning off the speaker impedance modeling on the FM3 helps it behave more like tube amps do when hooked up to it. Otherwise you can get weird low end response or overhyped lows and highs. I use a somewhat uncommon 4x10 cab so this might be less noticeable on a 4x12 that is used by most amp models as the speaker impedance curve.I know what you say is generally true, but the cumulative effect of so many high-cuts is often noticeable to me. Case in point - I bought a Fryette Powerstation 100 recently. To put my modellers through it. But I noticed the amp models were not as bright and aggressive as my real amps, nor as aggressive as the units sounded going direct to my Apollo.
So as it turned out, the PS100 filters a lot of high frequencies. So no only did I have the cab to contend with, but also the poweramp. It sounded too dark.
I sent it back in the end, and got a Suhr Reactive Load IR, which has a hi-cut function that you can disable. Much more to my tastes. I often feel that the aggressive 'take your face off' aspects of a cranked Marshall amp can have a lot of info in the 10-14kHz region. I've never measured it though, so I could be talking horsepucky.
Also the roll-off begins around 8kHz to me on the blue curve; way too low for me!
Not my experience with the PS-100 at all. I find it has a very marginal effect if I run my real tube amps through it vs running direct to a cab at the same volume. The line out to me sounds no different from my Bluetone Loadbox which like the PS-100 has a very marginal effect on the behavior of the real amp. When used with a modeler to the PS-100 line in I find that turning off the speaker impedance modeling on the FM3 helps it behave more like tube amps do when hooked up to it. Otherwise you can get weird low end response or overhyped lows and highs. I use a somewhat uncommon 4x10 cab so this might be less noticeable on a 4x12 that is used by most amp models as the speaker impedance curve.
Suhr RL models its reactiveness around a specific 4x12 cab impedance curve so maybe that's what you find more to your liking?
I need that as a bumper sticker!I can t understand nothing about everything
The continued mentioning of "sweep frequency" still has me mildly confused. I thought you were hitting it with a fixed-frequency impulse and then watching what happened through the full range of 20-20k. It appears the stimulus frequency has dropped lower, to allow more harmonics to reveal themselves. Perhaps you are hitting it with a narrow-ish sweep rather than single-frequency imuplse and that's my only point of confusion?Here's another comparison between the real amp (green), Axe-Fx (purple) and Product B (blue). I've reduced the sweep frequency so the first three harmonics are captured. The reason the Axe-Fx looks "better" than the amp is because there's less thermal noise than the real amp.
View attachment 89068
It's used quite a bit on these forums as well. Just do a search for the term. I have no idea what it means but others sure use the term a lot.ya the term harmonic content is about as useful as saying a left handed screwdriver. I have no bones with people who say "in this particular instance from the samples provided, A sounds better than B to my ears based on the current settings. " But to jump to conclusions and say that one product is better based on that is asinine.
It's a narrow band sweep, i.e. 6-8 kHz. If you use just a single frequency the aliasing products can be higher or lower so by sweeping over a small range you excite all the possible aliasing products.The continued mentioning of "sweep frequency" still has me mildly confused. I thought you were hitting it with a fixed-frequency impulse and then watching what happened through the full range of 20-20k. It appears the stimulus frequency has dropped lower, to allow more harmonics to reveal themselves. Perhaps you are hitting it with a narrow-ish sweep rather than single-frequency imuplse and that's my only point of confusion?
There are multiple ways to model an amp. The Fractal way involves modeling the individual components in the circuit.This is as absolutely fascinating. Gripping, almost.
I have a question and "prefer to take my answer offline." In other words, not attempting to derail this thread with this question:
@FractalAudio, I've been extremely curious as to how amp modeling actually works. Without asking you for any proprietary/IP details, can you point me to a resource explaining how this is actually done?
At first I guessed it would be like an amp version of a cab-IR shot, but I suppose that would only provide a model of the amp in one state. How do you incorporate all the different amp settings and differences in caps, pots, transformers, etc.? Is it approximation based on, you know, "graphs and charts" or do you have to sweep the controls while recording the amp's response metrics?
Or is there a framework of a tube circuit that you kinda fill in the data like a driving sim might approximate a car's performance allowing for tire grip, weight/balance, surface grip, etc. in their physics modeler? I can only imagine but the huge mystery to me is how the final product includes tweakable parameters while still being true to the amp being modeled.
If it's too technical I won't get it, but at the moment (as you can see) I cannot even speak intelligently on the issue because there is so little I actually know about the paradigm.
You don't understand the meaning of the graphs. Ideally there should be NOTHING between the plateaus. Anything between is noise and/or aliasing. The amp has lots of thermal noise (hiss). The Axe-Fx has less thermal noise and nearly no aliasing.
Is this the result of a sweep or a single frequency? If a sweep, you don't mean 20-20k but something narrower?Here's another comparison between the real amp (green), Axe-Fx (purple) and Product B (blue). I've reduced the sweep frequency so the first three harmonics are captured. The reason the Axe-Fx looks "better" than the amp is because there's less thermal noise than the real amp.
View attachment 89068
Reread the first paragraph of post 152.Is this the result of a sweep or a single frequency? If a sweep, you don't mean 20-20k but something narrower?
I think "sweep" might be confusing some commenters into thinking this is a full spectrum response...
Higher oversampling is the most common approach. There is a technique known as anti-derivative anti-aliasing. The problem with that is it really only works with waveshaper approaches. You use the anti-derivative of your waveshaper function and then take the derivative of the result.What is done to improve aliasing performance? Is it as simple as higher oversampling or is it more complicated under the hood?
I posted this before I got to that one...Reread the first paragraph of post 152.
d/dx of the integral of f(t)dt from 0 to t (that's hard to type out) is just f(x) isn't it? What's the benefit of doing it that way over just using f(x), does it sort of "normalize" the data?You use the anti-derivative of your waveshaper function and then take the derivative of the result.