Modelers and Aliasing

[Frankencat]

As an aside, if you're interested in ML researchers with SP backgrounds, check Joel tropp, Ben recht, Joan Bruna, Carlos Fernandez granda, Stephanie mallat.

I'll bow out now because I clearly don't know enough about the audio domain, eek.

You and me both brother. Even though I am pretty well versed in the software domain this audio stuff is over my head. But I am learning from this thread so thank you to the smart guys for that.

 

[Audio Transformer]

When I received my FM9 some months ago, I thoroughly compared it to my axe fx 3 before selling the latter, and one of the comparisons I made was about the aliasing performance. I haven't posted the results till now on other forums to avoid starting a flame war among both fanbois and haters, since these findings somewhat contradict what fractal has officially (un)said about this, and I perfectly understand that they avoid disclosing these details for those same reasons, on Fractal forum and TOP this would be blown out of proportion for sure.
But since we are on TGF and this interesting thread came out, here we go:

I used the method proposed by Cliff to measure the aliasing, feeding a 9-11 kHz sine sweep to the same exact amp model (VH4) and corresponding FWs between the two units. I also tested if there was a difference between having just 1 amp block vs 2 on the axe fx and it turns out there is indeed.

Axe FX with one amp loaded
View attachment 5352


Axe FX with two amps loaded (only one measured)
View attachment 5353


FM9
View attachment 5354


It's pretty clear and expectable that the one having less aliasing is the Axe FX with just one block loaded in the preset, it probably uses the whole 1 GHz horsepower of one of its core to process the amp alone, with an overkill amount of oversampling to reduce aliasing at minimum.
When loading two amp blocks that horsepower has to be shared between those two, so the oversampling rate is probably cut in half and the aliasing gets louder.
The FM9, having basically half the processing power of the Axe FX but split in 4 cores instead of 2, has to cut the oversampling in half again so it has the highest amount of aliasing among the three. And for the reason that the two amp blocks each have a dedicated core, running one or two amps in a preset doesn't make a difference.

The aliasing of the FM9 might seem quite high from these graphs since it's "just" 36 dB lower than the fundamental frequency, but it isn't really, cuz this test was made with ridiculously high amounts of gain to make the aliasing more visible in the RTA graph, something that nobody would likely ever use in real world (VH4 ch.3 with all the knobs cranked), and the dry sine sweep was at -12 dBFS, a level that probably no guitar DI signal can reach at those frequencies.
Making this test with more "real-world" settings and signals would push the aliasing below -60dB even on the FM9, and in fact I could not hear any difference at all between the two devices in my presets, that's why I decided to keep the fm9 and sell the Axe FX in the end, they really sound the same.

Maybe should you make some update on AxeFx3 VH4 vs ToneX vs NAM with TTS or newer one (the one shared on main group lately)
The pink curve is the wav3 of NAM to train the model without anti aliasing. The others are like the evolution and it’s not even including last ones.
So if someone can make the testing I believe it’s you. For some reason NAM Facebook doesn’t really possess any Axefx. I presume they’re different worlds, Axefx is out of price physical device while NAM is open source, totally free plugin

 

[Orvillain]

It could do it, but there'd be tiny little bits of egg shells around the upper limits of the system, and as much as most people would say they couldn't taste it, there would always be some super egg aficionado who would swear down that their analog eggs are way betterer coz they're fewer gritty.

This post should've gone down in history alongside nebcoromponcon.

 

[DLC86]

View attachment 39055
Maybe should you make some update on AxeFx3 VH4 vs ToneX vs NAM with TTS or newer one (the one shared on main group lately)
The pink curve is the wav3 of NAM to train the model without anti aliasing. The others are like the evolution and it’s not even including last ones.
So if someone can make the testing I believe it’s you. For some reason NAM Facebook doesn’t really possess any Axefx. I presume they’re different worlds, Axefx is out of price physical device while NAM is open source, totally free plugin

I will do it at some point but it's not as straightforward as you might think, cuz the anti-aliasing input files don't work if the source already produces aliasing, so making a fair comparison requires having some real gear that is modeled on the axe fx, create all the captures of that gear and compare it to the fractal model after matching its gain, settings and level to the capture... not an easy task.

PS: btw, I'm the one who made that graph and tests, I know it pretty well

 

[JCW308]

Are you using high and low cuts on your IR? Also I believe there would be high and low end fizz on a real amp too if we used a full-range speaker instead of a guitar speaker.

 

[pipelineaudio]

Wouldn't that be too late? As far as I understand it anyhow...It could hide the nastiest of them, but as a nonlinear process, I don't think the IRs would be creating any aliasing effects

 

[DLC86]

Wouldn't that be too late? As far as I understand it anyhow...It could hide the nastiest of them, but as a nonlinear process, I don't think the IRs would be creating any aliasing effects

Yeah, I feel there are quite a few misconceptions about aliasing, also in this thread... it's true that generally aliasing is louder on higher frequencies, but it's not a rule and can't be solved with a simple high cut cuz:

1. It depends on the amount of gain. If you have a lot of gain and hard clipping (pretty much what happens in a crancked fuzz), all harmonics will have pretty much the same level as the fundamental and this is also true for harmonics beyond Nyquist and reflected back, so the aliasing will be distributed quite evenly on the entire spectrum.

2. Putting a high cut filter reduces both the aliasing and the un-aliased signal by the same amount, so the level difference between the two doesn't increase and thus doesn't make it less audible. A high-cut filters out aliasing only when the filter is used with oversampling and applied before downsampling back to the original sample rate.

3. It's true that (as said in the OP) the level of high frequencies, which are those that tipically generate the highest amount aliasing, is quite low in a DI guitar signal, but put that signal thru a fuzz or some crancked amps and harmonics of mid-low frequencies will produce noticeable aliasing too, cuz the "slope" of the harmonics will become flatter as you increase the gain and/or use harder clipping.

Also, imagine what happens when cascading a digital fuzz model into an overdriven amp model (let's say a Plexi, a setup that shouldn't be uncommon at all) with no oversampling:

Firstly, those low-level high frequencies of the DI signal, after passing into the fuzz, will be boosted to a similar level as the rest of the signal.
Secondly, the whole signal, by being heavily distorted by the fuzz, will generate harmonics, intermodulation distortion and aliasing produced by both of these. Let's be optimistic and say that aliasing level is now 20 dB lower than the un-aliased signal.

At the input of the amp we will now have a signal that has about the same amplitude across the whole spectrum, with a lot of harmonics and intermodulation and a quite high aliasing.
The amp will also add its own amount of harmonics, IMD and aliasing, but this time these are not generated by the DI signal alone, but also by the harmonics, IMD and aliasing produced by the fuzz and already in the signal. What level do you think aliasing will now have? For sure quite higher than those -20 dB and it will be a lot more "messy" too.

 

[BenIfin]

Just a side question. Given what we know, is it accurate / correct (?) to rank the "lack of aliasing" [best] as follows:-

1. Axe 3
2. FM9
3. FM3
4. NAM Hyper Accurate Captures
5. NAM Standard Captures
6. Tonex Captures

[ not sure where the Helix fits ]

 

[DLC86]

Just a side question. Given what we know, is it accurate / correct (?) to rank the "lack of aliasing" [best] as follows:-

1. Axe 3
2. FM9
3. FM3
4. NAM Hyper Accurate Captures
5. NAM Standard Captures
6. Tonex Captures

[ not sure where the Helix fits ]

Fm9 and fm3 should be on par, the rest seems accurate but I still haven't tested the latest nam input file.

 

[DLC86]

seem like the guys from the NAM laboratory have found the ultimate solution here.. https://www.facebook.com/share/p/19qGsmkxF7/?mibextid=wwXIfr

He also found a neat workaround to train models at 96 or 192 kHz avoiding tonezone's checks.

 

[BenIfin]

seem like the guys from the NAM laboratory have found the ultimate solution here.. https://www.facebook.com/share/p/19qGsmkxF7/?mibextid=wwXIfr

I'm not a member of that FB Group.

Is the solution to (a) Train using the new "hyper approach" and (b) to do it at 192k - which makes the effective listning / alias free range up to 96k (?)

If so, I assuming (i) that makes the training process a lot longer and (ii) the NAM Captures size file would be massive (?)

 

[Boudoir Guitar]

Hey guys. I can’t hear the difference. Can someone rank them so I know which one to buy?

 

[JCW308]

Hey guys. I can’t hear the difference. Can someone rank them so I know which one to buy?

Get the Kemper and call it a day!

 

[northern_fox]

Hey guys. I can’t hear the difference. Can someone rank them so I know which one to buy?

Whichever one has more tubes

 

[deathbyguitar]

When I received my FM9 some months ago, I thoroughly compared it to my axe fx 3 before selling the latter, and one of the comparisons I made was about the aliasing performance. I haven't posted the results till now on other forums to avoid starting a flame war among both fanbois and haters, since these findings somewhat contradict what fractal has officially (un)said about this, and I perfectly understand that they avoid disclosing these details for those same reasons, on Fractal forum and TOP this would be blown out of proportion for sure.
But since we are on TGF and this interesting thread came out, here we go:

I used the method proposed by Cliff to measure the aliasing, feeding a 9-11 kHz sine sweep to the same exact amp model (VH4) and corresponding FWs between the two units. I also tested if there was a difference between having just 1 amp block vs 2 on the axe fx and it turns out there is indeed.

Axe FX with one amp loaded
View attachment 5352


Axe FX with two amps loaded (only one measured)
View attachment 5353


FM9
View attachment 5354


It's pretty clear and expectable that the one having less aliasing is the Axe FX with just one block loaded in the preset, it probably uses the whole 1 GHz horsepower of one of its core to process the amp alone, with an overkill amount of oversampling to reduce aliasing at minimum.
When loading two amp blocks that horsepower has to be shared between those two, so the oversampling rate is probably cut in half and the aliasing gets louder.
The FM9, having basically half the processing power of the Axe FX but split in 4 cores instead of 2, has to cut the oversampling in half again so it has the highest amount of aliasing among the three. And for the reason that the two amp blocks each have a dedicated core, running one or two amps in a preset doesn't make a difference.

The aliasing of the FM9 might seem quite high from these graphs since it's "just" 36 dB lower than the fundamental frequency, but it isn't really, cuz this test was made with ridiculously high amounts of gain to make the aliasing more visible in the RTA graph, something that nobody would likely ever use in real world (VH4 ch.3 with all the knobs cranked), and the dry sine sweep was at -12 dBFS, a level that probably no guitar DI signal can reach at those frequencies.
Making this test with more "real-world" settings and signals would push the aliasing below -60dB even on the FM9, and in fact I could not hear any difference at all between the two devices in my presets, that's why I decided to keep the fm9 and sell the Axe FX in the end, they really sound the same.

I'm just now seeing this post but you confirmed what I always suspected about 2 amps having more aliasing. Thanks so much for taking the time to do this.