Modelers and Aliasing

[James Freeman]

For testing and discussing modelers and plugins aliasing.

You can use this free signal generator plugin in DAW: https://plugins4free.com/plugin/2801/

The sine tone should be above 11kHz because 2nd harmonic is twice the frequency and is outside the audible band (> 20kHz).
With a real amp you should only see a single peak of the test signal when sweeping from 10kHz to 20kHz.
Real amps don't have aliasing, modelers and plugins will exhibit various degrees of aliasing and you will see and maybe hear frequencies inside the audible band.

I will also point that the 10kHz+ test tone should be at a level that guitar pickups produce.
This is the peak spectrum of an EMG 81 across the entire fretboard, so sending a 11kHz signal at 0.0dBFS into a plugin is not useful nor happens in any real scenario.

 

[James Freeman]

Sending 12.5kHz signal at 1vPeak into my high gain tube amp and other modelers.
Note: NO electric guitar produces 1v at those frequencies, this is only to demonstrate that real tube amps don't have aliasing.
The tested amp and modelers were set with the same amount of high gain that will make Wrestlemania proud.

Real tube amp.
You can see some power supply hum and ripple but zero aliasing at high frequencies, only the pure 12.5kHz signal all the way down to the noise floor.

Helix Native (Vitriol), Hum, Ripple and Sag at 0.
Some aliasing at roughly -60db below.

NAM capture of my Peavey.
A shit ton of very audible aliasing.

Again, the level of the test signal was unrealistically high that no instrument produces.
Maybe if you cascade gain stages that already have frequencies at those amplitudes, then yes, you will see similar levels of aliasing, BUT, if you already have distorted signal with such harmonics you probably already have fundamentals that much higher in amplitude that will MASK those parasitic aliasing harmonics.


You want to talk about masking and the actual audibility of aliasing in high gain tones?

 

[MirrorProfiles]

Also worth bearing in mind that a lot of aliasing will fall below the noise floor.

AND

not all plugins implement oversampling in the same way. This can mean results are different at different sample rates. It can also mean, that when in use, having a big signal chain in one plugin instance may sound different to having each distortion based processing on a separate plugin.

 

[BenIfin]

TONEX SOFTWARE ALIAS TEST

Hi all !

Firstly, thanks to James for setting up this thread and pointing me to the Sine Generator VST D/L.

OK ..... I have *no real* idea if I've done this right or not but here goes - what I did was:-

-> Created a new clean Helix Native Audio Track set to 48k / 16 Bit <- don't know if this is good or bad ?
-> Firstly, inserted the VST2 Sine Generator Plugin from the OP in this Thread and set it to 10097hz <- wouldn't let me select 10000hz exactly
-> Then in that same Track, I secondly added 1 x Tonex Software VST 3 Instance
-> The Tonex VST3 was only running 1 x "Rock / Distorted Dumble" D.I Amp Block Only .... no Cab, no Gate, no Comp, no Rev etc....
-> Then in the same Track -after the Tonex VST3 - I loaded 1 x Voxengo SPAN Instance
-> I opened SPAN
-> I then engaged the VST2 Sine Generator .... and the SPAN display once it settled - never changed from the screenshot below

=> am I even remotely on the right track on how to do this ?

=> does the above screen shot "show" anything of "significance" ?

From Zooming in to the original .jpeg - it looks like to me (?) - and I may totally be mis-reading this - that there is:-

=> at around 18k there is -63db of Aliasing ?
=> at around 12k there is -71db of Aliasing ?
=> at around 9k there is around -66db of Aliasing ?
=> after that I'm not really sure what / how to interpret or what the "spikes" show or mean ?

Happy to be educated if I've done this wrong :)

Thanks,
Ben

 

[James Freeman]

=> am I even remotely on the right track on how to do this ?

Yes, all good.
What output level did you set the signal generator?

--

I am going to get down and dirty with this aliasing debate, input levels, frequency ranges, harmonically rich signal and masking should be talked about.

 

[BenIfin]

Yes, all good.
What output level did you set the signal generator?

--

I am going to get down and dirty with this aliasing debate, input levels, frequency ranges, harmonically rich signal and masking should be talked about.

I ran the Sig Generator Freq at 10097k and the Level bang-on "9.00am" as for some reason that was "the loudest" squeal (?)

I also set the Helix Native Track level so it wasn't anywhere near clipping and this enabled me to run the Capture Volume up full.

Thanks again,
Ben

 

[BenIfin]

Also worth bearing in mind that a lot of aliasing will fall below the noise floor.

AND

not all plugins implement oversampling in the same way. This can mean results are different at different sample rates. It can also mean, that when in use, having a big signal chain in one plugin instance may sound different to having each distortion based processing on a separate plugin.

I seem to strongly recall Cliff posting about 18 months ago in the Fractal Forums that to him, anything at -60db or lower was irrelevant .... no idea if that's right or not about the " -60db " figure ?

Ben

 

[James Freeman]

=> at around 18k there is -63db of Aliasing ?
=> at around 12k there is -71db of Aliasing ?
=> at around 9k there is around -66db of Aliasing ?

What you are going to actually hear is the peaks at 2.7k, 3.7k and 6.5k, especially the giant -30dB peak at 3.7kHz.
Again, with a real amp you should not see anything below and above the pure test signal, just one pure peak.

"9.00am"

Number in dB please?

 

[James Freeman]

I seem to strongly recall Cliff posting about 18 months ago in the Fractal Forums that to him, anything at -60db or lower was irrelevant .... no idea if that's right or not about the " -60db " figure ?

Yeah, I've tested this in the past too, that's about the threshold you can hear two pure sine tones together without suffering the volume of the louder one, but that doesn't explain masking of a more harmonically rich signal like guitar.
We should test this.

 

[MirrorProfiles]

another easy way to hear aliasing is to sweep the sine frequency up, if you can hear the harmonics moving in a downward direction while the original sine is going up, that’s aliasing. you’ll see that visually too in an analyser

 

[BenIfin]

What you are going to actually hear is the peaks at 2.7k, 3.7k and 6.5k, especially the giant -30dB peak at 3.7kHz.
Again, with a real amp you should not see anything below and above the pure test signal, just one pure peak.


Number in dB please?

View attachment 5276

-27.53db .....is there a "better" figure to use ?

 

[James Freeman]

Progress in small steps.

Spoiler: Pedals

Strumming the open strings with EMG 81 and capturing the spectrum of Bypass, Rat, and a Tubescreamer, all normalized to peak.
Pedals are set to everything at maximum to generate highest harmonic distortion above 10kHz, obviously real pedals don't have aliasing so we are seeing the peak level of pure analog harmonic distortion.

The numbers below are for the 10kHz to 20kHz range you can see on the graph above, the input frequency range that generates aliasing if processed by a modeler or a plugin.

EMG 81: -70dB to -84dB.
Tubescreamer: -52dB to -66dB.
Rat: -39dB to -48dB.

Reminding that these numbers are form a normalized signal, ie. if the Rat peaks at -10dB, the values would be -49dB to -58dB, moved down by 10.

Alright, so now we can see that running a huge test signal above 10kHz into the modeler or plugin to test aliasing is meaningless.
We need to use real world 10kHz+ values, whether it is a pickup, a boost pedal, or anything we put before the amp.

More testing coming.

 

[James Freeman]

-27.53db .....is there a "better" figure to use ?

Ben, we are in the middle of figuring out a complex subject and we haven't talked about audibility and masking yet.
How can you ask the "best" questions already?

 

[BenIfin]

Ben, we are in the middle of figuring out a complex subject and we haven't talked about audibility and masking yet.
How can you ask the "best" questions already?

Cool. No probs !

Ben

 

[BenIfin]

Just re-did 3 x Tests as follows:-

-> Sine wave @ 10097k and Level @ 0.00 db .... again - am flying a bit blind

Test 1 = Amp DI Only - no Tonex Cab and no IR

Test 2 = Amp DI with my custom IR with no cuts - running full 20hz <-> 20khz

Test 3 = Amp + Cab Full Capture with no cuts running full 20hz <-> 20khz

Not sure if these add much to the discussion (?) but figured I would at least add the "IR" and "Tonex Cab" as I'm probably never [ :) ] going to play any modeler or profiler without a Cab or IR in "FRFR" mode :)

Still don't really know how to "read" which "spikes" are which .... ie:-

=> given I selected 10097k @ 0.00db ...... which are the "correct spikes" that are supposed to be there and which are the "aliasing spikes" which we dont want / want as low as possible ?

Thanks,
Ben

 

[FractalAudio]

Using a single frequency doesn't yield the best insight. Certain frequencies can yield aliasing products that are far below the average. Depends on the sample frequency and the amount of oversampling. There are frequencies that can yield almost no aliasing and increasing or decreasing the frequency by a tiny bit then yields large amounts of aliasing.

What I do is sweep a tone from 10-20 kHz and set the analyzer to peak hold. This will produce a plot of the "aliasing noise spectrum".

 

[FractalAudio]

Oh, and I should add that every neural network based product I've tested has tremendously bad aliasing performance.

 

[BenIfin]

Hmmmm ... this all got me thinking ... how does my actual Real World Live Preset with an EQ and Drive in front of the Tonex and the Tonex Cab with appropriate Live Hi/Lo Cuts for Rock Guitar "test" .... again done @ 10097hz @ 0.00db:-

Yet again ... not sure how to interpret this ..... but to my untrained eye .... it *seems* (?) "better" / very "good" (?)

Ben

 

[BenIfin]

Using a single frequency doesn't yield the best insight. Certain frequencies can yield aliasing products that are far below the average. Depends on the sample frequency and the amount of oversampling. There are frequencies that can yield almost no aliasing and increasing or decreasing the frequency by a tiny bit then yields large amounts of aliasing.

What I do is sweep a tone from 10-20 kHz and set the analyzer to peak hold. This will produce a plot of the "aliasing noise spectrum".

Oh, and I should add that every neural network based product I've tested has tremendously bad aliasing performance.

Thanks very much for chiming in .... your expertise is much appreciated !

Ben

 

[BenIfin]

....... What I do is sweep a tone from 10-20 kHz and set the analyzer to peak hold. This will produce a plot of the "aliasing noise spectrum".

^^^ You're not kidding ..... Voxengo SPAN has a hold function but it does not have a peak-hold ...... so "for fun" I "manually" very slowly swept the Full Freq Range in the Sine Generator Plugin back and forth multiple times from 20hz <-> 20k ...... the SPAN displayed "results" massively and wildly varied with even just very minor changes ...... I did notice - and I've no idea why - but at ~ 8k there seemed to be little to bugger-all aliasing showing up in SPAN, and what was there, was at "worst" -70db to much lower !?!? ..... this stuff does my head in :)

Ben