Chorus Comparison

FractalAudio

FractalAudio

Administrator
Fractal Audio Systems
Moderator
As another one of my technical comparisons I did a comparison between the Axe-Fx chorus and the chorus in the G*****. The G***** is widely regarded as one of the better processors out there and somewhat renowned for its chorusing.

The ultimate test of a chorus algorithm is how little artifacts it produces. The artifacts show up as noise clustered around the input tones. So I took a 500 Hz tone and input that into both units. I set the G***** to a speed of 0.1 Hz, a depth of 65% and 100% wet. The green trace is the result.

I then set the Axe-Fx to the same rate and visually matched the depth. I then set the depth of the Axe-Fx slightly greater so that the traces would be easier to distinguish. The blue trace is the Axe-Fx response. You can see the slightly wider main lobe due to the slightly greater depth setting.

As you can see the Axe-Fx has more than a 20 dB improvement in what I call "clutter". In fact the Axe-Fx clutter is so low it's in the noise floor. The algorithm is very processing intensive but as you can see it pays off. The Axe-Fx shows slightly more harmonic distortion at 1 kHz though. After I did the test I realized I was slightly overdriving the input but I didn't feel like doing it again.

chorus.jpg
 
That's freakin' crazy awesome!!! I agree the Axe-fx is the "cleanest" thing out there when it comes to effects!!!

It's almost as crazy as this dunk!

crazy-slam-dunk.jpg


Keep up the great work Cliff!!!
 
Cliff,


I'm sure you've answered this before, so I apologize in advance.


Seeing the graph you just posted shows me a little bit "behind the curtain" as to how you're able to so closely replicate sounds.

Is there a program available to someone like me (who isnt an engineer) that would allow me to visually try and emulate in the same fashion?


I probably should just train my ears to hear the subtleties, but compared to you, I've got stone-deaf ears...


Thanks!
 
FractalAudio said:
The ultimate test of a chorus algorithm is how little artifacts it produces. The artifacts show up as noise clustered around the input tones. So I took a 500 Hz tone and input that into both units. I set the G***** to a speed of 0.1 Hz, a depth of 65% and 100% wet. The green trace is the result.
Click to expand...

Is there anything about that noise/clutter that might be perceived as somehow enhancing the sound of the output?
 
Is the quality related to how you resample the delay line? Are you doing something much fancier than the normal cubic or cosine interpolation?
 
Technically this is clearly superior. But if you measured a Boss CE1 or CE2 they would surely not get such results, but still be perceived "better" by some. Guess the purely analytical approach not always is the only way, but hearing the quality of what you produce Cliff, I know you know that anyways :cool:
 
hunter said:
Technically this is clearly superior. But if you measured a Boss CE1 or CE2 they would surely not get such results, but still be perceived "better" by some. Guess the purely analytical approach not always is the only way, but hearing the quality of what you produce Cliff, I know you know that anyways :cool:
Click to expand...

exactly... just like i cant make the phase sound like a phase 90... it's too clean in the axe.. the phase 90 has some character.. and that character is it not being "prestine".

Just like everyone harps about true bypass... back in the day (the recordings all these guys are trying to reproduce) nobody had true bypass.. or cared... they had 100' coiled cords, no $100 power cables.. just thrown together... that is part of the charm.
 
Guess we could long for some parameters like "clutter bandwidth" and "clutter depth" to simulate those technical-not-so-optimal behaviour... :roll:
 
I think the boss ce1/ce2 has more to do with analog non-linear behavior than the digital clutter the TC is exhibiting. FWIW, I have never liked the chorus on the g-force or g-major. My favorite digital choruses has been on the Digitech 2120 and on the Eventide eclipse. But I like the Axe-fx better than them both.
 
Friedlieb said:
Guess we could long for some parameters like "clutter bandwidth" and "clutter depth" to simulate those technical-not-so-optimal behaviour... :roll:
Click to expand...
How about a new parameter called... "Shitty," the more you turn the knob the "Shittier" the sound will get with "artifacts" and such... :D
 
schnarf said:
Is the quality related to how you resample the delay line? Are you doing something much fancier than the normal cubic or cosine interpolation?
Click to expand...
Yes. It's possible he does bandlimited interpolation with a sinc of 128 points maybe.
 
Phil B said:
I'm still waiting for a nice TC 1210 sounding chorus. :?
Click to expand...

Any isolated examples? I've never owned one of those. I know it is 2 choruses that interact w/ each other. The quad-chorus can do that.
 
mortega76 said:
Friedlieb said:
Guess we could long for some parameters like "clutter bandwidth" and "clutter depth" to simulate those technical-not-so-optimal behaviour... :roll:
Click to expand...
How about a new parameter called... "Shitty," the more you turn the knob the "Shittier" the sound will get with "artifacts" and such... :D
Click to expand...
So we'd be polishing with a turd? :p
 
I was playing around with this some more and made the following comparison:
chorus2.jpg


I set the input frequency high which is a stress test for interpolation algorithms. I then set the rate and depth high which stresses things even more.

The magenta trace is the G*****. The yellow trace I used a linear interpolation algorithm as is used on L****, B**** and most other mainstream products. The green trace I used a cubic spline interpolation. The blue trace is the Axe-Fx proprietary interpolation algorithm. I offset the traces by 1 dB so that you could see them a bit better so take that into consideration. The Axe-Fx trace, FWIW, is nearly ideal.

The G***** clearly has a poor algorithm (which may explain why I never liked its chorus). I was able to get close to that "performance" using a cosine interpolation algorithm but even that worked a bit better. I believe it might have something to do with using fixed-point math.
 
FractalAudio said:
The G***** clearly has a poor algorithm (which may explain why I never liked its chorus). I was able to get close to that "performance" using a cosine interpolation algorithm but even that worked a bit better. I believe it might have something to do with using fixed-point math.
Click to expand...
You SHARC developers and your fancypants floating point math...
 
You must log in or register to reply here.
Back
Top Bottom