What is the internal processing resolution?

A word to someone who said that he doesn't see a need for high resolution in processing guitar tones...

After mixing records since around 1990 i can confidently say that Cymbals and Distorted Guitar tones are two of the most revealing signals as far as digital audio resolution.

If a distorted guitar tone does not need the highest resolution possible to sound somewhat close to natural, nothing else does...
 
Sure, @GlennO, didn't mean to invalidate what you said... I guess just working it through out loud.

The net result: the AF3 has a pristine audio signal.

Beats the hell out of the Yamaha FX500 I had in the early 90s, but boy that was a compact and fun unit! If the III could be put in a half-width 1U, OMG!
 
Last edited:
yeky83 said:
Isn't dithering unnecessary with 24-bit integer audio anyway? Quantization noise with 24-bit audio is lower than any analog output noise (thermal noise), let alone the gained up noise level of a guitar amp output.
Click to expand...
Exactly. The thermal noise in the processed audio is greater than the theoretical noise floor of 24 bits. If you were to dither by adding noise the added noise would be less than that already in the data.
 
GlennO said:
Oh my, that's not right at all. The AxeFX of course processes audio at 32 bit. And it outputs to 24 bit. Where you're confused is the conversion from floating point to integer. It's 32 bit *floating* and 24 bit *integer*. There is virtually no truncation with normal audio signals in that conversion.
Click to expand...

No confusion. 32-bit floating point converted to 24-bit fixed is exactly when you should dither (Bob Katz "Mastering Audio" is probably the best-known reference for this). To me, and others, 32-bit floating source files dithered down to 24-bit fixed sound better than files converted with no dither.

That being said, these two statements DO confuse me:

FractalAudio said:
Precisely. The Axe-Fx does not reduce the bit depth so no dithering is required.
Click to expand...

and

FractalAudio said:
The Axe-Fx III uses 32-bit floating point so that would be the best, if available. If not, 24 bits is fine. 48K would be my recommendation as any other frequency will require SRC.
Click to expand...

We know the Axe-Fx's A/D/A is 24-bit. If processing is done at 32-bit floating point, BUT the Axe-Fx does not reduce bit depth, how then do we get back to a 24-bit output without truncation/bit depth reduction?
 
finleysound said:
BUT the Axe-Fx does not reduce bit depth, how then do we get back to a 24-bit output without truncation/bit depth reduction?
Click to expand...

Dithering 32 bit float to 24 bit int is one of those topics that comes up from time to time in DSP discussions, but it's rare to hear anybody claim it makes an audible difference.

And there’s a good reason for that. The mantissa is, for all practical purposes, the same word length as the integer representation, so there is no truncation. Without truncation, there is no point to adding the dithering noise to your audio. Even if there were, you would need to immerse your rig in a helium cooled chamber for it to make any difference to the signal.

It’s hard to tell if you’re sincerely curious about this or whether you’re implying there is something wrong with the AxeFX, but either way I hope this helps explain why there is no need for dithering the AxeFX output.
 
GlennO said:
]
It’s hard to tell if you’re sincerely curious about this or whether you’re implying there is something wrong with the AxeFX, but either way I hope this helps explain why there is no need for dithering the AxeFX output.
Click to expand...

Event though im not the one who made the remark/question, i think people should feel free to ask questions or even assumptions without being scared of insulting Fractal. We are all learning on forums from one another and a discussion can take wings without constantly having to walk on egg shells. Just my .2
 
dumbeat said:
Event though im not the one who made the remark/question, i think people should feel free to ask questions or even assumptions without being scared of insulting Fractal. We are all learning on forums from one another and a discussion can take wings without constantly having to walk on egg shells. Just my .2
Click to expand...
historically, many questions like this thread are asked by people in an effort to diss the product, or create a fight about how some other product is better. it's just a reality of a product forum.

so people will definitely try to feel it out to see the intent of the question.
 
dumbeat said:
Event though im not the one who made the remark/question, i think people should feel free to ask questions or even assumptions without being scared of insulting Fractal. We are all learning on forums from one another and a discussion can take wings without constantly having to walk on egg shells. Just my .2
Click to expand...

You're right. I was just trying to say I wasn't clear on whether this was the kind of answer he was looking for. I hope I hit the mark with my answer. If I didn't, please don't interpret that as meaning there was some hidden message in my reply.

Also, I want to apologize if I come across as a "know it all". Those of you familiar with my plugins know I've been working in audio software for a while, so it's just that this an area where I feel I can contribute to the Fractal community.
 
Last edited:
chris said:
historically, many questions like this thread are asked by people in an effort to diss the product, or create a fight about how some other product is better. it's just a reality of a product forum.

so people will definitely try to feel it out to see the intent of the question.
Click to expand...

I enjoy the tech talk. Forum members here come from all walks of life with a common interest in music and the AxeFX. I myself come from a computer science background and really enjoy the conversation. I learn a lot from people with vastly more experience in other fields and am truly intrigued with how it all comes together in the magic box.

Discussion is good and I for one feel fortunate that the members here share their knowledge so selflessly. The cherry on top is Cliff responds and sets the record straight when someone is inaccurate and even encourages the discussion. Don't let the few trollers that come and go curb the enthusiasm for dialogue and knowledge. It's what separates this forum from most.

*taking my Axe III out of the helium cooled chamber* :D
 
Last edited:
Yeah, this is beyond my basic understanding of the inner workings of digital audio, I personally don’t believe anyone is being anything less than genuinely curious here. But that is from a point of view of someone who has the perspective of being able to kind of understand what’s being said and having to use my imagination off in “technical pretend land” to try & follow along with that which I don’t. Fascinating stuff nonetheless.
 
Here's a 32-bit floating representation taken from Wikipedia (s=sign bit, e=exponent, m=mantissa):

0 01111100 01000000000000000000000 = 0.15625
s eeeeeeee mmmmmmmmmmmmmmmmmmmmmmm
1 |--8---| |---------23----------| -> 32 bits

It's been said several times, but the advantage of 32-bit floating point is that it has effectively 24-bits of resolution (23-bit mantissa + 1 implied bit) with a floating scaling multiplier of 2^(8-bit exponent) leading to a huge range of values of 10^76 with 24-bit precision.

In the Axe (DSPs or plug-ins), changing the level or volume by a factor of 2 (or 0.5) is the same as adding 1 (-1) to the exponent. Thus unlike integer numbers, re-scaling of a floating point number has no effect on the resolution (except for small errors accumulating over many operations, although algorithms are optimized to minimize this).

For calculations that need higher precision Cliff mentioned that "64-bit processing is used in critical locations to minimize coefficient sensitivity."
 
GlennO said:
You're right. I was just trying to say I wasn't clear on whether this was the kind of answer he was looking for. I hope I hit the mark with my answer. If I didn't, please don't interpret that as meaning there was some hidden message in my reply.

Also, I want to apologize if I come across as a "know it all". Those of you familiar with my plugins know I've been working in audio software for a while, so it's just that this an area where I feel I can contribute to the Fractal community.
Click to expand...
What are your plugins?
 
You must log in or register to reply here.
Back
Top Bottom