Modeller "Dynamic Range" ... how meaningful is it for Guitar Modellers ?

If you roll the volume all the way off then you'll hear the noise.
That's the shorted-input noise. It's Pollyanna-level optimistic when you're considering real-world use. If you terminate a modeler's input with a reasonably representative resistive impedance (no alliteration intended), its noise floor will be higher, since resistors at temperatures above absolute zero generate thermal noise that increases with temperature and resistance. The guitar's noise floor with its volume control open enough to be useful will pretty reliably overhwelm a modeler's shorted-input noise.
The greater the DR, the less audible noise.
If shorted-input noise is enough to cause any kind of audible issue, a modeler won't remain in my possession.
 
That's the shorted-input noise. It's Pollyanna-level optimistic when you're considering real-world use.
No, it's not. When using high gain the noise referred to the input is amplified. If the DR is, say, 100dB and you apply 60dB of gain then the noise is now only 40dB down. This is audible. Very audible. If you improve the DR by, say, 20dB, now the noise is 60dB down. It's still audible but not nearly as much. If the noise is excessive with the guitar volume rolled off the customers will complain that, well, the noise is excessive with the volume rolled off.

The idea is to get the noise referred to the input as low as possible.

If you turn the guitar volume all the way up, assuming there is no interference, you want this noise to be less than the self-noise of the guitar. Otherwise you'll hear excessive hiss as you're playing. If we assume a 10K pickup impedance and maximum output level of 1V then we want a DR greater than 115dB so that the self-noise of the guitar is the dominant noise source.

If you only play clean, like you do, this isn't as important as the hiss is not amplified as much.

We manufacturers we go to great lengths to get the noise referred to the input as low as possible. This involves various techniques including choosing input architectures with low current noise, paralleling inputs on the A/D, using dual-gain approaches, etc.
 
No, it's not. When using high gain the noise referred to the input is amplified.
Apparently you missed my point, which is that shorted-input dynamic range represents a figure that nobody will ever be able to realize with an actual guitar plugged in and sufficient voltage gain to drive the modeler's amp block into nonlinearity (IOW, using the modeler as designed). IOW, it's "Polyanna-level optimistic."
If the DR is, say, 100dB and you apply 60dB of gain then the noise is now only 40dB down.
That only reinforces my point.
The idea is to get the noise referred to the input as low as possible.
Of course.
If you turn the guitar volume all the way up, assuming there is no interference, you want this noise to be less than the self-noise of the guitar.
FYI, there's almost always interference of some kind in a guitar signal. Most of it you can't do anything about inside the modeler. You can, however, take some well-established measures to keep it from getting into the modeler's input.
If you only play clean, like you do,
This is a mischaracterization of my playing and experience, and this is not the first time I've corrected you on it. You know better.
We manufacturers we go to great lengths to get the noise referred to the input as low as possible.
Wait! You mean "we manufacturers" aren't trying for maximum input noise? I'm shocked, shocked, I tell you! (Apologies to Bogey).
 
Last edited:
At what value does it become a moot point? How much difference can humans hear at Opus 101db vs. Helix 123db ?

So ... taking this above ^^ example ^^ ..... and assuming

(a) these figures measure and represent the exact same thing
(b) all other things in the modeler are equal and
(c) the guitar used in each is identical

Does the higher value -or- the lower value have *any* real world, noticeable / audible impact on the tone, dynamics, "feel" etc..... of an electric clean or dirty guitar signal played through each option (?)

Ben
 
Jesus Eating GIF
 
So ... taking this above ^^ example ^^ ..... and assuming

(a) these figures measure and represent the exact same thing
(b) all other things in the modeler are equal and
(c) the guitar used in each is identical

Does the higher value -or- the lower value have *any* real world, noticeable / audible impact on the tone, dynamics, "feel" etc..... of an electric clean or dirty guitar signal played through each option (?)

Ben
No
 
So ... taking this above ^^ example ^^ ..... and assuming

(a) these figures measure and represent the exact same thing
(b) all other things in the modeler are equal and
(c) the guitar used in each is identical

Does the higher value -or- the lower value have *any* real world, noticeable / audible impact on the tone, dynamics, "feel" etc..... of an electric clean or dirty guitar signal played through each option (?)

Ben
The product with 101dB will have 22dB more "hiss". 101dB is actually quite poor by modern standards.

This is a fundamental part of mixed-signal design but apparently is difficult for people to understand.

Here's an example:
Assume the maximum input level of a guitar processor is 0dBv (1V). Assume that the processor has unity gain when there is no amp model between the input and output and the input is shorted. Now assume that the processor is connected to a sound reinforcement system such that 0dBv in achieves an SPL of 100dB. Assume that the sound reinforcement system adds no noise of its own.

Now, if the processor's DR is 100dB the noise out of the sound reinforcement system will be 0dB (inaudible).

If you then put an amp model into the processor with a gain of 60dB, which is not an unusual amount of gain, the noise out of the sound reinforcement system is now 60dB. That's pretty audible. Equivalent to normal conversation levels. So you're sitting there with your guitar's volume off and your speakers are going "ssssssss". It's further exacerbated by the fact that the noise is "white". Human hearing is pink which causes white noise to be more noticeable in the high frequencies. This is why things like pre-emphases/de-emphasis were invented. To reduce the perceived loudness of white noise.

If the processors DR is, however, 120dB the noise out of the speakers is now only 40dB which is much less audible.

The noise is also audible when playing lightly or rolling off the volume. If the DR of the processor is worse than the self-noise of the guitar then the processor's noise will dominate. Ideally the processor should have a DR that exceeds the DR of the guitar.

101dB is poor. A typical guitar at room temperature is about 115dB. This means the processor's noise will dominate. 123dB is excellent and the processor's noise will be very low.

Now, obviously interference (hum, EMI, etc.) reduces the DR of a guitar but interference is almost always periodic. Humans perceive white noise in the presence of periodic interference. IOW we still hear the hiss. 60Hz hum doesn't mask it. We hear "zzzzzzzz" plus "ssssssss".

There are several ways to improve the DR of a processor:
1. Use multiple channels in parallel on the A/D. Every doubling of the number of channels improves the SNR by 3dB. Some AKM products allow you to do this right in the chip so you don't have to drive multiple inputs thereby reducing cost.
2. Use pre-emphasis/de-emphasis techniques. Boosting the high frequencies in the analog domain and then cutting them in the digital domain. This reduces the perceived hiss.
3. Use dual-gain techniques. Use two channels on the A/D where one channel has more gain than the other. This is similar to what digital cameras do to reduce noise.
4. Use a combination of the above.

The analog signal path is also important. You must use quality op-amps and, ideally, true differential design for best performance. But this adds to the system cost.
 
Last edited:
Back
Top