From 3257d0200520e6dc6d65540fb1b2c736f77e6377 Mon Sep 17 00:00:00 2001 From: "Robin E. R. Davies" Date: Fri, 12 Sep 2025 09:59:16 -0400 Subject: [PATCH] Docs copyedit --- docs/UsingNAM.md | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/docs/UsingNAM.md b/docs/UsingNAM.md index b2a7f30..1ac06ea 100644 --- a/docs/UsingNAM.md +++ b/docs/UsingNAM.md @@ -8,7 +8,7 @@ The results are astonishing. Models are capable of producing sounds that are vir PiPedal allows you to use Neural Amp Modeler models (.nam files) with the TooB Neural Amp Modeler plugin, which is bundled with PiPedal. -Although TooB ML models produce exceptionally good results, TooB Neural Amp Modeler produces significantly better results. Awkwardly, none of the built-in factory presets for Pipedal currently contain TooB Neural Amp Modeler plugins, because I am still working on obtaining NAM models that are licensed under terms that are compatible with Pipedal's MIT license. (Working on that, expect some progress soon). +Although TooB ML models produce exceptionally good results, TooB Neural Amp Modeler produces significantly better results. Awkwardly, none of the built-in factory presets for Pipedal currently contain TooB Neural Amp Modeler plugins, because I am still working on obtaining NAM models that are licensed under terms that are compatible with Pipedal's MIT license. (Expect some progress soon). So I would urge you, in the meantime, to experiment with TooB Neural Amp Modeler using free .nam files downloaded from [Tone 3000](www.tone3000.com), or perhaps purchase some commercial NAM models from any of several providers of non-free NAM models on the Internet. (Totally worth it!) @@ -22,7 +22,7 @@ Calibration allows you replicate the sound that your actual physical guitar woul physical guitar amplifier with the exact same control settings on the amp that were used when the NAM model was created. While this is certainly an interesting exercise, it is not necessarily a useful one. To properly calibrate your setup, you -will need to perform a fairly elaborate measurement procedure on the output of your physical guitar, using a fixed set of volume, tone control, and pickup settings. And once you have made that measurement, you will need to make sure that you always play with those exact same settings on your guitar. Any change to the guitar settings will invalidate the calibration. And of course, you will need to make sure that your guitar is always strung with the same gauge strings, and that you always play with the same pick, and that you always play with the same attack strength, and so on. So while it is an interesting exercise, it is not a particularly practical one. +will need to perform a fairly elaborate measurement procedure on the output of your physical guitar, using a fixed set of volume, tone control, and pickup settings. And once you have made that measurement, you will need to make sure that you always play with those exact same settings on your guitar. Any change to the guitar settings will invalidate the calibration. And of course, you will need to make sure that your guitar is always strung with the same gauge strings, and that you always play with the same pick, and that you always play with the same attack strength, and so on. So while it is an interesting exercise, it is not a particularly practical one in general. Nonetheless, there is a fairly active community of users who are interested in this sort of thing, and who want to get the most accurate possible sound from their NAM models. They know who they are, and they know what to do. But for the rest of us, it may just be simpler and more effective, and certainly far more creatively useful to twiddle the Input Gain and Output Gain controls in TooB Neural Amp Modeler until we get a sound that we like. @@ -30,7 +30,7 @@ There is no requirement to perform this calibration step. You can simply use the There, however, is some value to understanding the calibration process, as it may help you understand how the TooB Neural Amp Modeler works, and it may help you more efficiently dial in a sound that is close to what you want. -If you frankly don't much care about any of this (and why would you?), skip to the section "What if I Don't Want to Calibrate my Setup?", for useful guidance on trimming input and output levels for uncalibrated setups. +If you frankly don't much care about any of this, skip to the section "What if I Don't Want to Calibrate my Setup?", for useful guidance on trimming input and output levels for uncalibrated setups. ### What Calibration Actually Does @@ -79,13 +79,15 @@ Note also, that calibration generally only works for the first NAM plugin in an Well, then you won't get calibrated output. But you will still get perfectly good sound. You can simply adjust the "Input Gain" control in TooB Neural Amp Modeler to suit your own tastes and playing style. The sound of the amp model is determined by the selected model and its parameters, not by this calibration process. -There is some value to turning calibration on, even if you don't actually perform the calibration process. You can set a fictional dBU level in the TooB Neural Amp Modeler. Typical dBU measurements for guitar signals are usually in the range of -20 dBU to -2 dBU. Humbucker pickups tend to be in the -10 dBU to -2 dBU range, and are nominally somewhere around -6 dBU. Single coil pickets tend to be somewhere in the -6 to -20 dBU range, and are nominally about -11 dBU. So if you were to set the calibration value to -6 dBU, and trim your digital signal to 0 dBFS, then you would feed the model with data that is close to what the model would expect were you to be playing with a fictional guitar that produces -6 dBU voltage levels. So not calibrated, but almost certainly in the right range for what the NAM model expects. For models that explicitly identify themselves as models that have been calibrated for single-coil pickups, you may get better results if you set the fictional Calibration value to -11 dBU. +There is some value to turning calibration on, even if you don't actually perform the calibration process: specifically, turning Cailbration on usually makes it easier to set initial input gain settings when loading a new model. When working with older uncalibrated models, it sometimes take a bit of effort to figure out what input levels need to be fed to the model. For uncalibrated models, there is no real convention for how loud the input signal needs to be to get best results from the model. You may have to experiment with input gain settings until you find a good vaue. However, when working with newer calibrated models, default TooB Neural Amp Modeler control settings will usually be approximately correct. Not exactly correct, neccesarily. You may still need to tweak the input gain knob to get a sound that you want. + +The key to getting approximately correct outputs from a calibrated model is to set a __fictional__ dBU level of -6 dBU in TooB Neural Amp Modeler's calbration Value control. Typical dBU measurements for guitar signals are usually in the range of -20 dBU to -2 dBU. Humbucker pickups tend to be in the -10 dBU to -2 dBU range, and are nominally somewhere around -6 dBU. Single coil pickets tend to be somewhere in the -6 to -20 dBU range, and are nominally about -11 dBU. So if you were to set the calibration value to -6 dBU, and trim your digital signal to 0 dBFS, then you would feed the model with data that is close to what the model would expect were you to be playing with a fictional guitar that produces -6 dBU voltage levels. So not calibrated, but almost certainly in the right range for what the NAM model expects. For models that explicitly identify themselves as models that have been calibrated for single-coil pickups, you may get better results if you set the fictional Calibration value to -11 dBU. ### Recommended Settings for Uncalibrated Guitar Signals There are two types of .nam models that you may encounter in the wild: Calibrated models, and Uncalibrated models. For the calibration process to work properly, the .nam model must provide calibration metadata to TooB Neural Amp Modeler, that indicates the expected input level for the guitar signal. You can check if a .nam model provides calibration data loading the model and checking to see with the "Input Calibration" control is enabled in TooB Neural Amp Modeler. If the control is disabled, then the model is Uncalibrated. -At the moment of writing, Uncalibrated models are more common that Calibrated models. If you purchase a commercial NAM model, the NAM models ARE likely to be calibrated models. But models from Tone 3000 are almost (with rare exceptions) not calibrated. +At the moment of writing, Uncalibrated models are more common that Calibrated models. If you purchase a commercial NAM model, the NAM models ARE likely to be calibrated models. But models from Tone 3000 are mostly (with rare exceptions) not calibrated. When using calibrated models, We recommend that you trim input levels to TooB Neural Amp Modeler to 0 dBFS (peak). If the model is calibrated, then a good starting point is to turn calibration on, and set the calibration value to a fictional value of -6 dBU. This ensures that you are going to feed approximately reasonable values to the NAM model. You can then adjust the "Input Gain" control to suit your own tastes and playing style. @@ -95,7 +97,7 @@ When using uncalibrated models, things are, unfortunately, more complicated. The The problem of "overdriven" models is a difficult one. If you feed audio signals to a model that are outside the range of data that they were trained on, NAM models will often produce unpleasant artifacts. The sound is not as distinctive as that of hard digital clipping; but is unpleasant nonetheless. Overdriven models tend to produce output that is "fizzy", and/or "muddy", and just do not sound like the amp which is being modelled. If you are not getting good output from an uncalibrated model, try reducing the "Input Gain" control (and increasing the "Output Gain" control by a corresponding amount as well), until the unpleasant artifacts go away. -It's an awkward situation, because overdriven models tend to sound like poor quality emulations of amplifiers that previous-generation amp simulators used to provide. And—unlike hard-clipping—the onset of unpleasant artifacts is gradual and not always easy to detect. And doubly awkward because real amps that are heavily overdriven actually produce a similar effect. You might want to experiment with setting the "Input Gain" control to wildly excessive values so that you can hear the effects of an overdriven the model, so that you know what to listen for. You should be able to hear what sounds like a "noise floor" that rises up as you progressively overdrive the model. Once you know what to listen for, it becomes easier to avoid. +It's an awkward situation, because overdriven models tend to sound like poor quality emulations of amplifiers that previous-generation amp simulators used to provide. And—unlike hard-clipping—the onset of unpleasant artifacts is gradual and not always easy to detect. And doubly awkward because real amps that are heavily overdriven actually produce a similar effect. You might want to experiment with setting the "Input Gain" control to wildly excessive values so that you can hear the effects of an overdriven the model and know what to listen for. You should be able to hear what sounds like a "noise floor" that rises up as you progressively overdrive the model. Once you know what to listen for, it becomes easier to avoid. ### Output Calibration