Category: Uncategorized

A while back I wrote an article about using Impulse Response (IR) technology to model the sound to a microphone on an acoustic guitar and got pretty good results. What I didn’t mention is that same tech is more commonly used to model amps.

To be honest, I never cared much for that scene, since it seems to be catered more to metal and rock guitar sounds. Since I’m really into 30s and 40s music, the amp tones I’m interested in are more that of Charlie Christian, Barney Kessel, and Oscar Moore. They would play through octal tube amps of the era like the Gibson EH-150 and EH-185 that had a really nice, “wooly” break up to them when digging in. Unfortunately, my lot is pretty niche, and there aren’t any pre-made models of those amps. So I thought I’d finally try modeling one myself.

I don’t have an EH-150/185, but I do happen to have an EH-125 which is a little brother of those amps. So that’s the one I used:

Neural Amp Modeler

The open source project: Neural Amp Modeler has been around for a while now, and seems to have matured greatly over the past couple of years. Best of all, it’s free. You can use it as a standalone app or recording software (DAW) plugin to load models and play guitar through them. There is also a very well made companion site Tone3000 that allows folks to share their own amp models for free download. If it’s a common amp, chances are someone already made a profile for it which saves you a bunch of time trying to do it yourself.

There are good tutorials on how to set up a DAW rig to model an amp, but the gist is:

• Download a 3-minute long source audio file that sweeps through a bunch of sound frequencies
• Play back that audio file through the amp that you hope to model
• Record your amp playing the file with a microphone
• Upload both the source file and your recorded file to a neural processing program to generate a profile for the amp with some fancy math

Results

Here are the results with me playing some aimless Charlie Christianish noodling. The same audio file is running through the amp model, and the original amp with a Shure SM57 microphone on it, so there is no variance in the performance.

Actual Gibson EH-125 mic’ed with an SM57:

My “virtual” Gibson EH-125 model, completely digitally recorded using the NAM plugin:

Pretty close! I think the tone is a really good match. But the original does have more of the crunchy wooly sound in the first half with the chordal stuff that seems to have been lost in translation. Not sure how I could make that more accurate. Maybe if I turned up the amp volume or input gain a bit more during the modeling process, it would capture more of the tone. But in some ways I prefer a little less crunch, too.

I found that I really had to push the input volume to get it to break up in a manner similar to the actual amp. But the interesting thing is: I was able to tweak the input gain and EQ after I recorded it to get closer to the break up sound I wanted. Not the case with the real amp. You’re stuck with the input level and resulting distortion you got during recording.

Here’s me playing a full chorus of Rose Room using the NAM model, with me isolating Charlie Christian’s track and trying to match the tone exactly (I found that the treble had to be turned almost all the way up and the bass to about 75% to get in the ballpark).

Use cases

The most obvious use case of this model to me is that I can use it for home recording without pulling my amps out of the closet and mic’ing them up. I can do some late night recording sessions without waking anyone up.

If I went and purchased some kind of pedal that can load NAM profiles, I could load this model into them and use it for live performance. I could see a situation where I’d just run directly into a board in situations where using an amp is not practical.

Update: I found that the Gigfast Lite iOS app loads custom NAM profiles. I paired it with one of these very cheap headphone jack guitar input adapters and it works great. It has very low latency and is very flexible, with a very nice spring reverb and tape delay effects. Don’t see the need for more hardware, given I already have my iPad at most gigs.

Download my EH-125 NAM profiles

You can download my models at Tone3000 here: https://www.tone3000.com/tones/gibson-eh-125-amp-1940s-32994

I’d love to model some more relatively obscure amps of the 40s, so if you live near me in San Francisco and want to lend me an EH-150 or EH-185 for an afternoon I’d be thrilled to capture them as well.

Note: I am well aware of and have used the Jr. Barnyard and the JJ-150 pedals. They are fine, but I always wondered if I could get closer to the sound using free software. And the convenience of being able to record direct using plugins is appealing.

Please note: This DIY project is not affiliated with or endorsed by Tone Traveler or DrHerringbone

The Algorithm has been serving me a lot of ads for a product! Well that’s nothing new, but in this case, it’s a device that claims to break-in your guitar and improve its “tone” without going through the often grueling and non-rewarding process of actually playing it. That sure sounds good to me, but it retails for an eyebrow-raising $289, and they even offer a product that is literally called snake oil, if you can believe that.

Like all guitarists, I’m not immune to snake oil, especially if my “tone” (whatever that means) is involved, so I got very curious. Here’s a picture of this thing, and the ad that is now burned into my soul:

Of course, let’s not forget a lot of other things are going on during those years of guitar ownership: it’s hopefully getting played a lot, the wood is aging, stretching/contracting, and if it’s in a heated Queens apartment in the winter, probably dehumidifying and cracking up into a pile of firewood. The question is: can you really expedite all that with a few hours with some magic technology?

It’s probably true that playing a guitar for many years changes its sonic characteristics. My hand-wavy scientific explanation would probably include smart-sounding stuff like “the resonant vibrations of the sound is shaking up the cells in the wood, making the instrument more supple and rich,” and so forth.

So what exactly is this thing?

The device in question clips to the bridge of the guitar and emits sounds which will in turn vibrate the guitar. Now I’ve seen a bluetooth speaker before and this thing sure looks like just that. In fact, judging from the logo, it’s probably this one: EWA 106 on Amazon, which can be had for under $20.

There is the matter of hanging it from the bridge of a guitar, but a thick rubberband that I extracted from a head of broccoli and a bent paperclip did the trick. But if you’re a handy, childless person like myself maybe you’d go to the lengths of 3d-modeling a mount that slides right on this particular speaker, like so:

Here’s a link to that 3d model on printables if you want to print it yourself:

https://www.printables.com/model/1345295-guitar-bridge-mount-for-ewa106-bluetooth-speaker

Tone generation

There’s also the tone generation. It also comes with a tablet device and a pre-loaded app, which I assume is paired with the speaker out of the box. The app shows a GUI where you can select individual strings of the guitar to generate those notes.

Well, I’ve also seen a tone generator before, and there are plenty of apps that can do this, but the trick is to find one that can do multiple tones at once. This one: NCH Tone Generator seemed to work fine and was free and available for all platforms.

But in case you don’t need more apps in your life, you can also just find a web-browser based tone generator. Here’s one I found that is nice and configurable, I even pre-loaded the URL with all the frequencies of the 6 strings of a guitar, Online Tone Generator. Just click the on/off button next to each tone to get some creamy sine wave robo-guitar. You can look up the frequencies of specific notes and create another set of tones if you try this on your, say, banjo… but please check your local banjo amplification ordinances before proceeding.

Finally, you can also just download the following mp3 file that I created and play it on loop on the device.

The idea here is you’ll bluetooth pair the speaker to a spare phone, tablet, or laptop. Surely you’ve got a drawer with 3 generations of planned-obsolete devices somewhere. Then open one of these tone generators and let it go for a bunch of hours.

The test

The idea of “tone” is so subjective and non-scientific that I’m not sure that recording me playing a guitar before and after this vibration-bath would yield any tangible result. It’s also impossible to ensure that I’d play the exact same way with the exact same volume and mic placement.

That, and maybe I don’t want to obsess over what 30 second snippet I would play that is both demonstrative and sufficiently impressive in case anyone reading is thinking “can this internet guy actually play?”

But here’s one of those sorts of tests if you want to check it out. I skipped to the section where he does the before/after recordings: Jacob Schuler on YouTube.

The folks behind this product devised an experiment where they mic’ed a guitar and ran the device on it for 4 hours, observing a whopping 10b increase in volume over the course of that time period. That seems like as good a test as any to me.

Here is my test rig:

Over the course of the experiment, I would record a short snippet of the guitar and compare the amplitude and visuals sound wave during each of sample time.

Results

Here are the waveforms from the guitar with the “poor man’s tone improver” from the microphone recorded into GarageBand at various intervals in the test:

Test 1: 0 hours, 3 hours, and 11 hours

Test 2: 0 hours, 3 hours, and 8 hours

As you can see in Test 1 there is a visible 2db volume increase the peaks of the wave.

In test 2, there is no significant volume change. Perhaps half a decibel if I squint just right. The dynamic range seems to have increased though, as the distance between the peaks and valleys is higher on the later samples.

Enough “science” what about the “vibe test”?

Playing the guitar, it does feel a bit richer in overtones. Perhaps a bit louder too. But I really think these things are so subjective and can’t say for certain if its anything I can quantify or if it’s just my imagination. Honestly, I think we hear what we want to hear. If I dropped almost $300 on this thing maybe I’d want to hear lots of things.

While I believe my methods are scientifically precise, there is a lot to be said for recording actual music before and after and not just drone tones. I’ll give that experiment a shot when I find the time, but I’ve been too busy… well… actually playing music on my guitar.

Since doing this experiment over a year ago, I have not been compelled to repeat it on any other instruments. In fact I only thought to finally publish this article because someone asked me about it. I imagine if you’re anything like me, the same would happen to you: you’d mess around with it for a week and then would gather dust. The difference is that if you actually bought one, your wallet would be a lot lighter.

My violinist pal Benito Cortez has been showing up to gigs with an Audio Sprockets ToneDexter pedal for years and touting the benefits of impulse response (IR) processing to make lifeless acoustic instrument pickups sound like microphones. While his results sounded good, I always thought the whole rig looked too complicated and expensive. The ToneDexter retails for $449 new.

But lately I’ve been experimenting with making my own DIY piezo pickups and trying to find ways to make them sound better, which is exactly what these things claim to do. And it so happens that more of these IR pedals have been hitting the market from different manufacturers including Fishman and L.R. Baggs, which are comparable in pricepoint to the ToneDexter. But a very affordable one also entered the scene: The Nux Optima Air, which is only $179 new. I had to try it.

So how do they work? You attach a microphone to the pedal as well as your piezo pickup. Then you record a 10-second sample strumming chords up and down the neck. This records two samples of the same audio source, one from the pickup and one from the microphone. A frequency profile is generated for each sample and some fancy math figures out how to transform any signal coming from the piezo to match the sound of the microphone’s frequency profile. That’s about it — save your profile and you’re ready to go.

The Nux Optima Air has a USB port that acts as an audio interface so it was simple to record the output of it direct into GarageBand to hear the difference. Most demos I’ve seen are of folks strumming open chords, and my interest is more in using Django Reinhardt-style jazz and acoustic archtop playing.

First I recorded a tune* with the IR switch off:

Then I activated the IR effect switch on the pedal:

In short, it actually works. I honestly could have been fooled. The IR effect sounds like a mic’ed guitar to me. There’s none of that piezo “quack”. There’s a fuller warmer sound, with natural sounding harmonics ringing throughout.

There’s a bunch of presets that come with the pedal, but to be honest I think they’re pretty pointless. First of all, they are modeling folk guitars strummers I’m not really interested in (like Gibson Hummingbird / J45, Taylor 314) and secondly there’s no telling what input source pickup they used, where they placed it, and whether it responds even remotely like mine. I think the true benefits to IR are when you model your own exact rig.

I posted a full video of my experimentation with various settings and guitars here:

You may download my IR files here, but your success may vary depending on your pickup and guitar. It’s best to make your own, specific to your pickup:

https://www.paniquejazz.com/wp-content/uploads/2022/06/IRs.zip

*The tune is “Clouds” by Walter Donaldson played on Craig Bumgarner Selmer-Macaferri style guitar. A homemmade piezo pickup is affixed just behind the bridge on the treble side of the guitar an inch below the high E string. Both samples were recorded from this piezo direct from the pedal. Some light reverb was added in the mix.