Intro: Transform an Acoustic Guitar Into a Dobro Using Aluminum Pie Pans and Convolution
This Instructable is in the Art of Sound Contest. If you like it, show me some love!
I'm working on a long distance project with my former guitarist. He found himself in an alternative country groove one day, laid down some tracks, and asked me to do something with them. I was honored, of course, and began work right away.
For one song in particular, he recorded some slide guitar licks on his Ovation Celebrity; his playing fit perfectly, but that guitar sound didn't do it any justice. Acoustic guitars, when played with a slide, sound spitty and thin, with too much attack and little body. I just couldn't get it to sit well in the mix - it desperately needed dobro. EQ'ing was not enough. I had to find a way to model one using his existing track.
Fortunately, modern advances in digital audio and a little ingenuity paid off nicely. In my debut Instructable (debut meaning first one, so be nice) I will attempt to show you how to turn your acoustic guitar sound into one more closely resembling a dobro. Examples are included so you can hear it for yourself.
Step 1: Disclaimers
First, I know there are a few purists who will undoubtedly say, "Nothing you'll do will sound like my '37 National with hand-spun Quarterman cones," or, "How dare you defile the reputation of such a soulful instrument by saying it sounds like pie pans?" Let's get one thing straight: I respect and admire genuine dobros and dobro players alike. The sound I got, as you will probably get, does not sound exactly like a '37 National with hand-spun Quarterman cones. It does, however, sound like an Ovation Celebrity convolved with pie pans which, as I've discovered, is a surprisingly close approximation when mixed correctly.
Second, while we will not be modifying much more than a pre-recorded audio track, you will need to perform surgery on a loudspeaker. You will, in essence, destroy the speaker in the process. This means you will need to use a speaker you don't care about. Don't go ripping up your $1,000 near-field monitors and then come crying to me.
Third, while we will not use anything dangerous like power tools, if you maim or kill yourself in the process of replicating this Instructable I am not liable. I'll tell your grieving family, "I'm sorry for your loss; however, it just means (your name) is a moron and shouldn't have owned a (whatever you used to kill yourself)."
Step 2: How My Fascination With Pie Pans Started
In order to model a dobro, I needed to learn more about how they work. After kicking around on the web, I found that dobros are built in much the same way as a regular acoustic guitar. The only key difference is that they use resonator cones (cones fashioned from aircraft aluminum) to amplify the sound. This difference is what gives a dobro that warm, rich sound that works so well with blues, country, and bluegrass.
The cones are located in the body under the cover (which usually resembles a hubcap), and the bridge is seated directly on top of the cone. The strings vibrate the bridge, which vibrate the cone, which resonates in the body of the guitar.
To physically model a dobro, we need to process the track in a way that resembles this method of amplification. I have both the line out from his guitar and a single mic for air. Since Ovations have a piezo pickup in the bridge directly under the strings, we have a signal that is a clean representation of string vibration alone (I know, it also picks up the body, but I'm approximating here). We also have the mic, which represents what the body of the guitar does to the sound. The only thing we need to mimic is the resonator cone.
Since I don't own a dobro, it would be senseless for me to purchase a cone as they are about $60. I needed to find something cheap that was analogous; this is where the pie pans come in. They're lighter and less dense, and not tuned (I mean, really, do you think Betty Crocker stands at the conveyor belt of the stamper, tuning pie pans?), but they do have an interesting ring to them when struck. They just might work. Worst case scenario, I have some really weird samples to play with in the future.
Step 3: But How Do You Make the Guitar Sound Like a Pie Pan?
Excellent question. Before we get started, I think it's worth explaining a bit of the science behind what we're doing.
If you go into a room and clap your hands, fire a pistol, or pop a balloon, you're exciting the air in the room with a broadband impulse. The impulse is the sudden gush of air in all directions, broadband means that this impulse has a lot of frequency content. The resulting echo is known as an impulse response; it is the response of the room's reverberation to the impulse you created.
An impulse response is a sonic fingerprint of that space. For a long time, ideas and math said that you can apply this impulse response to recorded sound and make it sound like it was recorded in the same room; this process is known as convolution. Until the past decade or so, computers just weren't powerful enough to do the math.
We will be creating an impulse response of aluminum pie pans and using a convolution reverb to apply the impulse response to the track. This is how we will model the dobro.
Step 4: Ingredients for Dobro Pie
You will need:
1. A fast computer (slower than 2 GHz and less than 1 GB RAM is not enough to render in real time. And it doesn't count if it's clogged with viruses and adware that whirls you into a p0rnado when you go online)
2. Decent audio recording software with VST support (like Nuendo or Cubase; Audacity supports VST with a plug-in and it's freeware)
3. Voxengo Deconvolver (unlicensed version is fully functional but doesn't offer batch processing capability)
4. Convolution reverb software (like Voxengo Pristine Space or Waves IR; freeware equivalent is SIR)
5. Decent microphone(s) and a way to connect them to your audio card (crappy mics give crappy results due to noise)
6. Decent audio card (almost all these days can do at least 44.1 kHz at 16-bit resolution, but the higher the better)
7. The best speaker you are willing to sacrifice for science (I used a 3.5" speaker available at Advance Auto Parts for $12.99)
8. A bolt, two nuts, and some plastic washers (the smaller and lighter the better, but the bolt must be long enough to mount the pie pan with plenty of clearance from the sides of the speaker)
9. Aluminum pie pans (keep the rubber boogers on the labels, they'll come in handy later)
10. Noise reduction software (optional, but yields better results)
11. Most importantly, basic knowledge on how to use your computer, software, and hardware. There are too many variables here, such as different digital audio workstations and audio cards and mic mixers, for me to walk you through.
Step 5: WARNING! Science Content!
For the curious, here's a more detailed description of what we're doing. Those who just want to dive in blindly can skip this.
There are several ways you can create a usable impulse response. The quick and dirty method involves using the loudest, shortest sound practical (such as a balloon pop or starter pistol). Under the right conditions these yield decent results. There are a few problems, however.
First, noise is an engineer's enemy, and these methods are crippled by the fact that the maximum signal-to-noise ratio possible is limited to whatever you pick up from the mics. Any noise, such as electrical hum, HVAC, the flap of a butterfly's wings, &c. will spoil the response.
The second, and most obvious thing, is that these methods only work in real spaces. You can't pop a balloon inside a pie pan, and I'm pretty sure firing blanks at it won't work either.
So, we need to energize the pie pan by some means with some signal that we can use to obtain an impulse response. Going back to my explanation of an impulse, it's broadband - meaning the energizing sound carries a large amount of frequency content. There are two main types of signals you can use: a sine sweep (which we will use) and, as an FYI, a maximum-length sequence (preferred, but the software necessary to utilize this type of signal is beyond my budget).
The sine sweep we will create later will slowly pass through all audible (and some inaudible) frequencies, like smearing through a single impulse from lowest to highest frequencies. By comparing the original test tone with that of the pie pan, we can deconvolve an impulse response.
The beautiful thing about using sine sweeps is noise immunity. Since the deconvolution step uses an entirely mathematical method, it averages out noise. The longer the sweep, the more noise it's capable of eliminating. However, since amplitude in audio is a logarithmic beast, it takes an exponentially longer time to lower the noise floor by 6 dB. I find 60 seconds to be sufficient, but if you have a noisy environment you can do several minutes if necessary.
"Okay, so how do we energize the pie pan?" That's where the speaker comes in. We will modify it to work more like a solenoid (since it is anyway, mechanically speaking) and mount the pie pan to the speaker, thereby imparting the test tone through the object.
Enough, let's have some fun.
Step 6: Speaker Surgery
Again, use only a speaker you don't care about. However, the better it is the better your results. Ideally you want a speaker cone with reasonably flat response. It's okay if it's bass light; too much bass will bang the pan around in rather frightening ways.
First, trim or break off any screw mounts around the edge. You don't want the pie pan to take a random bang into any protrusions.
Second, we will need to fix the bolt to the center of the speaker cone. The little circle in the center of the cone is a dust cap. Rarely does it serve any genuine acoustical purpose; it's simply meant to keep grit from getting into the voice coil and wrecking it. So, we will carefully remove the cap.
If you have the same speaker I bought for this, there's a shortcut I didn't know about until I started peeling the dust cap away. If you turn the speaker over, you'll notice the magnet is actually a toroid. The hole gives you access, so you can skip prying up the cap.
Third, poke a hole in the center of the cap. Do the best job you can at centering this hole; an off-center hole will cause the bolt to tip out of plumb and can cause odd vibrations later. Make the hole just big enough that the screw will fit through.
Next, poke a hole in a rubber booger and slide it all the way onto the bolt. Fit this through the back of the dust cap. Check the fit; it the cap won't fit down onto the center of the speaker, you can either discard the cap use the booger (if it seats tightly inside the cone) or use a smaller booger. I don't think there's a real rule here, just do what feels best.
Finally, using super glue or rubber cement, carefully glue the cap back into place. With the speaker I bought, you won't need to do this unless you mistakenly pried it up like I did.
Step 7: Hooking Up the Speaker
There are several methods you can use. You can run the audio out of your sound card into an amplifier and use it to drive your speaker. I ran mine from the headphone output on my audio interface. It really doesn't matter, so long as you know how to hook it up.
Next, play some test audio to make sure it works. Start the audio low, then raise it until you start to hear distortion. Meanwhile, carefully watch the bolt to make sure it doesn't start wobbling around in crazy ways. It should have a straight up-and-down motion. If it starts moving side-to-side, back off until it returns to center. This should be a good volume to start our tones.
Step 8: Preparing the Pie Pans
This isn't too hard. Again, you'll want to center the hole the best you can. Also, make the hole just large enough so you can screw it onto the bolt. This will make for a more secure fit with less errant vibration.
Next, screw a nut onto the bolt just enough so the top of the nut is about 1/8" to 1/4" above the edge of the speaker cone. You want plenty of clearance so the pan doesn't hit the edge. Put a plastic washer over the nut, then screw down the pie pan. Next, apply another washer and finally another nut. You'll want to hold the pan steady when you do this, otherwise you won't get a secure fit. Snug the nut finger-tight.
Step 9: Creating Sweeps
For this step we will use Voxengo's Deconvolver software; I have to say this is definitely worth purchasing should you decide to continue this sort of thing. For now, though, just install the program and fire it up.
Click on the "Test Tone Gen" button, and a dialog will pop up. You should set the bit depth and sample rate to as high as your sound card can handle. The channels setting doesn't really matter here - either mono or stereo will work fine. For now, uncheck "Apply fade-in and fade-out".
The duration is up to you: the longer the better, but don't be silly and create 3600 second test tones. 60 seconds is fine, but if you're working in a noisy environment (like an apartment or in the same room with your overclocked computer with hair dryers as case fans) then 300 seconds might be better. Keep in mind, that's five minutes at a time you have to listen to piercing tones while resisting the urge to cough or scratch that pesky itch in your no-no zone.
Hit "Generate" and choose a location for your file. That's it for the test tone.
Step 10: Fire Up Your DAW Host
Open your host of choice and create a new project. Import the test tone we just created and solo the track. Next, connect your mics to the audio input and create new tracks for them.
As you'll see, I'm using two Nady microphones in a Blumlein configuration (one omnidirectional condenser mic for mid and one figure-eight ribbon mic for side). How you do this is up to you, but there are a few things to consider. You'll want a mic with a fairly flat response curve and low self-noise. If you're mic'ing in stereo, there are many ways to do this. Traditional stereo configurations will need two matched mics equidistant from the pan and spaced about 9"-12" apart. They should aim towards the center of the pan.
If we're all ready on the Dark Side of the Moon, play the sine tones.
Play the sweep and first watch (and listen to) the pan. You will hear all kinds of barking and squawking that, if broadcast over radio, would no doubt be interpreted as a "WOW!" signal by SETI. Honestly, the only way to describe it is a sound that will make you realize you're running test tones through an aluminum pie pan (which will make you either feel like an idiot or a mad scientist).
More importantly, though, you need to ensure that the pan doesn't bounce around or rattle. Rattling is difficult to discern at times, but it's considerably different than the nice ringing you get (usually in the higher frequencies). You'll get the hang of it. If it's bouncing, decrease your volume until the ill behavior subsides. If it rattles, make sure everything is good and tight; if it still doesn't go away, try decreasing the volume.
Next, make sure to trim your mics so they max out at around -3 to -6 dBFS.
Once you're satisfied, record. You'll want to leave some trailing silence on each track since the deconvolver expects the recorded file to be longer than the test tones. I left 10 seconds so I could have enough blank space to use noise reduction software later.
Remember to also record the speaker with no pan attached. This is for two reasons:
1. This will give you an impulse response of the speaker, which may come in handy for sound effects later on.
2. More importantly, we will also use this track as a reference. If we take the output of the speaker and compare the pan's output to it, we eliminate the speaker and mics as a variable in the final impulse. I've found this method to be a bit dicier, but when it does work the results are better.
Next, use your wave editor of choice to first remove DC offset and then normalize the files directly (be careful, most DAW host programs use non-destructive editing, meaning any changes you make in the host will not be applied directly to the file. That's swell for undo reasons, but not so good here. The deconvolver doesn't care about undos, it only sees the raw files).
Step 11: Perform Noise Reduction (optional)
That's one thing I hate! All the noise, noise, noise, noise!
As I mentioned before, this method is fairly noise immune. However, if you have any electrical hum or whirring fans, they will end up in the final file. Any noise in the file will get convolved with the guitar track, which makes it noisy.
Be careful with the noise reduction here. You only want to remove as much noise as you can before it tampers with the good signal. When in doubt, less reduction is better, because too much reduction will corrupt your impulse responses.
Why not wait and do this to the final impulse response? You get far better results if you do this here. First, there's less chance of altering your wanted signal since the noise reduction algorithm only has to concentrate on a small frequency range at any given time. Second, a higher signal-to-noise ratio here will result in an even better one after deconvolution, because whatever noise is left gets averaged and becomes quieter.
Step 12: We're Almost There
Now that we have our sweeps, it's time to deconvolve them. Start Voxengo Deconvolver. First, browse to your test tone file. Second, browse to your recorded sweeps (remember, the free version will only process three at a time). Then, browse to your output folder.
The only two options you should concern yourself with at the bottom are "Normalize" (check it, uncheck everything else) and output depth (set this to as high as your sound card will handle).
Hit "Process". Slower computers can take awhile.
Once finished, close out of this and play your impulse responses. If you did everything right, they should sound like someone struck a pie pan.
If you have any strange ringing (like a pie pan ringing isn't already strange) then your recorded files are corrupt. Either they have too much noise (like the pan rattling because it wasn't snug, or the speaker was overdriven), or too little (because you were overzealous with noise reduction).
In either event, they have frequencies present that weren't originally in the test tone and they were loud enough to confuse the deconvolver. Just record some new ones, and be more careful. I know, I've been burnt like this plenty of times, but believe me - it's worth it, and it's almost always due to user error.
After running through this, you'll also want to deconvolve a set of impulse responses using the speaker only tracks as the test tone file. Remember to output the files to a different folder than the previous set. When finished processing, check these out too and make sure they are okay. Again, you may not have good luck with these - but if you do, they'll have a more accurate response.
I have included two impulse responses. Both are mono 96 kHz, 32-bit (floating-point) impulse responses of the 4" tart pan and 9" pie pan.
Step 13: Okay, Now What?
Yeah, I know - it seems like a lot of effort to get a sound that you could get by hitting a pie pan with a stick. But it's not nearly as scientific, and not nearly as fun.
Now we're ready to use these impulses. If you haven't done so already, install your convolution reverb. Then fire up your DAW host again, and open whatever project you have that contains that nasty slide guitar track.
Next, add the convolution reverb as an insert effect and load the pie pan file into it. You'll want to mix it a little at a time, until you start to get that warmth. If you notice too much ring, play with the volume envelopes on the reverb to fade out any excess ringing (the timbre really comes from the first few milliseconds anyhow). You'll start to notice that, with a little effort, it sounds pretty close to the real thing. You get the sustain, the low end growl, and the attack is far smoother than before.
Now, do like I did and laugh hysterically at 4:30 in the morning like Dr. Frankenstein because you, my friend, have turned a guitar and a pie pan into a dobro.
The MP3's are samples of both the clean (Slide Ovation Celebrity) and processed (Pie Plate Dobro) tracks I have. Neither one have any additional effects. You will be able to tell an immediate difference between the two.
Step 14: Tips and Possible Improvements
I posted this in hopes that there might just be a dobro player with too much free time and a spare cone who would be willing to use this method to create an impulse response and send it to me.
Another idea is to see if using an embroidery hoop on the outside edge of the pie pan will help dampen the long ringing decay of the pan while still leaving the tone intact.
You can use this test tone method to take impulse responses of just about anything. Play it in a room and you have that room's reverb. Run it into the input of an outboard effects processor and you have its sound. The only limitation is that convolution only works with linear systems, meaning that a change in input is reflected linearly at the output. Tube amps and compressors are non-linear, therefore you'll get a phasey tonal coloring and not much else.
You can also use short samples (such as snare drums, choirs, &c.) as impulse responses in the reverb. Use your imagination here - you can get some really trippy effects this way. Just remember that the longer the file is, the more it eats up RAM and CPU - and it does so in a major hurry.
Thanks for reading. If you like it, star me up.