Convolution is the technology du jour for creating convincing reverbs, and if you're a Logic Studio bundle owner, you already have a utility that will let you get into it yourself, by creating custom impulse responses for the Space Designer plug-in.
Anyone who's been involved in music recording over the past few years must have noticed the increasing number of reverb units and plug-ins that work using a process known as convolution. Previously, digital reverb was created using delays and recirculating filters to approximate what happens to sound in real spaces, but with convolution, you 'record' the characteristics of a real space and then apply them to your own music. The process itself is actually fairly simple, but the reason why it has only recently become commonplace in Digital Audio Workstations (DAWs) is that it takes a lot of computational power to achieve. Since Apple bought Logic originators Emagic, the once-optional Space Designer convolution reverb plug-in has been bundled with Logic Pro as standard — and with the Logic Studio bundle comes a utility program that makes the process of creating your own Space Designer patches a whole lot easier.
Bearing in mind the fact that Space Designer has an expansive library of impulse responses (IRs) from great-sounding churches, concert halls, studios and clubs around the world, you might reasonably ask why you should bother recording your own reverb IRs. In most cases the honest answer is that there is little point, unless you come across a space with a very special character. Of course, sampling small or weird spaces can be useful and fun, but the real point of this article is to show that convolution isn't only applicable to cloning the reverberation of real spaces: it can be used to capture the character of many other 'linear' systems, by which I mean ones that don't include distortion or time-variant processes such as chorus, flanging, phasing, compression or expansion.
When recording a reverb IR, the speaker you use to play back your sine wave needs to be as flat and accurate as possible, otherwise the tonality of the speaker will colour your IR, and if you want to generate true stereo-in, stereo-out reverb rather than the more common mono-in, stereo out, you'll need to run the sweep signal through two speakers. You can record IRs at different locations in the room, and as long as you keep the background noise low and the recording level sensibly high without clipping, the rest is down to artistic judgement. A common approach for concert halls is to put the speaker where the performer would normally be and the mics somewhere in the audience area.
The theory behind convolution goes something like this: if I take a single sample of audio (in effect, a very short click or impulse), then play it into a room via a very accurate speaker system, I can record the resulting reverberant sound (impulse response) at some other point in the room and capture all I need to know about the way that room behaves in terms of reverberation. To recreate the result using a dry audio file elsewhere, all I need to do is multiply every single sample in my own recording by the impulse response recorded in the room, and the exact sound of that room will be added to my own audio. This actually works, but have you spotted the potential problem? A click the length of a single audio sample is so short that the amount of sound produced would be tiny. Furthermore, no speaker system I know of can reproduce a pulse of that short a duration without adding its own coloration.
Experiments have been done using bursting balloons and starting pistols to generate a short impulse, with a certain degree of success, but the most popular method used today involves one further step. Instead of relying on a short impulse, the speaker is fed with a sine-wave sweep that moves across the whole audio spectrum over a period of several seconds. Once recorded, a mathematical process is used to translate or deconvolve this into an equivalent impulse response. A sweep signal of this type produces a better signal-to-noise ratio than a conventional impulse and the loudspeaker system is no longer being asked to reproduce an impossibly short click with no coloration.
As already mentioned, Logic Studio comes with a separate Impulse Response Utility that generates a sweep signal, allowing you to create your own IRs. As you'd expect, the same software provides a way of converting the IR into a Space Designer preset.
The IR Utility also includes an Audition facility which allows a small range of test signals to be treated with the IR, to give you an idea of how it sounds. However, my own preference is to use the 'Create Space Designer Setting' option at the bottom left of the window, which makes your newly created IR instantly available within Space Designer. I tend to keep Logic open at the same time as I'm making IRs, with an instance of Space Designer active, so that I can quickly load in my latest IR and try it out. A point worth noting is that if Space Designer is in a channel insert point, which is where it would normally be used for amp/cab modelling, it defaults to 100 percent dry signal. You'll need to reset the mix to zero percent dry, with a wet signal level of around -6dB, and then resave the setting using 'Save As', in Space Designer's preset menu. I just keep the same name and replace the old settings.
An obvious and simple place to start creating your own IRs is with an outboard reverb unit. But first you need to set up a project in the Impulse Response Utility. Open the Utility and select New Project from the File menu. A dialogue box opens, asking you what format you wish to use (the options include mono, stereo and various surround formats); for making a stereo reverb IR, you'd select stereo. Next, assign the audio interface output channels that will be used to send out the sweep signal. Ensure that these outputs are connected to the inputs of the audio interface, and that you have the effects unit's outputs routed back into a pair of interface inputs.
Find a patch you like on the outboard reverb, and make sure it's sending out a 100 percent wet mix of the effect. (Note that if the reverb includes time or pitch modulation, as many of them do, those elements will be lost, so the end result may not sound exactly like the original — though it may well still be perfectly usable.) Next, click on the Sweep button and watch the waveform display as the sweep progresses, to ensure that you have a healthy level (as a guide, aim higher than 50 percent but never allow it to clip). By default, the sweep takes 10 seconds, and once it's complete, you'll see a dialogue box asking if you'd like to save the project. If you do opt to save, you can revisit the project later to make further edits, which I'd advise on the first few tries, but I don't find this necessary any more, so I just hit Cancel and then continue.
Click on Deconvolve, and after a few seconds, you'll see the sweep waveform replaced by a decaying IR. Because of latency through the system, there will be some dead space before the IR starts. The IR Utility gives you the ability to trim this off, so use the cursor to select the silence, then the Cut button above the waveform display to delete the silence at the beginning of the new IR. If you don't do this, you'll end up with an uninvited pre-delay on your reverb patch. I also find it useful to fade the ends of long impulse responses to make sure they end cleanly and smoothly, by clicking on Fade to apply a fade-out. Note that once you've started to edit an IR, you can't make a new sweep recording unless you first select New Project.
It's not only reverb units that are fair game for 'sampling'. I've heard very good impulses made from EQs, and you can make IRs of your favourite hardware EQ presets quite economically, as you don't need to capture very long impulses. You can take a similar approach as described for the hardware reverb cloning example, above, but you'd normally make a mono-in, mono-out sweep recording (though there's no reason you couldn't make a stereo-in, stereo-out IR for a two-channel mastering EQ or similar). As EQ can increase gain, check the resulting sweep recording for signs of clipping. Trim the gain using the Sweep Level parameter box (in the Sweep Generator section of the panel to the left of the waveform display) and redo if necessary. Remember that you won't be able to adjust the EQ retrospectively.
A technique called dynamic convolution can go beyond reverbs and delays to 'sample' time-variant devices such as compressors, but requires much more sophistication. As far as I'm aware, the only software implementation available is Acustica Audio's Nebula, reviewed elsewhere in this issue.
Creating IRs of an EQ is pretty simple stuff, and it doesn't take long to realise that there's a lot more creative potential available. Speaker emulators and those devices that improve the sound of DI'd acoustic guitars are essentially filters, and so can be captured quite efficiently using convolution — providing there's no compression or distortion going on, and that any time-modulation effects are bypassed. Reverbs and delays built into other processors can be captured accurately, but you will often get more flexible results by switching such effects off and adding them later using other plug-ins if needs be.
You can also create your own speaker-emulator plug-in by feeding the IR Utility's sweep tone into the power-amp input (effects return) of your favourite guitar combo and recording the result, either in mono or stereo. Where the mic is close to the speaker cone (the choice of mic and its positioning also affects the sound, of course), there may be little point in recording in stereo, but if you want to try capturing some of the room ambience with a more distant stereo mic position, then it may be worthwhile.
Essentially, the process is exactly the same as for capturing a room IR, except that this time we're using the guitar cabinet in place of the flat-response monitor, and it's the coloration of the speaker we're out to capture, not the room. In my example (see screen on the previous page) you can see that I'm using the Impluse Response Utility configured with a mono output from 'Analog 5', to route the sweep tone to the amp. To connect the two, simply plug a standard jack cable from the interface output to the effects return of the amp.
Ensure that the IR Utility's Monitor channel is set to look for the input from the microphone. Before you can record the impulse, you must activate the Record Ready button in the information line below the waveform screen (which should be empty), and I'd suggest using the 1kHz test tone box (in the panel to the left of the waveform display) to check the signal feed to your amplifier. Adjust the source level, if necessary, using the Sweep Level parameter box. All you need to do now is press record and the IR Utility will do its business. While you have all the gear set up, you might as well capture a number of IRs using different mic positions. Once you have all your IRs captured, 'top and tail' them using the cursor and Cut tool, as previously described. If there seems to be noise or other residual signal in the tail, you can select the end portion and click on Fade to apply a fade-out.
Because convolution can't capture dynamic (level-dependent) processes, this approach can't record the flavour of power-amp distortion or speaker-cone breakup (you should run the amp clean in any event, by keeping an eye on the signal level fed into it), but it will duplicate the filtering effect of the speaker/cab/mic combination, enabling you to process the sound DI'd from the preamp out of the combo, to get more or less the same end result as you would from miking it conventionally. In my experience, using convolution captures the low-end cabinet thump due to speaker 'overhang' rather more convincingly than conventional speaker emulation. Making up a series of IRs using different mics and different mic positions will ensure you have plenty of variety, and don't forget to try different mic positions across the speaker cones, as well as different mic distances. In my tests so far, the speaker emulations I've created have felt more authentic than the ones that come with Logic's Guitar Amp Pro plug-in, so I often bypass the plug-in's speaker emulation by selecting DI as the speaker type, then run it into Space Designer with one of my own speaker IRs loaded.
If your guitar amp has a particularly good-sounding spring reverb built in, and you can access it directly, you could even create a separate Impulse Response from that. Then you can use a 'dry' cabinet IR on Space Designers inserted on each guitar channel you want to process, with the reverb IR on an aux channel, so you can vary the amount of reverb just like on a real guitar amp.
As I pointed out earlier, sampling rooms to create impulse responses may not be that useful when so many really good ones already come with Space Designer, but for cloning the occasional hardware device setting or guitar-cab response, the results can be so much greater than the effort involved in achieving them. And the files are relatively small, meaning that you can email them to other Logic users if you wish to!
The longer the event you 'sample', the more CPU power is needed to recreate it, so sticking to a five- or six-second maximum IR makes sense. To ease CPU load further, Space Designer lets you choose a half-bandwidth mode, and in many cases this doesn't affect the audible result too significantly.