Recently i was talking with a recordist friend of mine, Watson Wu, and we had an interesting conversation about Impulse response recording. I have some notes worth sharing with other sound designers and recordists on this subject, hopefully it will provide some insight into my approach and how i went about capturing impulse responses for the sound library release Battle Rifles & Pistols.
First we should cover the basics, so i can add some reasoning to my method. There are three primary ways of recording an impulse response in the field.
Sine Tone Sweep
This method produces the most accurate impulse responses. I will avoid talking acoustics as I’m quickly out of my depth when it comes to the science behind impulse response technology. Put simply, it has the most complete range of frequencies that produces the initial reverb ‘ping’. Therefore, the recording you get has a much broader range of frequency content as a result. However, there are a few logistical problems with this approach. It involves using a speaker to reproduce the sine tone sweep on location. Thats often not easy or convenient to organise if you spontaneously want to capture the sound of an acoustic space when you’re out recording and only have access at that very moment.
I’ve heard of sound designers using this approach, but living in the UK i’ve been unable to find an actual starter pistol that doesn’t have a plastic mechanism and can produce a sound in excess of 100dB. Starter pistols are getting pretty rare in the modern world it seems.
The actual sound used can be manipulated, but commonly impulse response plug-in units are programmed to work with a clap recording as well as a sine sweep. I use Melda Productions MultiBandConvolution. Using multi band processing as a way of controlling the final result is extremely useful, the plug-in also has great sound design potential for that reason as well but thats another article!
As a sound designer and recordist, I’m grateful for the depth and scope of content that i can find in the world around me every day. However there are times when ambient noise becomes a real inconvenience, recording impulse responses is one of those times. In a studio or in a church, things get a little bit easier. Some locations you may be able to shut down all activity in the area (that would include air cons, people walking around etc and all other factors) but its quite likely in a ‘real’ world location that wont be possible. You may not even have an invite to be there! In which case you may want to quickly deploy a set up, record a few impulse responses in a quiet moment and then clear the location. Preparation and research when recording anything in a public or active area is crucial. Make sure you know peak and quiet times. If you want to record impulse responses then try to find a low point of activity where that might be possible. Next we need to look at the clap itself. We should get something clear straight away that will hopefully save time for those that are new to the subject and may be looking to record a clap impulse for the first time. Claps are pretty quiet, like too quiet! Even if you have huge muscles and bear hands its still going to be a small ‘peep’ when compare to say a firearm. This is a problem in areas where background noise may be an issue. The solution to this issue is raising the level of the initial clap as high above the noise floor as possible. This will produce a louder reverb ‘ping’ which in turn should produce more late reflections before they become inaudible due to the noise floor in an external location.
So, like how do i make a big clap then?
If you don’t have a really good old school starter pistol then i can recommend using a wooden clapper board with a large surface area and handles with strong grip.
I made mine using two lengths of soft wood around 2x6x100cm, a pair of wooden door handles (they were the cheapest) and a flush door hinge. The flush hinge is important as i wanted the clapper board to close evenly and apply as much force as possible to the surface area as it produces the clap sound. This type of hinge allows the boards to close completely without obstructing the wooden surface. Once complete its a reasonable enough size to take on recording shoots, i often just tie it to my stands in a bundle and it can produce a clap sound of around 120dB. Thats comparable to a shotgun so you will need some ear protection, my headphones are pretty well insulated.
I would suggest reviewing other IR libraries to get some idea of the distances that might be useful and the results you can expect. Go out into the field and try some recordings using one of the above approaches. One tip i would point out at this stage, gain compression on IR recordings is bad. You want to keep the gain of the clap as elevated as possible above any noise floor until you get back to the studio. Using a compressor would reduce the peak level of the clap and push your ‘ping’ back towards the noise floor. Give yourself some headroom and watch the gain carefully on your chosen field recorder, disarm any gain compression and limiting.
Back In The Studio
When i first took IR recordings back to my studio and tried them i would say the results were pretty un-usable! Once i used the clapper board approach and got a good signal to noise ratio with no compression this got me some of the way to getting a useable sound but the recordings still needed a lot of work.
As you can see in the featured image from the Melda MultibandConvolution plug-in, the continuous lower frequencies registering in the later part of the spectral display is the noise floor. If this file was used as an impulse response then you would hear a continuous noise in the background of any reverberation which will usually ruin the overall effect. In addition to the noise floor claps are are extremely bias towards high frequencies as they generate very little in the lower end of the frequency spectrum. In order to improve results, mastering IR recordings will help to reduce or in some cases completely resolve these problems.
IR Mastering Workflow
Denosing as a first step in the mastering stage is an option. The results that can be achieved on even fine detail sounds like reverberation tails are much better than it was say 5 or 6 years ago. If you’ve been lucky enough to record an IR with virtually no background noise at all, then well done… you can probably get away without denosing. If you find your results look like the example image above then you’ll have residual frequencies that will need to be removed if you really want to purify the results.
Multiband compression is a good technique to balance the range of frequencies produced from the clap recording. As mentioned, the results will be bias towards high frequencies so you may want to try and correct that by selectively turning down frequency ranges to balance the final output.
Ideally, don’t use any gain compression. The dynamic range of your recording is crucial to retaining a low signal to noise ratio. Obviously you want the peak of the clap to be as close to 0dB as possible so the output from the reverb unit will use the entire dynamic range available. You might find 0dB is a little hot so -2dB is perfectly acceptable in terms of a peak signal.