Digital guitar processing: everything you need to know about Impulse Responses

An impulse response (IR) file is a great way for saving the sound of a guitar speaker cabinet into a small digital file. A guitar speaker cabinet is known for its strong midrange and peak frequencies which are caused by the speaker and the reflections inside the cabinet.

You might think of the impulse response as a static sample, but it’s not. An IR filters the signal by convoluting it and is very responsive to input signals. Here’s my “for dummies” way (myself included) to describe what an IR is:

  • the IR contains the spectrum of all frequencies (20 – 20.000 Hz)
  • the IR contains the loudness levels of all these frequencies
  • the IR contains the length (sustain/ringing) of all these frequencies

Every frequency of the input signal will be convoluted in a way which vaguely reminds of a lookup table. If you convolute a perfect sine wave of 440 Hz, the output signal will contain nothing but that 440 Hz frequency. An IR cannot distort a signal because of the direct relationship between the frequencies of the input and output. So no harmonic distortion and no aliasing will ever occur. Convolution is a 100% lineair process so it can’t compress the input. This results in a very dynamic and lively sound.

The empty wav trick

What do you think will happen when you load an empty wav file (a wav with nothing but silence) into your IR-loader? Every frequency of the input is convoluted by the silenced file. Since no frequency can be found the output will be of total silence.

Try it yourself.

And you can also try this: load a Dirac file into your IR-loader. A Dirac file contains all frequencies (20 – 20.000 Hz) at the same level/loudness. This will result in input and output being 100% the same. So no frequencies will be affected.

What sample rate is perfect?

Professional speaker cabinet impulse responses often come in a package containing the IRs in all sorts of sample frequency formats like 44.1, 48, 88.2 and 96 kHz. So you might ask: what is the best format?

44.1 is perfectly fine to capture all frequencies which are needed for reproducing the frequencies of a speaker cabinet. Anything below 6 kHz is all we need. Even 22 kHz will be perfectly fine for a perfect speaker cabinet IR. Keep in mind that those IRs will be resampled in the IR-loader to the sampling frequency of your DAW settings.

So there’s no need to use a 96 kHz IR file just because your DAW is operating at 96 kHz. Only when your IR-loader is not able to resample you should use the same sample rates. But I cannot imagine an IR-loader not resampling the IRs. All plugins I’ve tested are capable of this.

My favourite IR-loader: NadIR

The NadIR loader by Ignite Amps is the one I highly recommend. The default setting of the NadIR has a low pass filter which is set at 6 kHz which makes perfect sense for guitar speaker simulation as I mentioned above.

Null testing nerds: look out!

For those of you who want to compare the different formats like 44.1 and 96 kHz against each other: no, they won’t nullify when inverting the phase of one channel. Must say though that I couldn’t hear any differences between 22, 44.1, 48, 88.2 and 96 kHz files, no matter the settlings of my DAW. Trust your ears, as always!

But that’s not everything you want to know about IRs. I know. There’s more to come about this soon…

Comments

2 responses to “Digital guitar processing: everything you need to know about Impulse Responses”

  1. Jannis Avatar
    Jannis

    I’m one of those null testing nerds you mentioned an stumbled upon this article while testing different sample rate IRs against each other. Could you elaborate on why they don’t null? I assume it has something to do with the internal resampling of the IR loader, but if it was explained in the article, I missed it 😀

    1. Marco Raaphorst Avatar

      I am not sure about this. Files of different sample rates have a gain difference. But it could be anything really.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.