How to create your own impulse responses w/ Andrew Wade - tutorial 

Thanks for watching!

It's like they read our minds. Thanks URM!

I didn't understand that part too. How can rotating phase makes the audio louder? And if you cut the white noise (btw it is actually pink noise I tried both and pink noise is more accurate) in 1 sample lenght there's no way that the waveform can oscillates both positive and negative ways because 1 sample is too short for that.

I don’t think it’s phase. I think He is referring to DC offset.

MIsweetshadow he’s talking about the positive and negative amplitude of the white noise. Positive amplitude made the speaker push out and negative makes the speaker pull in. If your speaker only moves one way it’s only working at 50% efficiency meaning you get 50% of the total potential volume the speaker can produce. What you have to do is have the white noise have positive and negative amplitude by rotating the phase of that one sample. You could also try generating white or pink noise for two samples. I don’t quite understand how those plugins work in practice but that’s the science behind what he is talking about

You can use it in any loader that takes .wav files, and most loaders take .wav

yes

Zoom in on the track till you can zoom in anymore, then attempt to make a selection and make it as small as it will let you.

It's not white noise per se.. Its more a step function. So what he's doing is creating a 1 sample that goes full scale. So if you were to render this the waveform would be all 0s (silence) then 1 sample with an amplitude of 1 (loudest output), followed by all 0s (silence). This is called the impulse. The name makes sense as its like going from 0 to infinity and back to 0 in the shortest time possible. When you generate a waveform like this it's the equivalent of feeding the amp a sine wave that contains all frequencies... And from that we can say that the response of the amp to this impulse will give us the frequency response of the amp.

@@Colin3Dart sort of. being a single sample, you are not actually capturing full cycles of any of those frequencies. This is NOT sending equal freq across the bandwidth to the speaker at all. It is sending a single positive pulse of transient voltage to the amp. assuming the amp has an extremely fast slew response, the amp then sends a positive (and only positive) voltage to the speaker. The speaker moves forward, and then resonates back into a resting position. when you record this, you are recording the natural resonance of the speaker moving, but it isn't accurate to an equal frequency response. You are kind of recording neutral speaker and box resonance, which gets you pretty close. BUT you are also capturing the response of the microphone and what ever preamp you are using too. This is definitely doing a thing, but not exactly like he explains it. A better way, but more complicated, would be to run white noise and take a long rms measurment to create a profile of what the speaker is doing. But if this works... it works. nothing is wrong

@@kniferideaudio5145 I guess from a purely theoretical standp point the impulse should indeed contain all the frequencies in the spectrum. I don't think this is a incorrect statement. Do an FFT on a sampled signal that contains 1 followed by all 0s and you'll clearly see all frequencies of equal magnitude in the frequency response. I guess where this method falls short is the application. There's no way to find a impulse that rises to infinity in 0 seconds. That's where your method of white noise or a sine sweep probabaly gives more accurate results

IRS should be universal

If you ask me this Total nonsense, all IT is indeed is Just a click, IT has nothing whatsoever to do with White noise or whatever source you use to take ONE sample of. One sample is too litle to represent ANY frequency. A click however does contain a wide range of frequrncies when players through a speaker or room Just by the tesonances IT invokes by exciting it.

I'd like to see some examples of convolution of this single sample impulse response with some dry signal. Actual results and how it sounds compared to the real thing.

I read your comment and ran some tests using Andrew's method at 44.1kHz, and Reaper's built in sine sweep convolve/deconvolve function in ReaVerb. I made three click files. A positive sample, a negative sample, and a positive sample immediately followed by a negative sample. (For what it's worth, 1 sample at a given amplitude sandwiched between two samples at 0 amplitude is just that. It doesn't matter if it's generated by white noise, a sine sweep, a square wave, Pro Tools pencil tool, etc.) I recorded all three of the click files through a mic'd speaker along with the sine sweep. Turns out the + and - click file doesn't work because it's not similar to a dirac delta function. I trashed that one. I made the polarity uniform between the three remaining IR files, normalized them, and compared their waveforms. Each sample's amplitude wasn't dead on, but they were extremely close. Sure, this method of creating an impulse may not be 100% mathematically sound, but practically, my conclusion is that the results are so close that it doesn't matter which method you use. Andrew's seems to be the faster of the two. After all, convolution is just multiplication in the frequency domain, so impulse files with similar peaks and troughs will sound similar. It doesn't matter how those peaks and troughs are created for this specific application.

I shoot 15 Ms out for verbs iirc. It should theoretically still be the correct EQ curve of whatever you sample.

You can do the peak to trough on white noise in the following way. First, record on a track you won't use a sine wave at 20hz. Then on a second track, record a white noise track. then with the two tracks zoomed enough that you can see the zero-peak-zero-trough-zero point of the 20hz wave, do your best to align the white noise track's 20hz cycle's peaks and troughs to the 20hz waves. Trust me, once you zoom out a little, its actually easy to see the 20hz oscilation amidst the noise. Just line that up as best as you can. it actually doesn't have to be perfect. It takes a few minutes to do the whole process. Then, with the white noise aligned to the 20hz, select the space from zero to zero, I.E. from the exact zero point, through the peak, to zero, through the trough, and then to the final zero. make a selection of that entire area, and make a copy of that selection or cut off the ends of that selection on the wHITE NOISE file. Delete the sine wave, you won't need it now. Now you have a 20hz complete frequency cycle length of white noise, which also contains all the other frequencies. Doing this also cuts out some of the lower frequencies by nature because the whole cycle can't complete. So you might want to move the white noise away from the zero on the timeline to maybe one second, make sure there's not "soft fade" on the edges of the wav, and then make sure there's a little empty space at the end, so you can put an empty wav file at the end for a hundred miliseconds or so, again, just to make sure the white noise plays in its entirety into the system you're capturing without being clipped. Then feed this white noise into the power amp and into the cabinet and mics. From there, do as described in the video here where you cut your empty space off the front of the file, and for me, I go all the way to the zero point where the wave starts at the beginning and cut everything else off. and I'd even recommend cutting the bulk of the space off of the tail of the recorded files too, basically from the point that there's nothing happening but ambient noise, cut it there. Then that's your IR file.

Matlab? Link or refer me?

can not

Because a guitar amp power section has a "tone" to it (which is influenced by tubes/transformers/etc), as opposed to a flat-response, linear gain amplifier which doesn't add power amp "color" to the signal that is coming out of the cabinet. The flatter the response of the amp you use, the better quality of IR you'll get.

Im very interested in using different noises to make irs like ppl use pink noise to mix there songs why not use it to make ur impulse response

any daw

the amp may be giving some tone to the final output the cab i guess?

JamieRuinsEverything The single sample is just the input to the system (a guitar cab/mic/room combo). What you record is what contains the information. I.e how the system responds to the impulse (hence the name). The response can be thousands of samples, especially if you also want to capture room ambience and reverb

Kevin Suhr I wish you can help us by make a walkthrough video about the precise technique to capture guitar cabinets IRs

Ya, while he used white noise to make the single sample impulse, the single sample is not white noise. The result of his method made an audio file that was something like [1, 0, 0, 0, 0, 0...], which is just a high audio pulse of 1 sample long. Google Dirac Delta Function, that's basically what he made and what you need. It's not the fact that he made a single sample pulse from white noise that it's capturing every frequency, but because he created an impulse function, which by definition contains all frequencies. But he could have taken a single sample from a square wave that starts high, and have gotten the same 1 sample impulse function file that he's going to use to make the IRs. The single sample he gets in this method will capture all the frequencies in the end, but not because it was made from white noise, but because he has a file with a single sample of 1 and that's all you need to find the response of a system like a speaker response.

@@kevinsuhr6342 How is it possible for one sample to contain two amplitude points? (the most negative and most positive)

cab sim

You are right on the money. These guys, either knowingly or not, are using the concept of white noise too freely. Plain and simple, they are literally approximating an impulse, which is technically supposed to be an infinitely short pulse with very high level. But we can’t make that signal in real life, and this single sample signal is the closest thing to it that is quickly and easily done. But it is conceptually and mathematically true that a very short duration pulse like this has energy across a wide spectrum of frequencies. Not all, like white noise, but close enough for the reasonable result we obtain. So he’s putting in an approximate impulse, and directly recording the response. That response is the real life convolution of the signal chain, speakers, and mic, and the digitally sampled recording of that response is a finite impulse response that will accurately reproduce the desire room/speaker response when convolved digitally with another input wave file, which in our digital world is just another train of “sort of” impulses. So the process here isn’t mathematically precise as far as information and transform theory goes, but it’s close enough to give us useful tools.

How do you fix the phase issue he talked about?

I think it is because white noise contains a roughly equal amount all frequencies. Pink noise on the other hand does not contain an even amount of all frequencies. The level of each frequency goes down as the frequency goes up, in other words more low frequencies relative to high frequencies.

@@fredrikreinholdsen9018 yes but low frequency requires more power to be pushed, personally i use the pink to have the same response from the guitar cab even in the lower part of the spectrum

White noise is equal power per frequency, pink noise is equal power per octave

Exactly what I am thinking... if you like your setup amp to cab, why wouldn't you capture that?

@@Polentaccio i am no expert, but i had to take a good guess, if you use your power amp in your amplifier, especially if its tube, you get added coloration to the tonality, what you want to measure are the speakers with clean power and no color so that you are getting the speakers true tone. Other wise if you did a side by side comparison of the amp through the cab, then impulse, you would get different results.