Sweep time length for capturing Cab IRs / Wav File time length for CabLab ?

unFILTERed

Power User
I am trying to capture some IRs of my cabinets. I am using a DAW and a different method than doing it with Axe FX for multiple reason which doesn't matter here.

Can someone tell me what should be the sine sweep length? I will use CabLab to convert the wav files to IRs. What's the ideal file length for CabLab?

And is it better to use a logarithmic or linear sweep?
 
Last edited:
So far what i found says me 170ms is the right lengt. Somehow i am my IRs regardless of what mic and how i position them have allways more lowend than mids or high mids. Any ideas what could be the reason for this?
 
I am using a matrix gt1000fx, my room is acoustically treated and i am not in a corner. I am using a logarithmic sine sweep 20Hz to 20kHz.

Sending the sine sweep to Matrix and capturing with multiple mics back in Cubase on seperate channels. Mic Pre is UAD Apollo 8.
 
Here is how my sm57 IR looks in CabLab. I tried Friedman Dirty Shirley Cab with a g12h and also another cab with Alnico gold. Result is pretty similar. More low End even if I place the sm57 almost to center.

Screen Shot 2018-05-02 at 2.47.05 PM.png
 
Yeah, but...how does it sound? ;)

Too much low end, and before all...not accurate. I am doing somewhere a mistake. If i don't use DAW, and measure with REW (room eq wizard) and export as 0.17s IRs and than import the IR file to CabLab it works fine.
 
I am using a DAW and a different method than doing it with Axe FX for multiple reason which doesn't matter hear.
The reason might not matter, but the method certainly does. Before going further, maybe we should know something about the "different method."
 
The reason might not matter, but the method certainly does. Before going further, maybe we should know something about the "different method."

I found my mistake, I shouldn't be laoding the exported-sweep wav files in CabLab. Those aren't the appropriate wav files which CabLab needs. With the right reference sweep file and the recorded cab+mic it works in REW. Seems like more steps than usual way but still better than sine sweeping for each mic. This way I can sine sweep one time and capture 5 mics at.
 
Last edited:
  • Like
Reactions: Rex
I found my mistake, I shouldn't be laoding the exported-sweep wav files in CabLab. Those aren't the appropriate wav files which CabLab needs. With the right reference sweep file and the recorded cab+mic it works in REW. Seems like more steps than usua way but still better than sine sweeping for each mic. This way I can sine sweep ones and capture 5 mics at once.
Good! I'm glad it worked out.

IRs in WAV format are actual impulse responses, not recordings of sweeps. If you play a WAV IR through a media player, it sounds like a recording that you would get if you tapped a pencil on the speaker cone—it's the speaker's response to an impulse (a tap).
 
Good! I'm glad it worked out.

IRs in WAV format are actual impulse responses, not recordings of sweeps. If you play a WAV IR through a media player, it sounds like a recording that you would get if you tapped a pencil on the speaker cone—it's the speaker's response to an impulse (a tap).

Yes, it took all my day to finally find that out :)

Now looking forward capture IRs of my cabs with some really nice mics (Sony C37A, Sennheiser MD441 - not 421 etc) and also i will capture IRs of some UAD stuff (pultec, studer, manley, helios, neve etc) and will share the IR&wav files and some mixes also presets with amps I like.
 
Use a logarithmic sweep. I use around a 20 second sweep.

I generate them to start and end in sine phase at every octave which is useful to extract harmonic distortion products not just shove them into negative time.
A useful thing about ending at sine phase with a sweep that extends to the nyquist limit is you avoid having to window the end, which in turns avoids pre-echo and ringing after deconvolution.

I start the sweep at 10 Hz, so I can use the first octave (10-20 Hz) to window the sweep to avoid low frequency ringing during deconvolution.

Note the length of the log sweep is unrelated to final IR length.
 
...also i will capture IRs of some UAD stuff (pultec, studer, manley, helios, neve etc)...
A note about capturing IRS using analog preamps or preamp simulations:

Much of the magic of those preamps comes from their nonlinearity — compression and mild distortion that fattens the sound. Neither compression nor distortion can be captured with an IR. When shooting IRs, accuracy is everything. It's good to keep your signal path as clean as possible, and leave the application of compression and distortion to the end user of the IR.

Hint: Explore the Preamp section of the Cab block on the Axe-Fx III. :)
 
Use a logarithmic sweep. I use around a 20 second sweep.

I generate them to start and end in sine phase at every octave which is useful to extract harmonic distortion products not just shove them into negative time.
A useful thing about ending at sine phase with a sweep that extends to the nyquist limit is you avoid having to window the end, which in turns avoids pre-echo and ringing after deconvolution.

I start the sweep at 10 Hz, so I can use the first octave (10-20 Hz) to window the sweep to avoid low frequency ringing during deconvolution.

Note the length of the log sweep is unrelated to final IR length.

Alberta, thx for all this information. Definitely learning new stuff as I am trying to follow and understand the things you wrote.

Can you enlighten me about the part, how to generate the sine wave you mention? Do you generate for each octave a seperate wav file and put them than together, if so with or without breaks in between?
 
Last edited:
A note about capturing IRS using analog preamps or preamp simulations:

Much of the magic of those preamps comes from their nonlinearity — compression and mild distortion that fattens the sound. Neither compression nor distortion can be captured with an IR. When shooting IRs, accuracy is everything. It's good to keep your signal path as clean as possible, and leave the application of compression and distortion to the end user of the IR.

Hint: Explore the Preamp section of the Cab block on the Axe-Fx III. :)

I am aware of this but still after you mentioned it made me rethink and i might generate calibration files for UA Apollo preamps and get things as clean as possible and skip the part about all those vintage gear flavor.
Probably it's not gonna make much sense.

Though one question; do you think it makes sense to use hi and low cut of mic pre (around 30hz and above 9kHz or 10kHz. When i don't do this i feel like unnecessary information is getting in the IRs. Spikes you can visualy see in REW or CabLab. At frequency ranges where you would anyway cut with Eq since those are not needed for Electric Gtr.
 
Though one question; do you think it makes sense to use hi and low cut of mic pre (around 30hz and above 9kHz or 10kHz. When i don't do this i feel like unnecessary information is getting in the IRs. Spikes you can visualy see in REW or CabLab. At frequency ranges where you would anyway cut with Eq since those are not needed for Electric Gtr.
Do what sounds best. There really are no other criteria. But in truth, there's no useful cab response below 30 Hz.
 
Can you enlighten me about the part, how to generate the sine wave you mention? Do you generate for each octave a seperate wav file and put them than together, if so with or without breaks in between?

Not quite.

The math is this:

http://ant-novak.com/swept-sine.php

It's essentially about choosing the right length and starting/ending frequencies so that every octave happens to start and end at 0 instead of some other level (like in the graphic below):

fig1.png


This is useful when you want to extract the non-linear products, depending on what you are doing (they'll be negligible for guitar cabinet captures.

After deconvolution you get something like this using such a sweep:

fig2.png

Where h3, h3, h4, h5, etc, are the non-linear products and h is of course the linear impulse response.


Here's a sweep I generated (and its inverse) that I created using my own code, which have the properties I mentioned above.

You use sweep.wav when capturing the response of the system in question.

The then convolve inverse.wav and the resulting capture, which will give you something like the picture above (not quite that dramatic but conceptually the same).
 
Last edited:
Not quite.

The math is this:

http://ant-novak.com/swept-sine.php

It's essentially about choosing the right length and starting/ending frequencies so that every octave happens to start and end at 0 instead of some other level (like in the graphic below):

fig1.png


This is useful when you want to extract the non-linear products, depending on what you are doing (they'll be negligible for guitar cabinet captures.

After deconvolution you get something like this using such a sweep:

fig2.png

Where h3, h3, h4, h5, etc, are the non-linear products and h is of course the linear impulse response.


Here's a sweep I generated (and its inverse) that I created using my own code, which have the properties I mentioned above.

You use sweep.wav when capturing the response of the system in question.

The then convolve inverse.wav and the resulting capture, which will give you something like the picture above (not quite that dramatic but conceptually the same).

:) I studied for 1 year Physics in College.......18 years ago. This is too much for me man, but still thx a lot taking time and trying to explain.

For now i am trying to find the best mic positions and with a 20 second sweep I am shooting IRs with multiple Pre Amp emulation of UAD, including also the Neutral Preamp with a calibration file captured with REW. This lets me take the impact of the preamp out so there will be also IRs representing a clear shot of just Mic&Cab, where as some other ones with API, Neve and UA610 emulations. They won't have the harmonic distortion information but still it captures their impact on the EQ curve.
 
Back
Top Bottom