Voxengo Premium Membership - All Voxengo Plugins For a Fixed One-Time Fee
Forums     Plugins     Impulse Modeler and Deconvolver To Aleksay: Canceling out the reverberation/sound of a specific space?


I was showing my friend your Deconvolver.  He then asked me this question:

"Ok.  I'm skeptical, but intrigued.  I poked around at this a little bit, and

I can see why this is really good for sampled source or DI recording.  But I

haven't seen how one might capture the inherent reverb characteristics of

the room you've recorded a live sound in, so that you could cancel that

pattern and apply a new one.  If we found something like that, we could

record the piano live in my living room, cancel the living room's

characteristics, and put the piano sound into Carnegie Hall if we wanted."

Is there a way to do that?  If we created an impulse of his living room, could we remove the sound of that space from a recording of his piano?

Thanks!!  Boogex rules!

-Gary


Deconvolution is an ideal process.  It will work seamlessly only when you have 'one point' sound source and 'one point' mic.

In case of piano, its keys and strings are spaced between each other.  So, when you capture piano, each note is played from a different point in space.  That is why it is not possible to perform 'dereverberation' with sufficient quality.

However, if you capture piano on a long distance (several meters), it becomes 'one point' sound source, and in this case dereverberation is possible.


Aleksey,

What about a very small room?

In my case, I'm currently limited to recording my Marshall 2x12 cab in a closet (roughly 3.5 feet square with a 10-foot ceiling).  I use two mics close to one of the speakers (though at different angles.).

I was wondering if there would be some way to create a replica of this closet in your impulse modeler plugin and then use it to *cancel* out whatever standing wave and harmonics are coloring the sound? (In this case it isn't the elimination of reverberation, in the traditional sense, that is needed.)

I'd really appreciate it if you could comment further on this, and provide a recommendation.


Impulse Modeler won't help in such case.

But you may use Deconvolver.  First of all, you should play the swept sine through your setup, then recover impulse, and then use it to deconvolve the guitar recording.

Of course, you should not change the speaker/mic arrangement between all stages (from impulse capture to actual playing) - otherwise you'll may get too much swishing.


Aleksey,

Wow - this is really cool...  I've never tried anything like this before, so a few follow up questions:

1) Are you saying I can run the "swept sine" through the Marshall 2x12 itself?  If so, I'm guessing the deconvolution process will remove the EQ effects of the amp as well as the room, right?  Can you comment on whether this might be a good or bad thing?

2) If I must use a different speaker in the closet to play the test tone, would it need to be exactly in the same place as the Marshall speaker?  Given the circumstances, do I have (hopefully) some margin of error in this approach (i.e. can I have good results if the monitor speaker and marshall cab speaker wind up being 3" out of sync?)

Aleksey - much appreciate the feedback, here.  This is the one problem I face in my studio that really bugs me.  I'm not sure that static EQ can accomplish the same thing as the method we are discussing.


wsc: 1) Are you saying I can run the ''swept sine'' through the Marshall 2x12 itself?  If so, I'm guessing the deconvolution process will remove the EQ effects of the amp as well as the room, right?  Can you comment on whether this might be a good or bad thing?

Sure thing, this will remove both EQ of the amp and the room response.  I think you may apply EQ again - should be fine.  However, a bit of cabinet response may be removed as well, so you may need to apply cabinet impulse as well (try www.noisevault.com for variants).

wsc: 2) If I must use a different speaker in the closet to play the test tone, would it need to be exactly in the same place as the Marshall speaker?  Given the circumstances, do I have (hopefully) some margin of error in this approach (i.e. can I have good results if the monitor speaker and marshall cab speaker wind up being 3'' out of sync?)

I suggest you to try Marshall first (running a dampened line signal through it should be fairly OK for this task), at its 'clean' setting.  Long swept sine test tone suggested, to reduce bad impact of possible distortion.

It's the easiest setup and won't cause any alignment problems.  Alignment problems may make this whole deconvolution thing unacceptable.

wsc: Aleksey - much appreciate the feedback, here.  This is the one problem I face in my studio that really bugs me.  I'm not sure that static EQ can accomplish the same thing as the method we are discussing.

No, EQ alone won't help, because when you EQing you are losing guitar body together with reverb.  So, you are basically doing nothing against the reverb.


OK - I just gave this a first try and had a few problems.  Can you verify the following steps and details are correct?

1) Create test tone (Sine Wave Sweep - generated by Deconvolver itself) [Clean Test Tone]

2) Record the test tone through the Marshall on cleanest setting, capturing with exact

mic and mic placement used to record the guitar tracks. [Closet Test Tone]

3) Create the "Closet Impulse" in Deconvolver:

- load "Clean Test Tone" into the "Test Tone File" field

- load the "Closet Test Tone" into the "File To Process" Field

- process with "Reversed Technique" (per instructions - the "Sine Sweep" test tone is being used)

- This creates the "Closet Impulse"

4) To remove the sound of the closet in a guitar track:

- load "Closet Impulse" into the "Test Tone" field

- select guitar track as the file to process

I've been unsuccessful so far. :-( I tried step #4 both with and without the "reversed technique" selected.  One way produced a silent file.  The other produced a loud, clipping, noisy file with no discernible content (just sounded like a machine or something).  I also experimented with the MP Transform option, but didn't get anywhere.

I'm not sure if the "Closet Impulse" I created was done right, either.  In case you have time to give a listen, I've placed the files created from steps 2 and 3 above here:

http://www.wscollins.com/decon/


You are correct with your steps.

However, you should not be using 'reverse technique' on the 4th step (if you do use it).


I also suggest you to add some silence to the "Closet Test Tone" so that "Closet Impulse" is not cut abruptly.  Generally, your existing "Closet Impulse" looks good (maybe a bit clipped?)

Anyway, I do not suggest you to do any post processing on the "Closet Impulse".  Use it "as is" for deconvolution, but make sure original "Closet Test Tone" has enough pre and post silence (there is an option in Deconvolver to lengthen this silence).

Also, please do use a 30 second sweep (without fades).  When recording the sweep, do not cut or post-edit the recorded sweep - in your case "Closet Test Tone" impulse seems to lack some of the leading oscillations.  Such "rude" editing does not work with impulses and deconvolution - all abrupt changes will be reflected on the final result.

This topic was last updated 180 days ago, and thus it was archived.  Replying is disabled for this topic.