The No.1 Website for Pro Audio
Removing small room sound via reverse convolution process?
Old 3rd September 2018
  #31
Lives for gear
 

Quote:
Originally Posted by Jack Luminous View Post
Quote:
Originally Posted by Bruce Watson View Post
I think you're making an interesting assumption that's not true: that the process of applying reverb to a dry source is symmetric to the processes of a de-reverb algorithm. And it's not at all symmetric. You can't just apply a "reverb algorithm in reverse to remove the ambient signature of the room" any more than you can apply a recipe in reverse and remove the salt from a cake.
Yes you can and it is called a deconvolution process. It's the exact inverse of a convolution process (what is currently used in commercial IR-based reverb plugins).
In a small room without enough absorbing material to absorb frequencies evenly (and stop all reverberation), the standing waves create a complex, harsh sounding eq which cannot be undone. Electronic processing makes the standing waves sound even worse.
Old 3rd September 2018
  #32
Lives for gear
 
Bruce Watson's Avatar
 

Quote:
Originally Posted by Jack Luminous View Post
Yes you can and it is called a deconvolution process. It's the exact inverse of a convolution process (what is currently used in commercial IR-based reverb plugins).

Theoretically, it should be possible to first record an impulse thru the mic doing the recording, then record, then remove the ambient reverb using the recorded impulse. It just happens that there's no available commercial product yet.
And theoretically, you really can remove the salt from a cake after it's been baked. It just happens that there's no available commercial product to do that yet.

Given the demand for de-reverb type products, which you can judge from the number of and the rising quality level of products already on the market, I'm thinking if somebody could make a deconvolution product work in the real world, we'd have a product. Since there is no product, I'm not persuaded that it's possible.
Old 3rd September 2018
  #33
Lives for gear
 
whippoorwill's Avatar
excellent preamps and ADs with dynamic microphones or really good hypercardioids seem to be the best bet. other patterns up close.
Lot of great smithsonian stuff done like that.
The real trick is just to go to a better room.
There's no taking out reverb really, its mostly better to just roll with it.
Old 3rd September 2018
  #34
Lives for gear
 
brockorama's Avatar
 

If a small bad sounding room is unavoidable, then accusonus era-d seems to do what it suggests on the tin. I don't use it , but have seen videos and read some testimonials.
Old 3rd September 2018
  #35
Lives for gear
 

Possibly check out: Sonible Freiraum plugin. It has a feature to adjust the relation of direct sound and reverb within any frequency range.
Old 4th September 2018
  #36
Lives for gear
On individual tracks, not a mix, I've taken out some of the small room sound with gate to silence the (short) reverb tail and some compression to emphasize the attack before the tail. It worked because the artifacts were masked by the rest of the mix.

On a stereo mix, you might get some success EQing certain frequencies in or out depending on their interaction with the room, messing some with mid-side processing, etc, but it will be serious trial and error / results vs artifacts. Can't imagine a program could be a one size fits all solution, or contain any tools that anyone with good enough training to use wouldn't already have.
Old 4th September 2018
  #37
Gear Addict
 

Quote:
Originally Posted by Jack Luminous View Post
Yes you can and it is called a deconvolution process. It's the exact inverse of a convolution process (what is currently used in commercial IR-based reverb plugins).

Theoretically, it should be possible to first record an impulse thru the mic doing the recording, then record, then remove the ambient reverb using the recorded impulse. It just happens that there's no available commercial product yet. I've not looked closely at the matter but I don't see any reason why it wouldn't be doable.

Of course, it wouldn't be a perfect process so reverb wouldn't be 100% removed, just like convolution reverb is slightly different from a real recorded reverb. I think it could produce interesting results if done properly.


I would ask two questions:
1) Is the transfer function of the reverb minimum phase?
2) Do you have an infinite precision computational device to do the inversion?

I think absence of either means any result would, at best, be quite imperfect. After all, a little noise presence really mucks up a Wiener filter, eh?


DG

Last edited by dgpretzel; 4th September 2018 at 07:42 PM.. Reason: correct bad term
Old 4th September 2018
  #38
Lives for gear
 
AlexK's Avatar
 

You would also need:

1) a speaker and microphone places in exactly the same location for the IR used for each instrument recorded
2) a speaker which has EXACTLY the same directivity as the instrument you’re trying to ‘de-convolve’ the room sound from (impossible - instruments are extraordinarily complicated in their three-dimensional sound radiatative pattern)
3) a perfect temperature and humidity controlled room as any change in density or volume of air in the space will cause a shift in frequency (subtly, but enough) of the rooms’ issues (or in the case of humidity the HF decay), leaving the IR/recording relationship incorrect.

If you can’t meet all of these criteria then your ‘reverse-convolution’ will actually add more noise.
Old 5th September 2018
  #39
Lives for gear
 
James Lehmann's Avatar
 

Some classic Gearslutz obtuseness in this thread.

It's as if the OP had asked: "Given that I have a recording with clicks and pops on it, might there be a solution that would mitigate the final result."

Why do folks opt to dumb down their replies with: "You shouldn't have a recording with click and pops on it"?

Firstly, replying to someone who has been here as long as 'jnorman' in such simplistic terms is surely a bit patronising.

Secondly, someone did indeed ask themselves a similar question to the OP's once upon a time and now we have really good solutions for clicks and pops like IzotopeRX etc, that would have been unimaginable a few decades ago.

What's wrong with informed inquiry, borne out of experience and curiosity?

I don't get this place sometimes.
Old 5th September 2018
  #40
Lives for gear
Quote:
Originally Posted by James Lehmann View Post
Some classic Gearslutz obtuseness in this thread.
I think you might be jumping the gun there, or at the least mis-characterizing the postings. What I see in the 38 responses are : encouragement to Jim to prevail yet further with more acoustic treatment (some has been done already...more necessary); some helpful suggestions of software to try; some alternate strategies (use cans, render backwards, use dynamic/more-directional/closer miking).... as well as rationale of why it's so difficult to 'unbake the cake' of undesirable reflections embedded in a recording.

That's a pretty broad church of replies, and the OP can opt to go down any or all paths of investigation, as he chooses. Acknowledging the difficulty of 'de-reverberating' isn't the same as flatly stating it's impossible....and if, at the end of exhaustive testing of all suggestions, Jim is forced to conclude that more treatment or another room is the only solution, at least he will have explored the alternatives first....assisted by well-intentioned contributors here ?

In summary...it appears to be a typical, constructive GS Remote Forum reply thread to me, and not at all dismissive nor unsympathetic to his plight.
Old 5th September 2018
  #41
Lives for gear
 

Quote:
Originally Posted by dgpretzel View Post
I would ask two questions:
1) Is the transfer function of the reverb minimum phase?
DG
A priori, I would answer yes. For the transfer function to be minimum phase , it would need to be both BIBO stable which seems to be the case intuitively (I may be wrong though) and also causal, which is clearly true in this case.

Quote:
Originally Posted by dgpretzel View Post
2) Do you have an infinite precision computational device to do the inversion?
Infinite precision doesn't exist in computational devices but that doesn't prevent said devices to perform very accurate computations in many situations. Deconvolution processes are routinely used in imaging. I don't see why it couldn't be used in audio, based on the same principles.

My background in signal theory is old and wrinkled but your arguments don't hold much ground at first glance.

Last edited by Jack Luminous; 5th September 2018 at 02:25 PM..
Old 5th September 2018
  #42
Gear Addict
 

Good points.

I made no argument.

Certainly deconvolution is doable and done often. I would also point out that, in your example of imaging, probably the most highly visible examples are images from space, which have been deblurred for decades. I would also point out that the point spread functions of space imaging devices are know rigorously, to a very high degree of precision. (Important when dividing by very small numbers, no?)

I wonder how precisely we measure the impulse response of a small room.

The devil is in the details, not the theory.

I intended no trolling with either of my posts.

Regards,

DG
Old 5th September 2018
  #43
Lives for gear
 

I would use the array that I was going to record the instrument with in the same location. Use a mono speaker to play the test tones at the same location as the players'
Insert the recorded test tones in the convolution vst.
Record the artist, dbl the recording on a second track. Insert the convolution on that track and set to 100% wet.
Invert the phase on the convolution track and bring the volume up from 0 and see if it helps attenuate the room. You might have to move the track forward by *ms to accommodate latency.
Old 6th September 2018
  #44
Lives for gear
One approach that hasn't been mentioned so far...but is a proven method of reducing room-based reflections....is to try a boundary layer/PZM type of approach.

I don't recall what range of SD mics you have, but simply trying an AB spaced pair of omnis and/or an ORTF pair of cardioids and placing them against the wall (parallel to) with stands (leave a 1mm gap between mic head and wall) should give you enough of an experimental feel to see if it's a candidate or not ? Best location would likely be high, and pointing down the wall, towards the floor .

Make sure it's an untreated wall though (or remove some treatment to make it so)....you need a hard plane surface to make the principle work.

Just try it in your studio room, then talk, sing or play flute at the wall (the wall becomes the microphone)

If it works encouragingly, then you could either kludge together a home made mount, and even double-sided-tape it/them to the wall....or spring for the official versions as below....

Boundary Layer Mount for d:screet™

BLC | SCHOEPS Mikrofone

Schoeps BLCG Boundary Layer Mounting Plate BLCG B&H Photo Video

Boundary layer microphones | SCHOEPS Mikrofone

If it works...cheap solution, sidestepping additional room treatment and purchase of new mics !

People poo-pooh boundary layer miking, but it works significantly well for theatre and opera productions....and if your back's against the wall (as it seems to be, in trying to make your studio work for student recordings) ...well, you might as well put your mics against the wall instead

Please try ....and post your audio trials here for evaluation !

PS....maybe a corner placement would give an even further reduction of flutter echoes, and more boundary reflection free gain for your mics ?
Attached Thumbnails
Removing small room sound via reverse convolution process?-flute.jpg  

Last edited by studer58; 6th September 2018 at 03:10 AM..
Old 7th September 2018
  #45
Lives for gear
Quote:
Originally Posted by studer58 View Post
One approach that hasn't been mentioned so far...but is a proven method of reducing room-based reflections....is to try a boundary layer/PZM type of approach.
Good suggestion. I've experimented also with mid-side boundary layer with even better results than spaced pairs. Taped a pzm to a low ceiling, then put a figure 8 directly underneath. Usable results in some situations.
Old 7th September 2018
  #46
Lives for snowflakes
 
12ax7's Avatar
 

Quote:
Originally Posted by studer58 View Post
One approach that hasn't been mentioned so far...but is a proven method of reducing room-based reflections....is to try a boundary layer/PZM type of approach.

I don't recall what range of SD mics you have, but simply trying an AB spaced pair of omnis and/or an ORTF pair of cardioids and placing them against the wall (parallel to) with stands (leave a 1mm gap between mic head and wall) should give you enough of an experimental feel to see if it's a candidate or not ? Best location would likely be high, and pointing down the wall, towards the floor .

Make sure it's an untreated wall though (or remove some treatment to make it so)....you need a hard plane surface to make the principle work.

Just try it in your studio room, then talk, sing or play flute at the wall (the wall becomes the microphone)

If it works encouragingly, then you could either kludge together a home made mount, and even double-sided-tape it/them to the wall....or spring for the official versions as below....

Boundary Layer Mount for d:screet™

BLC | SCHOEPS Mikrofone

Schoeps BLCG Boundary Layer Mounting Plate BLCG B&H Photo Video

Boundary layer microphones | SCHOEPS Mikrofone

If it works...cheap solution, sidestepping additional room treatment and purchase of new mics !

People poo-pooh boundary layer miking, but it works significantly well for theatre and opera productions....and if your back's against the wall (as it seems to be, in trying to make your studio work for student recordings) ...well, you might as well put your mics against the wall instead

Please try ....and post your audio trials here for evaluation !

PS....maybe a corner placement would give an even further reduction of flutter echoes, and more boundary reflection free gain for your mics ?
This is all good information!

...However, not of much use when you're stuck with tracks that are already recorded.

I think that what the OP was really more curious about was the potential of using IRs to improve already-recorded tracks. (Different problem.)

I (for one) happen to find the prospect to be quite intriguing, and think that one key to perhaps making such an idea work would be understanding how to deal with the propagation of various frequencies within the room (as exhibited by the image you posted):
The biggest problem I can see with using IRs to negate the effect of the room is that (as far as I can tell) any chance of making it work depends upon knowing ahead of time that it needs to be done, and going through great pains to make it happen.

...But if you're gonna bother to go through all that hard work before capture, then why the hell wouldn't you just spend a whole lot less time effort and energy by tracking in a better room in the first place?
(It would be less trouble.)
.
Old 7th September 2018
  #47
Lives for gear
 

Quote:
Originally Posted by 12ax7 View Post
This is all good information!

...However, not of much use when you're stuck with tracks that are already recorded.

I think that what the OP was really more curious about was the potential of using IRs to improve already-recorded tracks. (Different problem.)

I (for one) happen to find the prospect to be quite intriguing, and think that one key to perhaps making such an idea work would be understanding how to deal with the propagation of various frequencies within the room (as exhibited by the image you posted):
The biggest problem I can see with using IRs to negate the effect of the room is that (as far as I can tell) any chance of making it work depends upon knowing ahead of time that it needs to be done, and going through great pains to make it happen.

...But if you're gonna bother to go through all that hard work before capture, then why the hell wouldn't you just spend a whole lot less time effort and energy by tracking in a better room in the first place?
(It would be less trouble.)
.
That's why I came up with this.

Quote:
Originally Posted by emenelton View Post
I would use the array that I was going to record the instrument with in the same location. Use a mono speaker to play the test tones at the same location as the players'
Insert the recorded test tones in the convolution vst.
Record the artist, dbl the recording on a second track. Insert the convolution on that track and set to 100% wet.
Invert the phase on the convolution track and bring the volume up from 0 and see if it helps attenuate the room. You might have to move the track forward by *ms to accommodate latency.
Old 7th September 2018
  #48
Lives for snowflakes
 
12ax7's Avatar
 

Quote:
Originally Posted by emenelton View Post
That's why I came up with this.
Yeah, I'm with ya there, mate.

...Still, I think the idea might be worth perusing though, just because there are times when we are forced (for whatever reason) to try to do good work in bad places.
.
Old 7th September 2018
  #49
Lives for gear
 

Quote:
Originally Posted by 12ax7 View Post
Yeah, I'm with ya there, mate.

...Still, I think the idea might be worth perusing though, just because there are times when we are forced (for whatever reason) to try to do good work in bad places.
.
I thought a lot about it. It's the 'backwards' of tacking on a verb but needs to actually be precise. So I thought that the commonality of the mic placement for the IMPULSE recording and the performance recording being the same point would be where the CONVO subtraction effect could possibly be made to work.

Appreciate the positive remark.

Thanks

Last edited by emenelton; 7th September 2018 at 08:39 PM..
Old 7th September 2018
  #50
Lives for snowflakes
 
12ax7's Avatar
 

Quote:
Originally Posted by emenelton View Post
I thought a lot about it. It's the 'backwards' of tacking on a verb. So I thought that the commonality of the mic placement for the IMPULSE recording and the performance recording would be a point where the CONVO subtraction effect could possibly be made to work.

Appreciate the positive remark.

Thanks
I understand.

You certainly have given this a lot of thought.

I would go as far as to say that the proposition posited by the OP is not without merit (and the more I think about it, the more intriguing it gets).

Upon its face, it would certainly make much more sense to use a good room (if possible).

But what (if for whatever reason) that was not an option (and you had the ability to use something like this)?
.
Old 7th September 2018
  #51
Lives for gear
 

Quote:
Originally Posted by 12ax7 View Post
I understand.

You certainly have given this a lot of thought.

I would go as far as to say that the proposition posited by the OP is not without merit (and the more I think about it, the more intriguing it gets).

Upon its face, it would usually make much more sense to use a good room (if possible).

But what (if for whatever reason) that was not an option?
.
Well, I've never done an impulse recording LOL

When you do stuff like this you can normally pick up a few decibels before things go south. Still it would be interesting to see what happens as you raise the fader. Probably would end up sounding like a farmer had plowed a trough through your nice verdant field!
Old 1 week ago
  #52
Gear Head
 

Quote:
Originally Posted by robert82 View Post
This should be a sticky at the top of both the Acoustics forum, and the Newbie's forum. Minding this simple bit of wisdom would prevent a lot of needless frustration.
This is not strictly true.

Adobe Audition's "Capture Noise Print" can, in a sense, remove room acoustic. It depends on "what the character of the noise is" that it samples. If that ambience contains elements of the room acoustic, it will be deconvolved from the recording.

I've done it - to devastating effect!

A colleague composer had a recording of a solo trumpet piece recorded live in a church. He wanted the reverberant field of the church removed. I indicated to him what you said here - that it can't be removed, but to give it to me and I'd see what I could do - thinking EQ and all the rest of it.

I ran Adobe Audition's "Capture Noise Print" on a half a second of "ambience" of the church before the trumpet started - and when I applied removed that "noise print" using it's Noise Reduction process - low and behold the entire reverberant filed of the church was removed. It was _so_ devastatingly good that we could here that the trumpet player was nervous - there was a quiver in his lips that affected his embouchure - so we couldn't use the recording dry. But I could not believe how "dry' it sounded once that noise characteristic removal had been applied - it was nothing short of extraordinary.

I accept that that this isn't the norm - I've tried this since on other recordings but the noise print in those just removes a bit of noise and not the room acoustic - but in the instance described above - it really did - so convincingly that, as said, it revealed a problem with the performance that wouldn't allow us to use it dry!
Old 1 week ago
  #53
Lives for gear
 
jnorman's Avatar
Knolan - I believe audition’s “capture noise print” is a subtractive (phase inverted) eq used normally to remove hvac and background noises. Odd that it would have such an effect as in your trumpet example, but an interesting tale nonetheless.
Old 1 week ago
  #54
Lives for gear
 
jnorman's Avatar
BTW, for those interested in this subject, I have lately been using Acon Deverberate to address the problem of “small room” sound, which I face so often doing audition recordings in various pianist’s living rooms. There is a good discussion of the plugin, along with comparative samples demonstrating what the plugin can do, here:

How to use Avon Deverberate?
Old 1 week ago
  #55
Lives for gear
 
emrr's Avatar
Apologies if I missed it: consider using processing on a parallel copy and blending. I’ve gotten better results a few times. The processed copy can be more obvious that way when soloed.

With RX7 de-bleed I’ve had better luck at times de-bleeding a copy of the thing I want removed so as to have a better defined version for the target track de-bleed process. It might be possible to close mic the instrument and use that as a bleed source to create an ambient version of the intended recording which could be subtracted from an unprocessed version.

In short, combine techniques to increase the shaving angles.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump