Cancelling the effects of a room

Summing is taking it’s toll.

I’ve got this track where everything on it has the room – in other words it was recorded open-mic. 3 acoustic guitars and 6 vox all recorded in the same room. By and large, the room is decent–it’s a loft which spills into the [edit] FIRST [/edit] floor. The problem is that the summing of the room sound makes this track really stand out (in a bad way) with the other songs on the project.

Any thoughts on how I can minimize the summing w/o re-recording everything?

How many mics? Gating would be one obvious trick, but it won’t do a good job at all for your scenario. You need to remove the tonality that is added and bleed into other mics while the instruments are playing…while not gated. That’s like needing an anti-reverb plug-in.

I say use selective narrow band notching EQ to get rid of the EQ ranges you don’t want to hear. The room will have a few narrow ranges that can be removed this way. It won’t work well but it might help a little.

Another try it and see thing to do is to take a very quiet spot in the recording, but not a silent spot, and sample that for noise reduction. That’s normally done to silent spots as a way to remove tape hiss and other noise. It can be used to remove background sounds like whirring fans and disk drives. Likewise if there’s any where on the original track that there is “room sound” and no one playing our singing that can be used to as a removal sample. Don’t expect miracles. Remove as little as possible because it will take out A LOT of the instruments and voices since they were miced at a distance. They’re sounds are filled with what you are trying to get rid of.

Those are last ditch effort things. I’d be surprised is they work at all.

There is always the opposite approach - make all the tracks sound like the “live” one. This could be done with reverb or using “silent” sections from a recording of the room as a separate track.

I actually had the opposite problem, substituting a studio track for a live track in a CD of a live performance in front of an audience. I had to find sections of audience ambience from between songs to sum in to the begining and ends of the studio track. I found that it was only necessary to match the ambient sound at the transitions between tracks. Once the song was underway everyone’s attention is directed at the music and they don’t notice that the background “feel” has changed (if you fade it out slowly). If you don’t blend at the end they will also notice that discontinuity. I was fortunate to find enough quiet audience on trimmed parts of other tracks to do the job. In the future I would record audience ambience deliberately to have this possibility. It is surprizing how identifiable a section of “random” chair-creaks, breathing, coughing, etc. is. If you are trying to maintain the illusion of live you want to avoid repeating sections of audience ambioence.

You might find that the contrast is more of a problem than the ambience and that some way of controlling the transition works. Of course you now know the benefits of close-micing (and small-diaphragm mics which pick-up less ambience) and will choose the amount of ambience you want.

You might also consider high-pass filtering each track as high as possible for the pitch range of that particular channel. Much of the ambience is low-frequency and this might help. You may then be able to use regular reverb (perhaps band-limited) to mask any residual “funniness”.

Jim

Don’t try to make it what it isn’t. The only way to keep a room out is to keep it out from the beginning. Too late for that (unless you re-record it). Instead, try to frame it as a bonus track or intentionally set it apart somehow. It’s different: use the difference.

Just my $0.02, and I’m certainly no pro.

First, let me say thank you to you ALL for responding. Secondly, I think that Phoo should get the “reading my mind” award. :D There’s only one mic, and each of the tracks has been recorded individually. Or to say it another way, there’s six mics all of which are the same, in basically the same spot in the room which seem to be overtaking the ambience of the track.

So, to your point, Jeff, the room’s really not so bad IF it’s just one or two tracks along side of 6 or 8 other tracks recorded closed. It’s just that this is the ONE track on the whole project stripped down to just acoustic and vox. Most of the other tracks have drums or piano all of which are done closed, and thus the stark contrast.

So, to Phoo’s point, when you suggest trying to notch out the offending frequencies, would you point me to the individual tracks, the master track (or groups for that matter), or in the mastering stage?

Then, as a followup, since you brought up the (anti)reverb angle, would it make sense to back off on the reverb, or try and cover it up (OR: should I just use my ears?) ???

Seriously, is there any conventional wisdom that says something like, "adding reverb in that situation is pointless?"

Thanks much all.

Adding reverb to an already reverberant track is like trying to mask an odor with perfume, you can conceal it by making everyone notice a different stink. If it smells good enough it may be OK but most of the time it just makes things worse. My suggestion of using reverb after high-passing the tracks assumed that high-passing the tracks removed much of the natural ambience in that frequency band and that the additional reverb would be employed to blend with the remaining natural ambience which would still be present at higher frequencies.

I recently had to do something similar to mask a cut I made in a fiddle track. We recorded the session using loudspeaker monitors for a live-in-the-studio session (it is easier on the musicians than headphone monitoring and can work well enough if the playing is good). When I first checked things, the bleed was not too bad and I assumed that things would stay that way. Unfortunately, the fiddle player adjusted the mic position and then moved farther away. On mixdown I found that her level was low but more importantly, the bleed and ambience were almost as loud as the fiddle.

My harmonica sound was particularly loud in her channel which restricted how much I could drop my level without dropping hers as well. It also affected the tone of my harmonica by introducing a bit of delay and reverb. At one point in the mix I needed to drop her level to conceal a few bad fiddle notes, but doing so abruptly changed the harmonica tone due to the loss of the “reverb” component of the bleed.

In order to fix that I used an aux send envelope to increase the reverb level during the portion where I had to duck her level. This largely restored the tone of the harmonica and concealed the level manipulation. Not perfect but managable for a few notes here and there. If I had been more obsessive (and I might yet decide I want to be) I would have tried to exactly duplicate the delay/reverb characteristics from each channel and could probably make it inaudible. That’s more work than I am interested in at this point.

I would encourage you to try a few things that have been suggested, it can all be undone if it doesn’t work. You can do a “Save As” on your mix and do your fooling around on a copy of your .sng file. Reverting simply would entail using the original .sng and deleting the unsuccessful one.

Jim

I’m skeptical that Phoo’s tricks will help much (as phoo himself said).

I say again: this cut is different, USE the difference rather than trying to remove it.

There is no “unreverb” plugin. Noise gating is worth a try but is likely to be noticeable (when the gate opens and closing). Noise filtering using a wave editor works great if the problem is steady background noise in the room (hiss, hum, buzz), but it won’t remove the room’s reverb very well – at least, not without removing a lot of content. (If the room has serious fundamentals and you’re getting boominess or lots of reinforcement of certain frequencies, it would help a bit there – take the noise sample just after the vocals or whatever stops, just before the room’s verb fades to nothing. Work on a copy of the original, of course.

You might be able to dilute the room sound a bit by adding another track recorded elsewhere, like a duplicate of an acoustic guitar part. You could use the two as a mid-side pair, using the original as the “side”, which would turn the room color into a stereo imaging effect rather than purely coloration – sometimes that helps. Or use a comb filter on one of the tracks (e.g., MDA “stereo” plugin).

My only real world experience doing this was two drastically different projects. One was an old cassette of a piano done with a single distant room mic (portable with a built-in mic and auto-record leveler). The other was an attempt to salvage a good vocal performance that had way too much printed reverb on the track (I wanted to shoot the engineer, but he was a friend. I made him feel bad enough that he probably felt like shooting himself, but he didn’t, so we are still friends).

Both cases were VERY MUCH listen and experiment endeavors.

I used the sampling of the room sound method to remove what I could of the background on the piano track. There was some much room sound that it ruined the piano sound. I could only remove a few db of noise before artifacts were terrible. I did that in the old version of Cool Edit. That did help a little. After that I used a 31 band mono hardware EQ and a dbx compressor and gate to experiment with tone and further ducking-smoothing. That seemed to work better to my ears than plug-ins for some reason. After the EQ and dbxes I ran it through an Alesis Microverb to add a touch of stereo room back. That helped a LOT to cover the remaining artifacts and to fill in some dead spots created by the gating. The EQ was used as both cut and boost - a very little boost at the bright attacks helped, but that was some of the same frequencies that needed cutting to get rid of the room. The depth and lows were almost faked with the reverb. In the end I don’t think it was much better than the original, if any better at all. I was not happy with the results, but there was less noise and room sound, and the stereo spread was nice compared to mono. I’d say it was worse from a musical aspect.

The printed reverb track cleaned up much better. The reverb had a narrow band of frequencies that could be notched out pretty easily. Of course that also notched out the vocals at that spot and made them very thin. I put a hard gate on the track to get rid of ALL the reverb tails, then after the gate put the 31 band EQ. I notched out as little as I could so the reverb couldn’t be heard in the full mix. Fortunately, none of the other tracks had any reverb at all. Because there was no reverb on the other tracks notching and gating the vocal track made it sound dry pretty easily. It had just a slightly thin tone in the mix. At that point I added reverb to the snare track of the drums, and no reverb to anything else (the sound of the time was a wet snare fortunately). With that reverb on the snare I slowly removed the EQ on the vocals to the point it was almost not there. The vocal track sounded nice and full and the reverb on the snare masked the remaining reverb on the vocals. This worked great. It would not work at all if the mix was being done today where the typical mix is much drier overall. I did this long ago on an AKAI 12 track machine, bouncing the vocal over to another track to print it as a new track the way it was in the final mix. I was very happy with the result.

I suspect you’re dealing with the old piano problem unfortunately. One vocal track masked by a band is easier to manipulate that a bunch of tracks that will multiply the problem.

So, is this correct?

(1) You have a couple of tracks with the room ambience and some number of secondary tracks with little or no ambience.

(2) the EQ on the room reverberation sounds strange

I have a trick that you can use to fix (1), but on (2) you may be out of luck…

I had the soundman record some of my music during an open mic night in a bar. These base recordings were OK, but I wanted to add some more instruments into the mix, so I recorded some bass, additional guitars, backing vox, etc. (This sounds like the situation you are in?)

I wanted to make everything sound like it was recorded live, so I captured an isolated “clap” from the track and used that as the reverb impulse in SIR. As jankey at this sounds, it worked almost perfectly and the newly recorded tracks blended in seamlessly with the live tracks. If you still have access to the space, you could go back and capture an impulse now (using a hand clap or some other suitably short pulse. A firecracker, maybe?)

As for the resonant frequencies in the reverb, I don’t think that much can be done because EQing the verb will EQ the instruments as well. If you could use a gate like phoo suggested, you could split the tracks into “instrument+ambient” and “ambient” parts and apply a different EQ to the “ambient” part hide the unbalanced frequencies.

Maybe I misread. My understanding is he has a number of songs. On all songs, several tracks are recorded with room ambience. On most songs, more tracks were recorded directly, and these mixes sound fine – the direct tracks dliute the room ambience enough. But on one song, it’s just the miked tracks. This song sticks out in the collection of songs.

I think that just about anything he tries to do to the only tracks in a minimal mix for a given song will do more harm than good, because there’s nowhere to hide. I think there are really three meaningful options: (1) try to make that difference work and sound good by framing it inbetween the right two songs, or end the CD with it (maybe not the best idea) or make it seem like a “bonus track” or something, (2) add a couple tracks, or (3) re-record it.

I’m probably sounding like a crunmudgeon and pessimist. It’s worth trying some of the things mentioned above, but I wouldn’t spend a lot of time trying to do something that at best you can only partially do, and just about anything you do is likely to do more harm than good.

Most of the time, “less is more”. It’s better to filter out than boost, etc. In this case, though, I think that subtracting is unlikely to be satisfactory, so adding a track or two might be the ticket. Or just making what you have work by allowing it to have a “live” ambience. (Add a laugh track! :;): )

Maybe you could post a quick mix of the problem song and we’d have some ideas or more specific suggestions.

This is the reason I love N-track: The forum. Great discussion and sharing of ideas.

Ben:

You’re close. All of this is being done in my house, upstairs. I’m using an Equitek to record everything. It’s what I’ve got. I recorded two acoustics (open mic) panned left and right. Then I’ve got a lead vox, panned straight up. Two backing vox and then another 3 vox which are only present on the turnarounds and panned left and right as appropriate. On the turn arounds, there are a total of 5 channels all of which were recorded in the same room with the same mic in basically the same place.

So, just for grins, say that the room resonates at 2.5k. Now I’ve got 5 instances of 2.5k where it’s resonating. If you isolate one track, it doesn’t really seem that bad. It’s the combination of all 5 tracks that makes the room seem overly present.

Jeff:

I like the idea of using the difference. The one mitigating factor is that I really like this tune in comparison to the rest of the project. IOW, musically, I think this track is one of the better. It’s probably not one of the better of my “engineering” feats, but again, I think it’s just as much a function of the room as it is the engineering.

That said, I’d rather not make it a bonus or easter egg, but I AM open to using the difference as you said. Do you have any other suggestions? I’m really starting to think that I may have to re-record it, but even if I do, I’m still stuck in the same room with the same mic and requiring the same number of tracks – thus likely the same result. I’m open to any suggestions as far as rerecording at this point, OR how to mange the room better – whatever it takes (short of spending money). :D

I really do appreciate all of the input and suggestions.

Hi:
If I can add and not be in-the-way, some “Room” is great… It’s like how much is too sweet? Mabey you can try to put these tracks in-and-out of Phase with each other… If you haven’t tried that yet… There may be vairying degrees of altering the phase that might spike or peek at the correct frequency that is in question… I hear you say it is in-and-around 2.5khz. ? Mabey, this plug-in might help you in your quest to remove this “Room” sound/issue… Kelly Industries has this one as a FREE plug-in… Basiclly, it’s an adjustable phase plugin…

Stereo Image/Phase Tool VST

You might whantta try cloneing the track/Tracks individually… and applying the plug-in and then. re-rendering each track and apply the rendered tracks to a new time-line and then remixing the rendered tracks. This might remove the “Room” from the tracks… IT could be worth a rendering session…

Bill…

So far, I think the best bet is to sample the reverb tails after the music stops and just before the tails slide into oblivion – where you’ll get the most “room”, and denoise-filter using that sample (you know how to do this using a wave editor, right?)

For starters, try it on a mix, but it might be best to do it individually for each track and just take a smidgin off each track. Perhaps more in the back vox, since when they come in there’s other stuff to hide in – if filtering leaves a hole, that hole will be pretty much filled by the main vox.

After that, try messing with some band cut filters at what seems to be the room resonance frequency, but easy does it – and again, you might find the frequency while filtering the whole mix, but then apply it to individual tracks and vary the frequency a bit on each track, for 2 reasons (1) to tailor it for the track, but more imporantly (2) to spread it out and make the cut less obvious in the mix.

And I guess we’ve all learned a lesson: when recording lots of tracks in a room, consider doing the following:

1) move the spot
2) rearrange stuff in the room (gobos, sound curtains, the sofa)
3) change mikes
4) change soundcards

I mention the last only because the same problem, buildup of “sameness”, can happen when you have too many tracks with the same preamp. Or so I’ve heard: I don’t have enough serious experience to have learned this concretely but it sure makes sense. And every mike hears the room a little different, so that would help a bit too.

A little more clarification might produce a few more suggestions. Is the problem with the recorded ambience or with the effect of resonances on the tone?

If the problem is the recorded ambience you have had a number of good suggestions but the overall prognosis is not too good because that is hard to remove.

If the problem is the influence of ambience on tone you have more options. You can equalize each track separately and differently which will tend to decorrelate the tone problems so that each track shows more variation with respect to the others. You can try to identify the room-modes and apply EQ to reduce the strength of the signal so that the sum of the room and instument tone is flatter than you achieve by ignoring the contribution of the ambience in your track EQ.

One way to identify the roon modes and their influence might be to place a loudspeker playing white noise (FM interstation hiss is an easy source of this) in the location the musicians were, mic it as you mic’ed the original tracks and record the output. Look at the spectrum display in the equalizer dialog and see if you can identify the peaks and nulls. You can then try to EQ that “flat”, save the EQ as a preset then try it out on the other tracks or as an overall EQ for the master channel.

If the loudspeaker is not flat it might be useful to first run the experiment outdoors to see the “anechoic” performance of the speaker. You could then EQ the “anechoic” response to flat (over a reasonable band) and use that EQ as your starting point in the room analysis. You would then note the frequency, gain and Q changes necessary to flaten that response and make a preset using those values.

Jim

This procedure could help you find the frequencies that most affect the tone.

Hmmmm… well this is a pickle. I really don’t think you’ll get very far with any of the noise cancelling schemes unless you could’ve captured an impulse response in your original recording.

Do you notice the 2.5k resonance when (for instance) during the singing or only in the reverb tail? If it is, then plain old EQ should solve the problem. It may be that you are having a hard time cancelling the “strangeness” of the frequency response because the room has colored the sound in a complex way. You should be able to flatten the response if you can analyze the frequency spectrum of an impulse from one of your recorded tracks.

Who knows if this will even work, though, because boosting bands in the EQ will add in phase distortion that will make the signal sound strange, as well.

Best of luck.

Well, the problem has been solved. Although crow has never been one of my favorite dishes, this will be a heaping dish of it.

First, let me say thanks to each of you for your kind helps along the way. I believe that you’re questions, thoughts and ideas are what finally solved this for me.

I went to the console and started by isolating the vocals group. The problem was still there, but not as bad. I also noticed that there was a load of distortion on the lead track, but the wave didn’t look pegged – not even close. I isolated the lead vox. Same thing. I opened the lead vox in Wavelab, and the problem went away – distortion and all. ??? Hmm. Something’s rotten in Denmark (no offense to our European brethern). :;):

Back to N and the list of lessons learned:

1. Did you know that N-track’s compression tool has an output gain that should be used to keep the level below 0 db? Well, I didn’t and every one of those tracks were absolutely slammed. Been using N for over a year now and JUST NOW figured that one out. Go figure.

2. Overly compressed reverb (complete with distortion) doesn’t sound good on a lead vocal or backing vox for that matter. At least in this genre.

3. When you have a problem, break it down into parts. That rope you think you’re looking at may just be the tail of an elephant.

4. Learn to be contrite. There’s enough arrogance in this world as it is. You may not know everything.

5. That little master effects button is more useful that it may seem. When in doubt, go back and listen to everything with OUT ANY EFFECTS.

6. If the tracks sound good without any effects and start sounding like c*ap with a ton of effects on it, then take the effects off!!!

Ok, so the nut was that the room isn’t really that bad as long as I’m not compressing the snot out of everything and keeping the output from overdriving the tracks. Back everything down and wow!!! It’s no longer different.

Thanks much to everyone for the help.

<!–QuoteBegin>

Quote
1. Did you know that N-track’s compression tool has an output gain that should be used to keep the level below 0 db? Well, I didn’t and every one of those tracks were absolutely slammed. Been using N for over a year now and JUST NOW figured that one out. Go figure.


Compression will generally increase the effect of reverb tails, for obvious reasons.

Other than clipping in the master channel (which obviously we avoid), going over 0dB on the output of a plugin, especially one like a compressor, usually isn’t very significant as long as there’s no clipping in the master, or the next plugin in the chain after the compressor isn’t being overdriven.

Genearally, any compressor has a post gain. If there are no plugs after the compressor, adjusting the post gain is identical in actual effect to adjusting the track fader (assuming the compressor is on a track). Note that “after the compressor” includes plugs on aux tracks, for post fader&fx sends.

Bottom line though, you’re right about your approach to solving this problem.

The “always on” effects in n-Track V4 were a terrible idea, IMHO. No effects should be “always on”, they should only be on when specifically turned on. That might not be a culprit here, and since I only used V4 for a little while, I might be misremembering it. (Plus I disabled all “always on” effects first thing, since I don’t want effects unless I want 'em!)

<!–QuoteBegin>

Quote
Note that “after the compressor” includes plugs on aux tracks, for post fader&fx sends.


This would also include groups with limiters on the vox’s, yes? That might have been the real problem, but it was solved by “unslamming” the compression. Really do appreciate the help and insights.

yes, that’s it.