Help needed on denoising a 2ch classical recording

[Gaston69]

Hi Gearslutz,

I need some advice/help.

Last friday I recorded a classical group of 7 musicians playing barok on string instruments like lute and mandoline and flute from the 16 & 17th century. The venue was a small Luthers Church with very very nice acoustics and was loaded with approximatly 150 people (almost sold out).
My mic stand was approx 4 meters in front of the group and about 3 meters high, and the audience was just sitting behind my mic stand, unfortunatly one listener hit the stand with her foot during the show (shit happens).
On the stand I had mounted 2x Bruel & Kjaer DPA 4006 A/B 60cm which run into my AMEK DMCL pre-amp and all recorded in Logic 8 and parralel to my DAT in case my Macbook would crash. Due to the level of applaus my average recording levels were about -20/-12 db.

Now I am home and listening to it I can hear quit a lot of noise such like:

Back ground noise from audience.
Noise from the microphones/equipment.
Sub-Lows from cars passing the church.
People entering the church (slamming doors, however it's far in the back ground)

I tried X-Noise from Waves, Low Cut, R.Compressor, L2 and it did toke away a lot of noises and it pulled up the levels to a 0.2dB however it doesn't sound like classical music anymore i.e. it start sounding to compressed, not open anymore, loosing 3D etc etc.

Currently I am mixing in the box with Peak, Logic 8, Waves Platinum.

Can you give me some advice on how:

Pull up levels
Denoising (using above mentioned plug-ins?)
Maintaining it's 3D panorama, life

The music is played very nice, acoustics fantastic, the mics picked up the vibes and ambience perfect however I need to find a way to make a good professional CD for this band.

Please share your ideas with me because the music is it worth, lates I will post some takes.

Thanks,

Gaston

[nickynicknick]

Re Levels...if the applause is way too loud....find your loudest peak (music without applause) and make that your high point (-4ish)...then you'll have to mix your tracks down when the applause kicks in so that they don't peak at more than -4....don't use limiting for this type of music.


Re noise reduction....I'm afraid there's not much you can do that won't make it sound worse IMO....it's a live recording...people were making noise around you...you.can't change that....maybe next time try a closer micing situation...a stereo pair over the musicians can be very nice


Maintaining 3D panorama....don't use limiting, noise reduction plugs or drastic eq.

Good Luck.

Nick

[ghellquist]

As for levels - ride gain, generally no compression on classical. Limiter yes, working only on a few places here and there.

In your program, turn down applauses and raise music levels using automation or whatever works for you. My favourite program, Samplitude, does this quite easily.

I have had surprisingly good results on noise reduction using Samplitudes built-in tools, spectral cleaning is one of them. What you do is to "paint" the offending frequencys and they will be diminished. Does not always work but can be quite good at times. I have been told there are even better tools but the built-ins work for me. May take a bit of testing though, sometimes creates artifacts.

If you have the possibility to publish a short segment of music, say a minute or so I can show what can be done and you can hear if you find it worth it. I believe some of the people on this forum might be even better at reducing noise. Preferrably wave files (mp3-s are sort of already mangled).

Gunnar Hellquist

[nickynicknick]

Originally Posted by ghellquist

As for levels - ride gain, generally no compression on classical. Limiter yes, working only on a few places here and there.

In your program, turn down applauses and raise music levels using automation or whatever works for you. My favourite program, Samplitude, does this quite easily.

I have had surprisingly good results on noise reduction using Samplitudes built-in tools, spectral cleaning is one of them. What you do is to "paint" the offending frequencys and they will be diminished. Does not always work but can be quite good at times. I have been told there are even better tools but the built-ins work for me. May take a bit of testing though, sometimes creates artifacts.

If you have the possibility to publish a short segment of music, say a minute or so I can show what can be done and you can hear if you find it worth it. I believe some of the people on this forum might be even better at reducing noise. Preferrably wave files (mp3-s are sort of already mangled).

Gunnar Hellquist

seriously...if you want to maintain authentic classical sound, don't use noise reduction plugs. They're garbage. The only noise reduction that is good is phase cancellation....if there is a noise i.e a hum or buzz that you can capture and record on a separate track then flip the phase. It will cancel the original, this only works for consistent even hums and buzzes though

[pkautzsch]

Getting the mics closer to the music and farther away from the audience is key. Often that will mean using more than two mics to get a well-balanced image.

I had a similar situation recently: lute, baroque room seating about 150, very impolite audience (lots of coughing, coming late, and even TALKING!!!). Sound of my main pair of KM83s is great if there wasn't that loud audience. Fortunately I had a pair of cardioid spots quite close. I used these for most of the sound and had the KM83s fill in the "pressure transducer" decorrelated bass range only. Took a bit of track delaying (due to different distances) and filtering (LP @~500 on the omnis, HP @~400 on the cardioids), but worked out quite well. The image isn't as spatial as with the omnis only, but way less distraction, and that's more important to me in this case. Can't really get rid of the noises, but you can get them more into the background, so the coughing isn't louder than everything else.
There's a very expensive NR software by Algorithmix, which is said to be very effective even for coughing and door-slamming. No first-hand experience with this though.

[Gaston69]

Well guy's I think you are all right, what I will try is to normalize up to -0.4 without the applause and use maybe a little limiter and cut off below 60hz because in order to cancel out the sub-low from passing cars.

Peter, buy the way the band I recorded was Cappela Asquigrana, they are from Koln/Aachen/Vaals NL, do you know them?

Gaston

[thenewyear]

Automate the freq on the HPF so it is higher when the cars are passing. Up to 200Hz is probably possible momentarily on baroque programme material especially if the curve isn't too steep. This may also help to reduce the impact of the stand contact.

Coughs and splutters are VERY difficult to deal with. Take a look at the Cedar restoration tools if your budget would stretch to a rental of those you *might* get some useable results.

As others have said ride the gains down by around 12dB during applause (also consider editing down the duration of the applause to 6 seconds or so). You might want to leave the gains down for the wild track/room noise before the next track starts if this is distracting before gently bringing it back up again 2 seconds or so before the next track (although don't start this ride on the track ident as it will sound artificial/distracting if someone is selecting specific tracks on the disc).

Careful with the limiter, this shouldn't really be necessary as baroque music doesn't have anything like the same massive dynamic shifts that you get with classical music. If I were you I'd ask yourself why you are considering using it both technically and sonically/aesthetically by limiting this material.

[0VU]

A few thoughts

Simply recorded, natural sounding, things are much harder to hide post production 'fiddling' in. Tread very lightly and do the minimum processing possible.

Compression/levels:
Your average levels sound fine for live recording. (Assuming the recording was 24bit - if it's 16 bit I'd normally go higher (say, -6dBFS ish maybe more on predictable programme) and let the applause clip if necessary - it's not easy to hear clipping in appluase unless you really hammer it into the end stops - and even then it's not always easy!)

Bringing up levels on the music will also bring up the background noise.

'Conventional' compression (i.e. it reduces things above a certain threshold) tend to be rather unpleasant on classical music - especially simply recorded and simply scored music. As Gunnar suggests, if you really must compress, the best way is with musically sensitive, brain (not electronically) driven manual gain riding. Using fader/level automation makes this easier. If you can't bear to do this and have to use 'a compressor', try setting it up for parallel compression as it'll be a lot less obtrusive and retain much more of the natural dynamics of the music.

I'd never expect to use 'a limiter' on any kind of classical music unless specifically requested to by artistes/clients. Ime, there's always a better answer. That's not to say I never 'limit' - I just don't use a limiter for it.

Again, I'm with Gunnar when it comes to music/applause leveling. Do it manually and with consideration to the natural dynamics of the programme. You're not aiming for a pop radio broadcast balance with no dynamic range - let things breathe. BEware of running your levels too high - intersample peaks on pure sounding programme can be a very audible problem; I'd suggest leaving around 1-2 dBFS of headroom on the finished master and, for CD, not worry about 40-50dBFS of dynamic range below that.


Noises
How bad actually is the noise? It's a live concert - noises happen - people cough, chairs creak, busses and motorcycles drive past. Realistic expectations are important on the part of the clients - it'll never sound like a closed session and trying too hard to make it so will only harm the good aspects of what you have without achieving the objective. If there are really bad problems detracting from the performance then I'd aim to reduce/remove the worst noises but never try a complete clean up as it's rarely successful even if you have the right tools - which you don't.

LF rumble can, depending upon the musical/spectral content of the programme, often be reduced enoguh by the use of some gentle LF rolloff. Knowing the sepectral content of the instruments and music recorded will advise an appropriate frequency to try. If you can get hold of one, you could try using a Weiss EQ1-DYN Dynamic EQ to apply gentle, programme dependent EQ, if you can't get one, you could perhaps do something in the automation in your DAW. Aim to do as little as possible and only when and where it's really needed.

My experience of using phase cancellation as noise reduction is that it very rarely achieves anything other than making things worse and I'd never bother. Whilst it can remove some types of static noise, it's useless on any kind of dynamic noise or complex spectral content and it doesn't differentiate between musical content and noise so removing wanted alongside unwanted is a problem.

Impulse/discrete noises are best removed using an appropriate tool such as CEDAR ReTouch (which would be my first choice)/Algorithmix ReNovator, or similar. These can work virtual miracles if used carefully and can achieve things which nothing else can - including seemless removal of things like coughs, door slams, stand kicks, chair creaks, bird twittering, shouts, bangs and bumps. It can also work wonders on reducing things like wailing sirens, burglar alarms/other bells and some traffic noises. I've use both these products, particularly ReTouch with great success over the last six or seven years or more and they still amaze me with what they can do. Removing these noise types with EQ/compression/whatever is usually impossible and very unlikely to give sufficent improvement without causing greater damage than it fixes. If you have a lot of noise problems, prioritse as it's not a quick process but it is effectve.

Ideally, if possible, it might be worth trying to arrange a patching session with the group in the same setup in the same venue. The patches are unlikely to match perfectly but if what you have is really problematically bad, inserting a slightly mismatched cover might be the lesser evil. (You should come clean about editing it on the CD booklet though - as it'll no longer be a 'live' recording. Explaining that it was lightly edited to cover problematic noises in the concert is usually enough to keep buyers happy.)



You must've used a rather large amount of gain if you're hearing noise from a pair of 4006s. Was it a very very quiet source or were you a long way away? Might be worth looking at for future projects. Sometimes it's better to be a bit closer than is ideal, especially on stereo pair live work. You can usually do a bit of remedial work to repair a slightly too close balance that'll yield a better final result than trying to get around a lot of noise from the outside or the audience.

[David Rick]

Originally Posted by pkautzsch

There's a very expensive NR software by Algorithmix, which is said to be very effective even for coughing and door-slamming. No first-hand experience with this though.

The Algorithmix reNOVAtor software that Peter mentions has a stellar reputation, but I can't afford it either. Samplitude/Sequoia's spectral editing function is clearly inspired by reNOVAtor, even if not in the same league. But it's still quite worthwhile for mitigating impulsive noises like coughs and chair squeeks. I don't try to make them disappear entirely (reverberance makes that impossible), but to reduce them, so it doesn't sound like someone is coughing in your ear. My policy is to process only the really distracting noises. If there's some loss of fidelity in a few isolated one-second spots, I'm willing to accept that in the interest of musical continuity.

I agree that conventional noise reduction tools intended for ventilation noise and the like extract too high a sonic cost to be used in classical work. But I'll sometimes use such a tool between movements or before the downbeat.

David L. Rick
Seventh String Recording

[audibell]

Originally Posted by David Rick

The Algorithmix reNOVAtor software that Peter mentions has a stellar reputation, but I can't afford it either. Samplitude/Sequoia's spectral editing function is clearly inspired by reNOVAtor, even if not in the same league. But it's still quite worthwhile for mitigating impulsive noises like coughs and chair squeeks. I don't try to make them disappear entirely (reverberance makes that impossible), but to reduce them, so it doesn't sound like someone is coughing in your ear. My policy is to process only the really distracting noises. If there's some loss of fidelity in a few isolated one-second spots, I'm willing to accept that in the interest of musical continuity.

I agree that conventional noise reduction tools intended for ventilation noise and the like extract too high a sonic cost to be used in classical work. But I'll sometimes use such a tool between movements or before the downbeat.

David L. Rick
Seventh String Recording

I have found that selecting out SOME of the attack of a cough or chair knock on L then R channel yields a better result than trying to paint it all out, channel-linked. I agree that hiss removers should be used sparingly at the beginning and in-between movements and again, they are reducers, not removers.

It's hard work, moving through an hour-long program marking up coughs, squeaks, street-car rumble and candy-wrapper rustling for attenuation.

I can't work miracles with any of these apps but I can shift intrusions "back" so the music stays in front. I don't apply any single process to a whole file, work with 16/44 and never give back any "tinklies", the sound of filters working too hard. Since Katrina many things once taken for granted have vanished and most agreably, one of these is air conditioning during the playing at concerts.
WT

[Remoteness]

For future reference:

I know this doesn't help your noisy 2mix today, but since the venue was a noise situation, I would have close (spot) mic'd the instruments to maximize the music output.

[Roland]

As has been said above, give the compressors and limiters a miss, they will make noise problems worse. Phase cancelling noise can also be a problem. There are great tools out there, like Renovator and Cedar Retouch, these can really make a difference if used by experienced, skilled operators as can some of the other noise reducing plugins (again if used sensitively by an experienced operator). As you haven't posted a sample it is difficult to be more specific with advice. depending on the final use of this recording you will have to make a judgement as to whether it justify's the additional expense of getting it tweaked.

Regards


Roland

[dsomeister]

For your coughs, chair squeeks, etc. Dowload Rx from Izotope. If you like it, it's much cheaper than RenoVator. It is working just fine for several classical cd releases for us!

Roy cherryhomes
Smu Meadows School of Music
Recording Engineer

[Jim vanBergen]

I love Cedar for noise issues....but with classical music, just about anything you do CAN do as much damage as good.

[DrDeltaM]

Reducing noise on critical music like classical music is possible, but it's always a question of how far you want to go.

Reducing hiss (I wouldn't go as far as totally removing it, makes things sound a bit 'dead') and noise such as passing cars, coughs, thumbs (like the lady hitting your stand) can all be done. The hardest part is judging how much you can reduce them without doing more bad then good.

Of course, with powerful tools like Cedar, you'll be able to dig deeper then with less advanced tools.

Maybe I can help you out

[Gaston69]

As promissed,

Here are a few samples from my recording last Friday 11/01/2008

Recorded by using 2x Bruel & Kjaer 4006 A/B approximatly 60-70cm distance, 3-4 meters in front of the stage and about 3 meters high.
Used Cordial Top Performance cable 10m Gold plate Neutrik XLR into AMEK DMCL Pure Path used the Amek's internal converter > Spdif > Mbox2 > Logic 8, edited with Peak > Normalized > Pow'r Type 3 Dither > Itunes > Mp3 > 128 Kbs.

No EQ & No Effects, everything Pure

Capella Aquisgrana live in Vaals

Please your oppinions!!!

Gaston

[Plush]

That complement of instruments is very soft to begin with. Lutes and recorders and such are a challenge. When you apply enough gain from the mic amp to get normal recording levels, you also are bringing up background noises. It's unavoidable.

Other than the electronic noise (hiss) I heard, this is a totally normal type of sound in that venue.

You should not be getting any hiss from your mic amp chain. If you are, something is broken or the way you have set your gain staging is misaligned.

best from Chicago,

PhlushPHonnic

[Gaston69]

Plush,


Well that's what I also thought, I found the amount of his way above then what I have been used to. You can see in my previous thread how it all has been routed, actuall straight forward, I got the his also on my back up DAT.

The Mic Gain was set at 52dB ( Amek DMCL ) this unit is in general very quit with other mics I have. The samples you heard were my first recording with Bruel & Kjaers, but on Thursday I will be recording a choir without audience i.e. will have another change to try them.

Any advice?

Gaston

[DrDeltaM]

Here's a little try-out:

[Gaston69]

Mathijs, first of all thanks.
The samples you've posted are really a big improvement i.e. noise was reduced in a very musical way. Have informed to band about your services.

In a couple of weeks I have more classical recording projects comming up like:

- Chambre Choir (24 musicians)- Jonannes Passion
- Chambre Orchestra about 35 musicians playing Schubert Balletmusik I & II
Schubert Symfonie n°8 "Onvoltooide" D 759
Beethoven Vioolconcerto in D opus 61
- JOHANN SEBASTIAN BACH: SONATAS AND PARTITAS FOR VIOLIN - 1
Xavier Julien Laferrière (baroque violin) and Claudio Barduco Ribeiro (harpsichord)

I will try to pay more attention to the mic-placement, so I get a good stereo panorama while maintaining a acceptable signal to noise ratio.

Will record with same equipment as you've heard before:

2x Bruel & Kjaer DPA 4006
AMEK DMCL 2ch. Class A pre-amp using it's AD converter.

Samples will be posted soon.

Cheers

[DrDeltaM]

Thanks

Good luck with your upcoming recordings

[XLR]

May I ask how You achieved this magic?

[DrDeltaM]

Originally Posted by XLR

May I ask how You achieved this magic?

Dehissing was done with Cedar AutoDehiss, general background public-noise diminishing with Cedar DNS. The micstand-hit was made more subtle with Cedar Dethumb (because the thumb was just together with the attack of a note, I didn't fully delete it to affect the note as little as possible). I also removed some cough with Cedar Retouch (spectral editor).

[Alexey Lukin]

For comparison, here's what I've been able to get with Denoiser and Spectral Repair tools from RX.

[DCtoDaylight]

Originally Posted by Gaston69

In a couple of weeks I have more classical recording projects comming up like:

- Chambre Choir (24 musicians)- Jonannes Passion
- Chambre Orchestra about 35 musicians playing Schubert Balletmusik I & II
Schubert Symfonie n°8 "Onvoltooide" D 759
Beethoven Vioolconcerto in D opus 61
- JOHANN SEBASTIAN BACH: SONATAS AND PARTITAS FOR VIOLIN - 1
Xavier Julien Laferrière (baroque violin) and Claudio Barduco Ribeiro (harpsichord)

I will try to pay more attention to the mic-placement, so I get a good stereo panorama while maintaining a acceptable signal to noise ratio.

Will record with same equipment as you've heard before:

2x Bruel & Kjaer DPA 4006
AMEK DMCL 2ch. Class A pre-amp using it's AD converter.

If you haven't already, you might want to take a look at this Michael Williams paper - I found it super-helpful in understanding stereo mic placement, and he makes it very easy to scope out a given situation.

http://www.microphone-data.com/pdfs/Stereo%20zoom.pdf

Good luck!

[NorseHorse]

Just bumping this thread cause it has some helpful discussion.

[Gaston69]

It has helped me a lot indeed, since then my quality of recordings are improved a lot. Listen on my website the sound samples.

[sonare]

Originally Posted by nickynicknick

Re Levels...if the applause is way too loud....find your loudest peak (music without applause) and make that your high point (-4ish)...then you'll have to mix your tracks down when the applause kicks in so that they don't peak at more than -4....don't use limiting for this type of music.

I have found (the hard way when running WAY late) that if you set the applause peak to -5 or -8 (or so) your levels will be ok UNLESS the audience is very small, you are recording the Berlioz Requiem, or it is an opera chorus.

Rich

[aracu]

Loud coughs, clicks etc. can usually be lightened up with volume envelopes.

[springer]

I record in very noisy environs and over the years have considered this when placing mics above the "correct" stereo placement for "optimum" stereo image. Results? listenable without distraction - and after all isn't that the desired result?
Every once in a while the crowd is quieter and I wish my mics were further back, but this is rare.