Become a Site Supporter and Never see Ads again!

Author Topic: Best way to remove excessive background hiss from otherwise good recording?  (Read 7027 times)

0 Members and 1 Guest are viewing this topic.

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
I expected an artist to play a mostly rock set and he ended playing a mostly jazz set, so I had the input levels on my PCM-M10 set too low.  As a result, an otherwise excellent recording is marred by excessive hiss, especially during the between song speeches to the audience and during quiet passages of slow songs.  The recording was made at 48k/24 bit. 

Can someone suggest the best way to remove background hiss?  I usually use Audition 3.0.  In the past, when I’ve attempted to use Audition’s de-hiss feature, the software removed the hiss, but left behind what I can only describe as a “burbling” and “gurgling” sound that was even worse than the hiss.  I believe this is referred to as “artifacting.”  As a result, I have never bothered using hiss removal until now. 

Should I be using different software, or am I just using the wrong settings?  Any suggestions would be greatly appreciated. 

Offline yousef

  • Trade Count: (2)
  • Taperssection All-Star
  • ****
  • Posts: 1411
  • Gender: Male
If this was me, I think I would first try some fairly focused (but light) EQ and then apply some *very* conservative noise/hiss reduction - probably not much more than 3dBs of reduction, if it allows you to specify.

I find that the ear can be very forgiving of hiss over the course of an entire recording. Less so of the artefacts that occur at the more brutal end of hiss-reduction.
music>other stuff>ears
my recordings: http://db.etree.org/yousef
http://www.manchestertaper.co.uk
twitter: @manchestertaper

Offline ScoobieKW

  • Trade Count: (3)
  • Taperssection All-Star
  • ****
  • Posts: 1667
    • ScoobieSnax Audio Archive
I've found that you can perceptually clean it up by dealing with the between song stuff (even volume envelope) and leave the music as is. My brother in law made a mix of live recordings of his bands for my wife years ago. I didn't touch the music at all, and his response when he heard my CD transfers, was how well I did the Noise Reduction.
Busman BSC1, AT853 (O,C),KAM i2 Chuck Mod (C), Nak 300 (C),
M10, UA-5, US-1800, Presonus Firepod

http://kennedy-williams.net/scoobiesnax/

Offline TSNéa

  • Trade Count: (6)
  • Taperssection Regular
  • **
  • Posts: 155
Hello jj69,

You could try MAGIX Audio Cleaning Lab:
http://www.magix.com/us/audio-cleaning-lab/

I think MAGIX Audio Cleaning Lab is probably the same than Magic Audio Cleanic MX that I used sometimes for conferences, meetings, a show for children with tales and songs. Not really music, yet. You have a choice between presets for six different types of noise or enter your own parameters.
They offer a 30 day test version here :
http://www.magix.com/us/free-download/

It's made by the developpers of Samplitude, an (object oriented?) audio editing sofware that seems to be greatly appreciated in documentary and radio work. I got once a free LE with a cheap audio interface: very attractive. The price of the Samplitude full version is pretty high...

Hope it helps.

Offline bryonsos

  • Omni addict
  • Trade Count: (28)
  • Needs to get out more...
  • *****
  • Posts: 2062
  • Gender: Male
  • If it's important, tell me to write it down.
    • LMA uploads
Obvious question, but... Did you normalize the recording?
Mics: 3 Zigma Chi HA-FX (COL-251, c, h, o-d, o-f) / Avenson STO-2 / Countryman B3s
Pres: CA-Ugly / Naiant Tinyhead / SD MixPre
Decks: Roland R-44 / Sony PCM-M10
GAKables
Dead Muppets

My recordings LMA / BT / TTD

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
Obvious question, but... Did you normalize the recording?

Yes, of course. 

Offline it-goes-to-eleven

  • Trade Count: (58)
  • Needs to get out more...
  • *****
  • Posts: 6702
Maybe you could use parallel compression techniques to boost the transients, while leaving the hissy bits behind.

New product - Get your NEUMANN active mounting bars
Active microphone mounting bars for Schoeps and AKG, and Naiant AKG, ready to ship

Offline anr

  • Trade Count: (1)
  • Taperssection Member
  • ***
  • Posts: 363
You shouldn't necessarily blame the record levels on the M10.  Other factors can produce what you perceive as "hiss".  It may simply be the natural room noise, which is more broadband.  If you attack that with hiss reduction plug-ins you'll take out a lot of low/mid range music, as well as top end "hiss".  Study the file carefully to find the quietest part and look at it with a spectrum analyser.  Try what yousef said - a little EQ.  Quite often, running a mains hum filter programme (50/60Hz comb filter) gets rid of a lot of room "noise". 

When I really feel it is necessary, I use Sony Noise Reduction 2B.  As stated above, go carefully in very small steps using the above quietest part as a reference.  It needs at least 500ms to work with. 

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
I'm almost positive it's hiss.  The levels were set really low.  I was not using a pre, and I typically set the input knob on the Sony to 5, which I've been told is unity gain.  However, I typically record hard rock/metal shows in very loud rooms. 

I appreciate all the advice, but much of the terminology is "over my head," so I'm not quite sure how to apply it. 

Would it help if I post a sample of the audio?  If so, should I post some between song chatter or some music?  How much?  Is a few minutes enough? 

Offline anr

  • Trade Count: (1)
  • Taperssection Member
  • ***
  • Posts: 363
jj

By all means post a sample.  Try to include a section of quietness between songs as well as music. 

Offline Ekib

  • Trade Count: (1)
  • Taperssection Member
  • ***
  • Posts: 258
I use Nero Wave Editor . It's so easy to work with .
But I have to say, I don’t mind it. I do object when I see people sticking microphones up my nose, in the front row. If I see anyone doing that [laughs] I’m going to have security remove them. Because that’s just obnoxious. But I don’t mind if people come and discreetly at the back make a recording of it. And I know that it’s just for their own use, for the superfan.
(Steven Wilson , interview http://blog.musoscribe.com/index.php/2011/01/25/interview-steven-wilson-on-audience-taping/ )

Offline BlindGuyEars

  • Trade Count: (5)
  • Taperssection Regular
  • **
  • Posts: 151
Izotope RX 2 if you can afford it.
But, echoing what everyone else said. use a tiny bit of hiss reduction and some very careful and conservative EQ.

I'm not sure how parallel compression would help. What about downward expansion though?

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
Okay.  Here are a couple of samples from the recording, both in FLAC format.  These are recorded in 24 bit at 48kHz.  The only thing done to the recording at this point is that it was normalized. 

"Sample 1" is  a one minute sample.  The first 30 seconds or so is between-song banter and the remainder is beginning of the old standard, "Beyond The Sea."  You'll notice that the hiss is very distracting during the banter, and much less so during the song. 

Sample 1:
https://rapidshare.com/files/3167617211/Sample1.flac

"Sample 2" is just 10 seconds of near-silence, ending with an announcer beginning to introduce the band.   The two seconds before the announcer begins speaking is probably the best place to take a sample of the background noise.  Prior to that you can hear the house music played through the PA system. 

Sample 2:
https://rapidshare.com/files/4280393094/Sample2.flac

Most of the suggestions people have made in this thread are a bit over my head, so if someone could give me a more simplified step-by-step approach, that would be greatly appreciated. 

Offline anr

  • Trade Count: (1)
  • Taperssection Member
  • ***
  • Posts: 363
JJ

http://www.sendspace.com/file/fcwsgx


The background noise you hear isn't "hiss" in the accepted sense of the word.  In fact, during the gaps between speech in the spoken intro there is very little above 2kHz.  It is what I described (above) as room noise.  Sometimes you just get rooms that are noisy when empty (and there is a formal method of measuring and classifying this).  Things like whether the room walls are external or load bearing effect this.  All I did to your sample is run a 60Hz hum filter, which made an appreciable difference, and then a very little amount of noise reduction using Sony NR2b. 

The recording is very good indeed.  The problem, if indeed there is one, is that the room is a little unsuited to that type of music.  Fortunately, artists and promoters aren't too anal about such things, otherwise there'd be nowhere for them to play!  I'd say there is very little you could have done to "improve" the recording at the time. 

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
anr,

Thanks for the info.  I can't listen to your sample until I get home from the office tonight, but I can't wait to hear it. 

I am a bit skeptical that there would not have been much, much less of the background noise - whatever it is - had I set my input levels much higher (or used a pre).  If it would be helpful, perhaps I'll post an un-normalized sample of the same audio tonight.  The normalization process added at least 20db to the signal. 

Do you know how I can duplicate the process you used in Audition 3.0?  Does Audition have a similar hum filter? 

I'm not familiar with the Sony software you mention, but could someone recommend NR settings for Audition.  In fact, Audition has two NR filers, a "hiss" filter and a "noise" filter.  I'm not sure what the difference is, or which one is best for this application. 



The background noise you hear isn't "hiss" in the accepted sense of the word.  In fact, during the gaps between speech in the spoken intro there is very little above 2kHz.  It is what I described (above) as room noise.  Sometimes you just get rooms that are noisy when empty (and there is a formal method of measuring and classifying this).  Things like whether the room walls are external or load bearing effect this.  All I did to your sample is run a 60Hz hum filter, which made an appreciable difference, and then a very little amount of noise reduction using Sony NR2b. 

The recording is very good indeed.  The problem, if indeed there is one, is that the room is a little unsuited to that type of music.  Fortunately, artists and promoters aren't too anal about such things, otherwise there'd be nowhere for them to play!  I'd say there is very little you could have done to "improve" the recording at the time.

Offline anr

  • Trade Count: (1)
  • Taperssection Member
  • ***
  • Posts: 363
JJ

I'm not familiar with Audition, sorry.  But I assume it runs common plug ins.

Don't get fixated on "noise reduction".  As you said earlier, it can introduce nasty artifacts and ruin the recording.  Better to have a little background noise.  I only used it to illustrate what could be done to your sample, but in practice I'd probably have EQ'd the bottom end.  If you look at the spectral analysis of a "silent" part, the bottom end looks like a "shelf" with very few peaks.  For example, in Sony Paragraphic Eq you can create a Low Shelf, select a frequency range and how much gain or attenuation you want; this would reduce the level of just that annoying part of the sound.    It's not quite as simple as this, but less is best as it causes less harm.  It is difficult to offer a precise answer as you need to familiarise yourself with the necessary plug-ins and adjust it so it suits YOU, on your playback system.   As ever, always create a back up file of the original.   

If you google DX or VST plug-ins, you will find a raft of freebies.  Run the files and the next time Audition loads, it should recognise them and make them available to you.     

 

Offline Ben Turnbull

  • Don't ask me about Audacity...
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 8996
  • Gender: Male
  • A Clark Nova ate my lunch... Don't ask, just pee
JJ

Well, since my apartment complex's wifi lounge has banned rapidshare as a P2P site  ::) I'm going to comment on my experience with broadband background noise/hiss and hope that it applies somewhat to your problem.

First, you say that you didn't use a preamp so I'm assuming that you used the internal mics on the M10.  They're not bad but they may not be the cleanest sounding when the noise floor is raised.  You said you normalized.  (God I hate normalization.)  and that you hit it pretty hard... 20dB   :o
I haven't done any M10 testing or read any specs so this is just speculation, but if the mics have a fair amount of self noise and you raise it too much (20 dB is well...) and then listen to the product, what you are likely hearing isn't room related hiss but a function of your post process.  Just because one records at 24bits it doesn't mean that it's possible to blow those suckers up without consequences.  I fear you've exceeded the ability of even 24bit density and 48hz sampling rate to save a recording originally laid down at too low a level.  JC could do wonders with a few loaves of bread and a couple fish... I fear you may have needed a lot more fish to start with before you even got started.

On the other hand...

Most rooms that have air conditioning will have an ambient hiss associated with the air moving through the ducts and exiting the vents.  Some are designed to be pretty quiet, some not.  If you find that your location was near such a vent you might be able to find a place on the recording where the room chatter is nill and all you hear is the "air".  Some PA systems may also have a resident sssssssisssss to them when not getting a full signal so that is also perceived as hiss.  Again the a fore mentioned plug ins that have been mentioned will serve to mitigate that, but as you found out, trying to remove it completely has serious consequences.  Sample that noise and start to gently experiment with the noise reduction tools available.  When I've had to do this, I've applied it to minimize not eliminate the noise and have done so only on the quieter passages where louder sounds can't mask the  noise to start with.  I've used Audition's Noise reduction, Hiss reduction (two different tools) and iZatope RX and find that RX is the best for my needs.  Lots of RX vids on You Tube that, even if you don't use it, you should find instructional.

Last thing... and this happened to me last week.   ::)  If you are running the internal mics on the M10, where was the mic sensitivity switch set?  That could make a difference as well.

Good luck
Only seeing two shows a month now...
-My name is Ben, and I'm an AQhaulic ranting hysterically in my dotage ...
Read "Toward the End of Time" by Updike to know my name sake.
Official Archivist for Jazz Central Studios, MinniMN
DPA4021/SKM140/NakCM1k/ADK A51s/Countryman B3/ATes943/SPc4(och)/S'lux s502... >
TinyBox OT> PCM-M10/DR-680/R4/T+mod UA5 needs attention

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
Ben,

Thanks for the lengthy reply.  Just to clarify, I was definitely not using the internal mics on the Sony.  I was using AT943s built by Sound Professionals (SP-CMC-8 cards), along with the Sound Professionals 12V battery box plugged into the Line in on the Sony.  The mics were built with the SP "low sensitivity" mod because I typically record VERY loud concerts. 

What happened here is that I was expecting to get a loud rock set from this band, and they did a mostly jazz set instead.  As a result, I had the Sony's input level set much too low.  I'm used to recording metal shows where I just set the Sony at about "5" on the input knob and forget about it.  If I have to normalize a metal show by 10db-20db it's never a big deal. 

I'm sure the normalization process (which may have been closer to 30db) is not helping here, but without normalization, the music would be inaudible. 

I really don't think the noise is normal, or had anything to do with the room.  The room was in the middle of a casino, but designed specifically for music and typically has very good acoustics.  The ceiling is VERY high, if that matters.  The two PA monitors hang about 20' about the stage. 

Offline Ben Turnbull

  • Don't ask me about Audacity...
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 8996
  • Gender: Male
  • A Clark Nova ate my lunch... Don't ask, just pee
OK, so with the exception of not having a pre in the chain you're running my rig.  Works well.  Sounds like you just ran out of usable bits to stretch while normallizing.  Can't make bits where there aren't any to start with, sad to say.  This is how we learn... or at least how I've learned! 
Only seeing two shows a month now...
-My name is Ben, and I'm an AQhaulic ranting hysterically in my dotage ...
Read "Toward the End of Time" by Updike to know my name sake.
Official Archivist for Jazz Central Studios, MinniMN
DPA4021/SKM140/NakCM1k/ADK A51s/Countryman B3/ATes943/SPc4(och)/S'lux s502... >
TinyBox OT> PCM-M10/DR-680/R4/T+mod UA5 needs attention

Offline vanark

  • TDS
  • Trade Count: (29)
  • Needs to get out more...
  • *****
  • Posts: 7262
  • If you ain't right, you better get right!
    • The Mudboy Grotto - North Mississippi Allstar fan site
In the past, I had a fair bit of success using the Noise Removal tool in Audacity.  I know everyone is giving you good advice, but if you can't follow it, you might give the Noise Removal tool a try to see if you like the results.  I've used it in the exact situation you describe and been satisfied with the results.  The key is to have a sample of noise that is nothing but noise.  The sample you provided has other sounds in it (house music) and really can't be used well to create a noise profile.  I know everyone has been telling you to use EQ and compression and whatnot.  That will work if you have the time and patience.  If you don't, you might want to try something else.  Good luck and YMMV and all that.
If you have a problem relating to the Live Music Archive (http://www.archive.org/details/etree) please send an e-mail to us admins at etree(AT)archive(DOT)org or post in the LMA thread here and we'll get on it.

Link to LMA Recordings

Link to Team Dirty South Recordings on the LMA

Mics: Microtech Gefell M21 (with Nbob actives) | Church Audio CA-11 (cards) (with CA UBB)
Pres: <empty>
Recorders: Tascam DR-60D | Sony PCM-M10

Offline jj69

  • Trade Count: (3)
  • Taperssection Regular
  • **
  • Posts: 165
Just wanted to post a follow-up to my own thread.  I tried some of the suggestions you've all made - at least the ones I could understand and figure out how to apply in Adobe Audition.  After messing with AA's features and coming up with absolutely nothing in the least bit satisfying, I decided to try different software.  I just got iZotope RX 2, and I think I've found my solution! 

Simply put, the "Denoise" feature in RX is astounding.  Keep in mind that I'm familiar only with the Noise/Hiss Reduction in AA, and to a lesser extent in SoundForge (and that was years ago), so that's my basis for comparison.  Believe me, there is absolutely no comparison between RX and the other two programs.  Even using the stock settings, applying 10db of Dehiss in RX left almost no noticeable artifacts.  If I tried that in AA, the result would be un-listenable. 

What I've decided to do with my recording is use about 3.0 db of RX's Dehiss over the entire recording.  Then I'll go back and add another 10.0-12.0 db of Dehiss to the soft passages (mostly the between song comment).  I'm going to use the highest quality settings available in RX.  For those familiar with RX, that means "Smoothing" set at 10.0 and using Algorithm D.  At those settings, processing takes a long time (almost real-time on my system), but I'm hoping the result is worth it. 

I'll report back on how it comes out if anyone is interested. 

 

RSS | Mobile
Page created in 0.355 seconds with 48 queries.
© 2002-2019 Taperssection.com
Powered by SMF