naturerecordists
[Top] [All Lists]

Mic Self noise and sound filtering (was Digest Number 1557)

Subject: Mic Self noise and sound filtering (was Digest Number 1557)
From: Walter Knapp <>
Date: Sat, 27 Nov 2004 16:48:08 -0500
From: Terry King <>

> 
> At 12:46 PM 11/24/2004 +0000, Walt wrote:
> 
>>>> >The noise reduction software can be used for fairly minor noise
>>
>>>reduction. It will not come at all close to a good recording made with a
>>>good low noise mic. Noise reduction software introduces it's own noise
>>>too. It can only be applied very lightly.
>>>
>>
>>>> >Go for as good a mic as you can afford, get close enough that you don't
>>
>>>have to use a lot of gain, and work on technique. You make the recording
>>>by the time it's fed to the recorder. The rest is minor changes.
> 
> 
> Very good points!
> 
> Walt, I'm trying to understand the noise that is output by a typical good 
> condenser mike: that is it's "self noise".  Is this wideband uncolored 
> (white) noise, typically?  I'm thinking about designing a preamp, and 
> wondering if I want to be able to do any frequency-domain adjustments or 
> cutoffs, or none.  I have a couple of Shure WL183's that have a rated 
> self-noise of 22.5 dB SPL.  Am I correct to assume that a desired signal 
> that is 32.5 dB SPL would give a SNR of 10 dB?  Do you have any recordings 
> of a microphones self-noise in an acoustically quiet environment? I'd be 
> interested in doing a spectral analysis on the noise.

Mic self noise is not generic in it's character. Each mic will have it's 
own self noise profile. Some may have a nice smooth hiss (MKH mics, for 
example, though there are others). Others it may be more static like, 
crackles, sputters, etc., obviously much more of a problem as it does 
not hide well.

No single filter will adjust for all mic self noise. Certainly no preamp 
where you really want clean gain across the board. And since the mic 
self noise generally wipes out most everything else below it's level, 
you have little to recover. It's the floor on what's possible.

I have recorded in lots of quiet environments. Some even quiet enough to 
challenge a MKH mic's self noise. Though generally something else is the 
limit.

In general, for quiet environments you want a mic with well below 20 dBA 
self noise figures, the lower the better. Some examples I use, all 
Sennheiser:
Sennheiser MKH-20  -  10 dBA  Omni I use this in the mod SASS
Sennheiser MKH-30  -  13 dBA  Figure 8 the side mic for M/S
Sennheiser MKH-40  -  12 dBA  Cardioid  I use this M/S
Sennheiser MKH-60  -   6 dBA  Short Shotgun also M/S in my kit
Sennheiser MKH-70  -   5 dBA  Long Shotgun Not yet in steady use, M/S?
Sennheiser MKH-80  -  10 dBA  Multi pattern, I set up a pair as M/S

To relate self noise to ambient signal levels you also have to take into 
consideration the mic's sensitivity. Self noise is tested relative to a 
94 dBA signal at full scale. Recording quiet ambiance you will probably 
have more gain as the wanted sound is lower level. Self noise is a 
number you use (carefully) to compare mics you are considering buying. 
It's a important number for nature recording, or anytime you are 
recording in quiet environments as it provides a floor to what you can 
record.

Note it's somewhat risky comparing self noise figures between different 
brands, but we are all forced to take this risk.

> 
> On another related subject:  Some online sample soundscapes have known 
> human-caused backgrounds such as distant power plants and trucks on a 
> distant highway.  Has anyone tried the sampled-noise reduction techniques 
> on these type of noise sources??

Ah, most of what you use filtering for! Sampled-noise reduction is only 
one technique you throw at this. And kind of minor at that. You throw 
everything but the kitchen sink at it. You'd throw the kitchen sink if 
it did any good! And learn that nothing substitutes for not recording it 
in the first place. Filtering is a patch effort after the fact. Everyone 
does it, and wishes they did not need it.

I use several other techniques, primarily frequency dependent filtering 
of various types first and follow by a sampled-noise reduction applied 
lightly at the end of the filter stack. Most of the time no more than 6 
to 10 dB reduction. That followed by a realtime sonogram display to 
check what I'm doing. I may also use dynamic range filtering (which is 
in different software so can't be added to the stack, and is really 
tricky to use). Checking what you are doing with a sonogram is important 
to help avoid filtering to your ears and bias only.

> Any suggestions, pointers appreciated.

A lot of this is ground covered numerous times in the group, but you 
can't cover it too much. Reading through old posts could help. Subjects 
like this tend to develop info through back and forth discussion as 
specifics come out.

Try playing around with filters on your own recordings. That's really 
the best way to learn what's really possible. In reality much less than 
theory seems to predict. Grind up samples with all kinds of different 
filters over and over on the same samples different ways and you will 
develop your own methods that work for how you record. I spent several 
months just playing before even attempting filtering on anything 
critical. And was still going back and redoing things a lot after that, 
still do. It's almost it's own art form.

Jump in with both feet, and when you get stumped on specifics we will 
try and haul you out of whatever mired you down if we can. That's the 
way to go about it, more or less. It's very much learn as you go.

Walt





________________________________________________________________________
________________________________________________________________________

<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU