Actually 16dBa isn't that low anymore.

It is a bit of a bear to measure. The preamp is probably quiet enough, be sure your phantom supply is quiet too, that can influence things when you are looking at this low a level.

Lowest I've seen have been in the 7 to 9dBa range, A weighted.

You've got the gist of it anyway.

Terry,

I've done that. (The first part, for measurement equipment and microphones).
Can't you find somewhere a school or college with an anechoic room in your surroundings?

Or someone with a high insulated control room.

Maybe for a small cost they allow you to use it for a while?
Or by using charm or relationships even for free?

This small box is as well complicated (frequency dependency).

Dan, the lowest self noice Mic I know is a B&K omnidirectional 1" measurement mic with a self-noise of -2dB (indeed minus) including preamp.

Yeah, the B & K's are in a class by themselves. I haven't looked at their measurement stuff in quite awhile, I was referring mostly to the commercially available studio/PA mics I've seen.

The omni mics are interesting in that the capsules themselves are inherently quieter than a directional mic, then you have the benefit of the much higher output voltage from the large diaphram and very high polarizing voltage, which in most B & K's is about 200 (or used to be) which is twice the usual in studio mics. Then add some very tasty low noise electronics, and there you go.

What do they cost tho?

$1500 for a capsule, $1000 for a mic body, $1500 for a supply/preamp plus any screwy B & K connectors.

Just a guess on the costs, like I said, it's probably been 20 years...

Dan,

Yes, I know it's an exception.
I can't remember the price, I know together with the preamp was a lot.

How can a microphone have negative self-noise? Does that mean that, by repeatedly adding the microphone in front of a noisy signal, you can clean it up?

Eric said: Terry,
...
>>Can't you find somewhere a school or college with an anechoic room in your surroundings?

Apparently not, here in Tunisia.

>>This small box is as well complicated (frequency dependency).

Eric, if I do not need to have ANY sound sources (if I am not measuring frequency response, etc) and all I need is a very acoustically quiet place to measure microphone self-noise, would not a box work OK??

>>Dan, the lowest self noice Mic I know is a B&K omnidirectional 1" measurement mic with a self-noise of -2dB (indeed minus) including preamp

RE: Negative number here: This is relative to the human threshold of hearing, so negative numbers are appropriate,

patrick_like_static wrote: How can a microphone have negative self-noise? Does that mean that, by repeatedly adding the microphone in front of a noisy signal, you can clean it up?

Patrick,

A negative dB value is STILL a positive arithmetic factor.

Expressing zero or negative in the arithmetic sense is mathematical impossible in dB's (logarithmic scale).

0 (zero) dB does not mean nothing but the factor 1.
Any arithmetic factor >0 and
As such one can see arithmetic factors > 0 and
If you see -30 dB on your gear it means a factor 10^(-30/10) = 0.001 or 0.1% of your max. output which is 0 dB equaling the factor 1 or 100%.
And this factor 1 comes from 10^(0/10) = 10^0 = 1 = 100%

As such -50 dB is still noise which is present.
a dB value is a dimensionless factor versus a reference unit.

You can't express in dB's that there is no sound. Try in Excel or your calculator: 10*log(0). You'll see you get an error.

Edited: Sorry Terry read it to late:
And as Terry said: 0 dB meaning a factor 1 is set to the treshold of hearing (at ca 1000 Hz, but should check it).

TerryKing wrote:

Eric, if I do not need to have ANY sound sources (if I am not measuring frequency response, etc) and all I need is a very acoustically quiet place to measure microphone self-noise, would not a box work OK??

You are measuring to extreme low levels in the whole spectrum also in the lower audio frequencies.
A small box will also show a modal behavior (assymetric or not).
So you will have somewhere a crossover between a plain pressure field in that box (where wavelengths are too long for the box measures) and a soundfield.

With absorption it's very difficult to get to those low frequencies.

I don't say it isn't worth testing, but I shouldn't be sure until testing such a box in itself, to know what I'm measuring.
If you only want total dB(A), it diminishes the problem somewhat (more exclusion low frequencies).

In fact to measure those very low levels, one needs MLS techniques or work with statistical analysis (multi-spectra).
Not much must happen to boost a 0 to 5 dB background level per band. Such measurement results are very fast poluted by disturbing events.

Now I agree it's worth testing, I only should feel safer in a known environment if you should have access to it (and maybe cheaper than making and testing a good box).

In order to find the self noise you should create something that best is at least 8 dB lower than the self-noise per band, and for the most difficult bands the absolute minimum the backgound noise in the box should be (per frequency band) = 3 dB lower than the self noise, otherwise you get a summation, not knowing which is which.

Just some thoughts.

Thanks very much, Eric and Terry; did I understand correctly that a mic with -7dB of self-noise is noisier than a mic with -9 dB, but a mic with -7 dB is less noisy than a mic with a +2 dB? If I'm way wrong here, perhaps giving me a handy link pertaining to said topic would be easier for you than having to reiterate.

Also, If 0 dB is anything below the threshold of human hearing, does the curve of our perception start at a flat 1 dB, or at every increment above 0 dB (e.g. 0.00001 dB)? If the former is true, why are such marginal numbers important in signal-to-noise?

Patric,

patrick_like_static wrote: did I understand correctly that a mic with -7dB of self-noise is noisier than a mic with -9 dB, but a mic with -7 dB is less noisy than a mic with a +2 dB?

Perfectly correct.
:P Only I doubt that the -9 and -7 dB mics exist but let's forget that here.

patrick_like_static wrote:
Also, If 0 dB is anything below the threshold of human hearing, does the curve of our perception start at a flat 1 dB, or at every increment above 0 dB (e.g. 0.00001 dB)? If the former is true, why are such marginal numbers important in signal-to-noise?

1) loudness is frequency dependent, individual and subjective. There are the Fletcher loudness curves, substituted by newer ones, and numerous others, for broadband and pure tones. So don't see such numbers as fixed axioms. They are based on statistical data, valid for certain boundary conditions. Lots is written already about that, and recently I heard that ASTM is busy to revise things.

2) The treshold is at 0 dB not 1 dB (remember 0 dB is not nothing but 1 time the reference unit)

3) The importance is relative. Loudness is also related with the masking background noise. There is little chance you ever will notice a 15 dB noise in normal circumstances, while in anechoic rooms, you can hear your hart beat. You don't need to feel your pols to count the rate. The same 15 dB tone is audible there (if you have good ears and no tinitus).

4) In order to use the maximum dynamics of nowadays digital possibilities it can be usefull that the dynamic range is maximized.
Having 95 to whatever signal to noise ratio in a room with 40 dB background noise has little sense. The isolation must be good but also the self-noise of the mic and whatever.

However to measure this self-noise the background noise must be lower otherwise background and self-noise adds.
0 dB + 0 dB = 3 dB.
If you have a mic with say 15 to 16 dB(A) self noise, seen per frequency band those levels are lower (they are 16 dB(A) all together).
As such one must make sure that the possible pollution by background noise or events will not influence the measurement. Practically this means that the background noise must be ca 8 to 10 dB lower, to make sure one really measures the self-noise. One can go to 3 dB lower but must be sure then about the stability of the measurement (at those levels not so evident).
From the measurement results one must subtract the background noise. When one is sure that the background noise is 8 to 10 dB lower one can about ignore this subtraction. However if not, one MUST subtract the background noise from the measurements.
In order to be able to do that, one must have other measurement equipment capable of measuring this very low level background noise.
Not many do have equipment capable of doing that

To get the feel you should check logarithmic calculations etc.

This is all a bit fast written (please don't mind my English or spelling, I'm Dutch).

Those very low levels are extremely rare in real live.

Eric:

Wow. It took me quite a few reads, but I gathered the gist of that post. Thanks for being patient with me. I'm appreciative of members like you on this 'board.