Forums

Improving the SNR of a mic

Started by Michel Rouzic February 23, 2006
I use a low end speaker connected to my mother board's AC97-type ADC to
input sounds. The problem is that I have a lot of noise.

How could I improve the SNR and get rid of a big part of the noise?
(I'm not talking here about post-processing but about getting the raw
signal from the ADC to a better SNR). Is my low end speaker the problem
or is it rather my ADC that's introducing the most noise? In other
words do I need a better ADC or a better speaker? Anything about the
two that I can do to make it better?

"Michel Rouzic" <Michel0528@yahoo.fr> wrote in message
news:1140737808.283548.99860@j33g2000cwa.googlegroups.com...
> I use a low end speaker connected to my mother board's AC97-type ADC to > input sounds. The problem is that I have a lot of noise. > > How could I improve the SNR and get rid of a big part of the noise? > (I'm not talking here about post-processing but about getting the raw > signal from the ADC to a better SNR). Is my low end speaker the problem > or is it rather my ADC that's introducing the most noise? In other > words do I need a better ADC or a better speaker? Anything about the > two that I can do to make it better? >
You need to use the full dynamic range of the ADC without clipping of course. Tam
Michel Rouzic wrote:
> I use a low end speaker connected to my mother board's AC97-type ADC to > input sounds. The problem is that I have a lot of noise. > > How could I improve the SNR and get rid of a big part of the noise? > (I'm not talking here about post-processing but about getting the raw > signal from the ADC to a better SNR). Is my low end speaker the problem > or is it rather my ADC that's introducing the most noise? In other > words do I need a better ADC or a better speaker? Anything about the > two that I can do to make it better?
What kind of noise? Hum? The fan in the computer? Low-end speaker or high, the impedance mismatch to the sound input is very high. Try a transformer. And shielding. Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Dougal McDougal of that Elk wrote:
> "Michel Rouzic" <Michel0528@yahoo.fr> wrote in message > news:1140737808.283548.99860@j33g2000cwa.googlegroups.com... > > I use a low end speaker connected to my mother board's AC97-type ADC to > > input sounds. The problem is that I have a lot of noise. > > > > How could I improve the SNR and get rid of a big part of the noise? > > (I'm not talking here about post-processing but about getting the raw > > signal from the ADC to a better SNR). Is my low end speaker the problem > > or is it rather my ADC that's introducing the most noise? In other > > words do I need a better ADC or a better speaker? Anything about the > > two that I can do to make it better? > > > > You need to use the full dynamic range of the ADC without clipping of > course.
umm yeah, good point, but without an amplifier I don't really have the choice.
Jerry Avins wrote:
> Michel Rouzic wrote: > > I use a low end speaker connected to my mother board's AC97-type ADC to > > input sounds. The problem is that I have a lot of noise. > > > > How could I improve the SNR and get rid of a big part of the noise? > > (I'm not talking here about post-processing but about getting the raw > > signal from the ADC to a better SNR). Is my low end speaker the problem > > or is it rather my ADC that's introducing the most noise? In other > > words do I need a better ADC or a better speaker? Anything about the > > two that I can do to make it better? > > What kind of noise? Hum? The fan in the computer? Low-end speaker or > high, the impedance mismatch to the sound input is very high. Try a > transformer. And shielding.
The noise is originating somewhere between the speaker and the ADC. It seems to follow a pattern that is about 1.05 seconds wide. I'm quite ignorant when it comes to electronics, what is a transformer (is it those long wires rolled around metalic bars?) and shielding?

Michel Rouzic wrote:
> I use a low end speaker connected to my mother board's AC97-type ADC to > input sounds. The problem is that I have a lot of noise. > > How could I improve the SNR and get rid of a big part of the noise? > (I'm not talking here about post-processing but about getting the raw > signal from the ADC to a better SNR). Is my low end speaker the problem > or is it rather my ADC that's introducing the most noise? In other > words do I need a better ADC or a better speaker? Anything about the > two that I can do to make it better?
Look at the Tascam US-122. Amazing noise performance for the price (<$200). I've measured Ein of about -127 dBu. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein
"Michel Rouzic" <Michel0528@yahoo.fr> wrote in message
news:1140758561.557678.170880@v46g2000cwv.googlegroups.com...
> > Dougal McDougal of that Elk wrote: > > "Michel Rouzic" <Michel0528@yahoo.fr> wrote in message > > news:1140737808.283548.99860@j33g2000cwa.googlegroups.com... > > > I use a low end speaker connected to my mother board's AC97-type ADC
to
> > > input sounds. The problem is that I have a lot of noise. > > > > > > How could I improve the SNR and get rid of a big part of the noise? > > > (I'm not talking here about post-processing but about getting the raw > > > signal from the ADC to a better SNR). Is my low end speaker the
problem
> > > or is it rather my ADC that's introducing the most noise? In other > > > words do I need a better ADC or a better speaker? Anything about the > > > two that I can do to make it better? > > > > > > > You need to use the full dynamic range of the ADC without clipping of > > course. > > umm yeah, good point, but without an amplifier I don't really have the > choice.
You always need a pre-amp!
>
Bob Cain wrote:
> Michel Rouzic wrote: > > I use a low end speaker connected to my mother board's AC97-type ADC to > > input sounds. The problem is that I have a lot of noise. > > > > How could I improve the SNR and get rid of a big part of the noise? > > (I'm not talking here about post-processing but about getting the raw > > signal from the ADC to a better SNR). Is my low end speaker the problem > > or is it rather my ADC that's introducing the most noise? In other > > words do I need a better ADC or a better speaker? Anything about the > > two that I can do to make it better? > > Look at the Tascam US-122. Amazing noise performance for the price > (<$200). I've measured Ein of about -127 dBu.
Maybe it's a slightly too big investement. Plus inputting a signal from a extra-low end speaker with such a device seems to be pretty much like putting the engines of a Ferrari in an old shitty Fiat. So far my plan is to get the best of what I already own.
The noise waveform you describe (1.05uS pulse) sounds like digital
interference, which would probably be there regardless of choice of
transducer.

The last time I tried a (el-cheapo) speaker as a mike it was good for
about 20 millivolts p-p.  Your sound card ADC input probably wants
about 2 volts peak to peak.  The 100:1 voltage gain required (40 dB)
which you are presumably doing in software will bring up the level of
the noise by that amount.

Jerry Avins' comment about a transformer is the simplest way to fix
this.  Get a speaker transformer of reasonable quality (minimum size of
1 cubic inch - you'd be surprised at how many things are better when
they are heavier).

The primary winding needs to be rated as having a primary impedance of 5
Kohms or greater, and the secondary at 8 ohms or less.  Now connect the
speaker (mike) to the 8ohm secondary.  Connect the 5K winding to the
ADC input & signal ground.

The transformer is now running "back to front" and giving you a step-up
voltage multiplication of 25 to 1.  This is about 28 dB gain. which
should allow you to wind down your software gain by that amount.

Jim A.

Jim Adamthwaite wrote:
> The noise waveform you describe (1.05uS pulse) sounds like digital > interference, which would probably be there regardless of choice of > transducer.
I mesured on a sample of ten to be about 1.05 second long, but I noticed on other parts of a recording that it can be shorter than that.
> The last time I tried a (el-cheapo) speaker as a mike it was good for > about 20 millivolts p-p. Your sound card ADC input probably wants > about 2 volts peak to peak. The 100:1 voltage gain required (40 dB) > which you are presumably doing in software will bring up the level of > the noise by that amount. > > Jerry Avins' comment about a transformer is the simplest way to fix > this. Get a speaker transformer of reasonable quality (minimum size of > 1 cubic inch - you'd be surprised at how many things are better when > they are heavier). > > The primary winding needs to be rated as having a primary impedance of 5 > Kohms or greater, and the secondary at 8 ohms or less. Now connect the > speaker (mike) to the 8ohm secondary. Connect the 5K winding to the > ADC input & signal ground.
ok thanks alot for the tip. i got alot of transformers hangin around so i'll test so of them.
> The transformer is now running "back to front" and giving you a step-up > voltage multiplication of 25 to 1. This is about 28 dB gain. which > should allow you to wind down your software gain by that amount.
Just wondering, how do you calculate the voltage multiplication out of the impedance of each part? Anyways I guess a 230 to 12 volt (or less) transformer will do it, right?