How to limit level under special decibel level

Audacity 2.3.2, Win 10 64 bit.

This is quite difficult explain, but:

I use special audio recorder circuit, Aplus Aivr. It is very simple OTP circuit with D/A converter. Simply you can record any audio file inside it.

Reason circuit is very limited, signal need edit. Eg. under 100 Hz there is no any possibilities. Also over 16 k. There is also many limits, but need jumping with signal it hear … enough ok.

Ok. One limit is, “if signal level is too low, it make strong noise to output”. Signal level 0,0 is ok, but “more than 0,0, eg. 0,05 is poor”.

Ok, ok. “0,1” of course contain important information, but I want reduce it. For test.

Easier to explain if use picture:

This higlighted section is speech, yes. Anyway I want reduce it— test purpose. Yes yes, if I remove it, I hear speech is not good and end of words disappear and cow fly and and

So: How I can make filter: “If signal level is under <adjustabledb, eg. 0,05, 0,01 etc> AND longer than REPLACE IT with silence”?

That’s called a “noise gate” (or sometimes just “gate”).
There’s a noise gate plug-in available here:
This is a “Nyquist Plug-in”.

Installation instructions for Nyquist plug-ins in Audacity 2.3.2 on Windows:

There is a Noise Gate plugin.

Removing background noise or noise between words is very difficult. It’s normal for the effect to sound mechanical or distorted.

Much better to record a clean, undistorted voice than try to fix the distortion later.


Thank you, this is right plugin and works.

Problem is not poor record. Problem is Voice circuit. It need some special jinx.

  • Sample rate max 20 k, but memory is 1 M. If need long speech sample rate 8 k… (!!)
  • Freq range limited

Circuit is about same as in singing greeting card.

I try make excelent speech; my sample speech is 4 second long, and using 20 ksample and 8 bit unsigned wav all memory of the circuit is used. Only way make working speech using this circuit is reduce sample rate, reduce band, cut away signals (as in this question). Maybe any working way is replace all silence with DC-offset, maybe.

(Of course, better speech is possible, eg. any Pic-processor or any etc, but… )

I assume that your device is also limited to using PCM audio (compressed formats such as opus not supported)?