The soundboard that a friend used produces steady background ringing at a certain frequency. I have used a nyquist prompt script in the past to eliminate a similar problem with different audio hardware (ringing at a different frequency), but don’t know how to pinpoint the frequency of this ringing sound. Could someone look at this audio file to find what frequency the ringing is happening at? I’m guessing once that’s found then that frequency can be plugged into the following script that I used for my Blue Snowball mic?
(let* ((q 50) ; set the base Q for the filter
(freq 1000) ; set base frequency
;set the number of iterations
(iter (truncate (/ *sound-srate* (* 2 freq)))))
(dotimes (i iter s) ; start the DO loop
(setf s (notch2 s (* freq (1+ i)) (* q (1+ i))))))
Thanks. I’m using Audacity 2.1.3-alpha on Windows 10.
Yeah, I did the same, though there are about three frequencies that stand out in the first half of the clip that only contains room tone. In the second half with speech those frequencies are subsumed and aren’t as discernible amongst the speech.
That’s the custom filter for The Dreaded Yeti Curse. The Yeti has problems with getting rid of computer USB noise from your voice. It was “relatively” easy because most USB systems have similar data management and they all make about the same noise. If you have different noise. it may not be coming from the USB system and it could be a lot harder to suppress.
Drag-Select a relatively quiet part of the performance (noise only). Analyze > Plot Spectrum. Log Frequency, Size at 512. Grab the window and pull it so the window is as wide as possible. The accuracy goes up when you do that.
When we do that with The Yeti Curse, the display have nice peaks at 1000, 2000, etc. And that’s how the correction is written. You may need two different filters. One for 4000, 8000, 16000; and a second one for 6000, 12000 18000.
Plot Spectrum cursor will get stuck on the peaks and it will tell you the frequencies.
You picked a good batch of tones. Between 3000 and 4000 is babies screaming on a jet with their fingernails on a blackboard. If you suck enough different tones out of a performance (that’s what you’re doing), it’s going to affect the show quality.
I don’t think the single existing filter will work on that series of frequencies without being re-written.
I suggest a wider notch : q=30 rather than q=50 , for the 1kHz harmonics,
and another two notches at 269Hz & 474Hz …
(setf s (notch2 s 269 30)) ; apply 269 Hz notch
(setf s (notch2 s 474 30)) ; apply 474 Hz notch
(let* ((q 30) ; set the base Q for the filter
(freq 1000) ; set base frequency
;set the number of iterations
(iter (truncate (/ *sound-srate* (* 2 freq)))))
(dotimes (i iter s) ; start the DO loop
(setf s (notch2 s (* freq (1+ i)) (* q (1+ i))))))
Thanks everyone for the help! Trebor’s solution worked. A questions about Trebor’s code. What does the “q” represent (with a smaller number for q being a wider notch).
That’s the Quality of the filter. The higher the quality, the sharper and more surgical the notch, but the more likely it is to distort the show. If you make it too wide and sloppy, it starts deleting parts of the show that should be there.
