After I posted the "Intro to VoIP Security panel podcast, I received a nice note from a listener suggesting that I try a high-pass or notch filter to remove the buzzing that in the podcast.  The note prompted me to investigate further because I, too, had been rather annoyed to have this sound in the recording, especially because I had been recording directly from the conference room mixer!  I heard it while there at the Internet Telephony conference, and tried tweaking the mixer a bit to see if I could drop out the buzz, but found no way to do so and had to conclude that it was originating somewhere in the audio equipment they were using.  So I did really want to get rid of it... but then when I came back wanted to get at least one of the panels out and didn't have time to track down the problem.  But after the comment and before I did the next panel I searched...

...and found in the Audacity wiki - "Eliminating a Continuous High-Pitched Whistle-Like Noise from a Recording", a great little tutorial that helped me figure out what I needed to do.  It involves a few steps, basically:

1. Select part of the recording that is the closest to silence that you can get - where you only hear the buzz.
2. Go to the "Analyze" menu and choose "Plot Spectrum".
3. Identify one of the peaks.
4. Go to the "Effects" menu, choose "Nyquist Prompt" and enter "(notch2 s freq value)" (ex. "(notch2 s 1019 25)")
5. Play the sound and listen for the buzz.
6. If the sound is still there, go back to step #2 and identify another frequency and repeat the process. (writing down the frequencies you are using as you go along)
7. When you have eliminated as much of the buzz as you can, select another segment of sound (preferably several seconds) that includes human voices and repeat step #4 for each of the frequencies you wrote down (unfortunately, per another forum post, there appears to currently be no way to set up a notch filter for several frequencies).
8. Listen to the resulting segment to ensure that it still sounds okay (hopefully sans buzz).
9. Select the entire audio file and apply the notch filters to the entire selection.
10. Listen to your clean(er) audio file.

That's basically what I did... although in thinking about it I might have had a step between 8 and 9 where I used "Undo" to remove the notch filters on the small human voice segment before applying the filters across the entire file.  For my own record, here's the sequence I did:

(notch s 1019 25)
(notch s 2046 25)
(notch s 21649 25)
(notch s 21650 25)
(notch s 21652 25)

That all seemed to do the trick. There still a bit of a low hum, but I also tried a high-pass filter that would basically wipe out everything under a certain threshold, but using numbers down like 100Hz I didn't discern any real difference - and I was reluctant to go too much higher and start impacting voices.

Anyway, you can hear the difference on the "VoIP Security Best Practices" panel.  Still a small buzz... but at least the high-pitched one is gone.  (And suggestions on killing the low buzz are always welcome. :-)

Tags: audacity, audio, podcasting, post-production

Had a bit of trouble today when I went to upload a new Blue Box episode.  It sounded fine when you downloaded and played it, but when you used the flash-based player I have on the site, it sounded like we were a bunch of chipmunks!

It turns out in reading through the LibSyn support forums (because I was thinking of switching to using the LibSyn flash player instead of the one I use now) that the issue is with the sampling rate in the MP3 export rate out of Audacity and LAME.  Sure enough, I looked in iTunes and my freshly-exported MP3 file had a sample rate of 24Khz versus the standard 44.1Khz.  What happened?

Well, it turns out that Audacity lets the LAME encoder control the sampling rate! You can set the bit rate in Audacity, but Audacity lets LAME adjust the sampling rate.  I never had this issue before because I've always used the default bit rate of 128Kbps, which must just use the default sampling rate. However, as I wrote earlier, I'm looking to use a bit rate of 56 to get much smaller file sizes.  This was my first upload at the new bit rate.

Only way I could work around this right now was to go to the command line to force an output sampling rate of 44.1:

lame.exe -b 56 --resample 44.1 inputfile.wav outputfile.mp3

Ta da... an MP3 file that plays well in the flash MP3 player!  Just had to delete the old one on LibSyn and upload a new one... hopefully I didn't delete the old one while someone was in the middle of downloading (but I probably did).

Ah, the fun of all these different audio issues... :-(

Tags: audacity, audio, lame, mp3, podcasting

One of the things that had continued to puzzle me was why Shel Holtz was able to get the FIR podcast audio files to be so much smaller than the audio files I was getting for Blue Box.  I knew that he used the same tools as I do - Audacity and the LAME MP3 encoder - but yet my files were averaging around roughly 1MB per minute and his were half that, i.e. a 10 minute podcast for me would result in a 10 MB MP3 file while for Shel it would be 5 MB (not that either Shel or I really ever produce a podcast as small as 10 minutes!).

Yesterday in the midst of some other things I fired Shel a message and he clued me into the one setting that is different between us.  I had the bit rate for MP3 export set to 128... and he had it set to 56.

Ta da... one little change and now my audio files are about half the size!  (Hmmm... I wonder if I set it to 64, would they be exactly half the size?)  On a quick listen, I can't really hear any difference between the two files (one exported at a bit rate of 128 and one at a bit rate of 56).  Given that our show is a talk show, we don't necessarily need the audio quality of a music podcast.  I'll have to export the next Blue Box episode at that bit rate and see if anyone has any comments.

Note: Here's a nice little piece in Podcasting News that talks a bit about adjusting the bit rate and gives some samples recorded at different rates.

Tags: audacity, audio, fir, podcasting

In working with Audacity and needing to fade a section of a podcast but not wanting to fade to zero (which could easily be accomplished by Effects->Fade out...), I stumbled upon this great tutorial which introduced me to the truly wonderful envelope tool.  What a truly awesome tool!  Fantastically easy way to fade things... but even more so to adjust gain when you have a very varied recording.  I've already received the comment that Jonathan's voice was much better on our latest podcast (#6) and part of that is the method we used for communicating, but part of that is also that I used this tool to raise the places where his voice was low.

Very cool tool!  And thanks to the transom.org folks (whomever exactly they may be) for creating this tutorial... I'm checking out more of their site as well.

Tags: audacity, audio, podcasting
Current Mood: impressed