We are recording voice educational lectures in wav with an H2 recorder and would like to compress them to place on a website. We notice a bad ringing at the 16000 Hz MP3 joint stereo setting compression in Audacity and there is still some ringing even at 40 kHz. The 40 kHz files are too large for email or slow downloads.
We got no noticeable ringing when using iTunes to do a mono conversion of the wav file at 16 kpbs. We would prefer to stay in Audacity because iTunes does not allow other editing and also requires us to export a wav file from Audacity for conversion.
I analyzed the spectra of the various recordings and attached a picture. The wav file and iTunes show a fairly linear drop in signal with increasing frequency and then it stops. Both the Audacity compressions show a precipitous drop to -140 dB. The fall off gets increased with the 40kbps but is still there.
Could this be the source of the ringing?
How can I make a clean sounding mp3 file in Audacity at 16 kbps?
(We had previously used a Sony ICD-P620 for the task and got great results and small file sizes, <20Mbyte/2hour lecture, but the AV techs wanted the H2. I tried a low pass filter to get rid of -140dB signal, but that didn’t work.)
<<<2. How can I make a clean sounding mp3 file in Audacity at 16 kbps?>>>
You probably don’t. The minimum bit rate for a mono show is 32. Stereo is 64. Lower than that, the show rapidly gets bubbly, honky, and muffled.
You are warned that compressors only work when they’re presented with perfect, clear, crisp original work. MP3 compressing a show that started out MP3 or otherwise damaged is a shortcut to gargly disaster. The transfer from the H2 to Audacity should be as high a quality WAV as the system supports. Do Not use MP3 in that step.
Segment the show in blocks of time that fit. Few people can email an hour show that sounds perfect.
You might also try AAC instead of MP3. Anybody with an iPhone, an iPod or a Mac can play those. You can get iTunes to create those or you might be able to get FFMpeg to do it in Audacity 1.3.11.
You don’t really need stereo for a lecture, for the same kbps rate mono will give better quality sound than stereo: less burbly compression artifacts.
For voice-only (not music) recording, 10Khz bandwidth is plenty (you could even get away with 5KHz but it’ll sound like a telephone).
Resample* a copy of your original wav recording at 22050Hz (instead of the 44100Hz default) this will reduce its size with negligible loss in quality for a voice-only recording. Then make a mono version and save as mp3. Attached are examples of what it’ll sound like at different kbps (mono not stereo).
If I had to listen to an hour I wouldn’t use less than 64Kbps for a 22050Hz sample rate voice-only recording (mono).
-140db is silent, you’re trying to remove something that can’t be heard: the threshold of audibility is about -100db.
[* “Resample” is in the “Tracks” menu in Audacity 1.3 }
nothign wrong with using H2
dont like H2 then buy another sony
set the H2 to record at the lowest mp3 rate you can tolerate
and set it for mono
then leave the file alone
you cant compress it any more and make it better
not sure what benefit audacity is bringing to this scenario
all you can do is make the file worse if you overprocess the minimum (and already compressed some) size mp3 file
If you let the H2 compress the work, then you can’t bring it into Audacity for further processing and editing. Audacity will not edit MP3 tracks. It imports them, converts them to high quality format (with no increase in show quality) and then edits them. If you want an MP3 later, Audacity has to export one – in effect doubled-compressing the work which always sounds terrible.
If you go that route, you don’t need Audacity at all. You should be using one of the fine MP3 editors out there.
<<<If I had to listen to an hour I wouldn’t use less than 64Kbps for a 22050Hz sample rate voice-only recording (mono). >>>
Yes, 32 is seriously on the edge. 64 in mono is much better – if you can stand the file sizes.
However.
What happens if you leave the original work at 44100 instead of down-sampling to 22050? Did you ever actually try that? I think downsampling is a serious wife’s tale.
Destructive compressors all process much better with the best, highest quality work you can supply. Their job is to preserve as much quality as possible. Pre-damaging the work usually creates sound (or video) problems with no great benefit anywhere else.
<<<If they do that they won’t have the oppertunity to make a high quality presentation, say on CD, or via the internet when bandwidth improves.>>>
Yes, that does make my teeth hurt. I come from a background of “camera master negatives” and “studio master sound tracks.” You can always go downhill from those to any other product you wish. If you already have a degraded product and you haven’t even finished speaking, that’s just begging for trouble.
If Palmeroo isn’t aware how to convert a stereo track into a mono one on Audacity
“Stereo track to Mono” is on the “Tracks” menu …
Sacrificing stereo will reduce the file size and compression artifacts for a given kbps.
that is the point –
if you let the H2 compress while recording and make mp3 you don’t need to process with audacity. h2 will even normalise it.
note the h2 does not compress the mp3 after recording it.
if you are going to use audacity then start with a wav file at least 16x44100.
<<<the world today thinks mp3 is high fi
and anything “better” sounds strange to them>>>
Unless the bubbling gets too bad and then it sounds strange to everybody. Editing is about control. If you do production in a compressed format it gets progressively worse and you have no control over it. You can go down to MP3 from a high quality format, but you can never come back.
Trebor’s guidance on converting to mono before converting solved the ringing problem and should allow us to get adequate sound quality for the lectures in a compact format.
My apologies to all of you audio masters, but these are walk around lecture audio notes for docents at a major metropolitan art museum. The docents understand and appreciate at least the visual arts and want to share their appreciation of real 3D objects and not some over compressed jpeg thumbnail photographs. Most docents have attended the walk through lecture and the mp3 has been a helpful supplement for their training and memory.
We would have continued with the Sony mono recorder but changed to the H2 to please the AV Techs. The H2 allowed direct plug in into the audio system and avoided ruffling feathers with multiple microphones on the speakers . We have had problems getting from the H2 wav format to the size and adequate quality we had been used to with the Sony. Thanks to your help it looks like we now have a workable solution with the H2 and Audacity.
We will probably continue to use the H2 in the wav recording mode. The minimum H2 recording frequency for mp3 is 44 kHz, I believe, and that would be too large an mp3 file. Audacity has been helpful to allow us to edit interruptions and gaps, and normalize or adjust the volume. We just had trouble getting the Audacity mp3 export small enough as a joint stereo file without the mono tip from Trebor.
We had managed to get a satisfactory quality and size by using iTunes to convert the wav file into a mono mp3. But using iTunes would have required us to export the Audacity edits into another wav file to be processed by iTunes.
Staying in Audacity will now more simply give us all we need, even if we have had to sacrifice much of the audio fidelity. Sometimes you just need a peanut butter sandwich to keep going and a five course meal is not needed or useful. Please understand our needs and constraints may force us to compromise audio quality. We would not think of trying to use these files for anything other than a small group of personal users, nor mean to imply the extra bits are not necessary or useful.
Here’s some spectral output from the various recordings for a noisy period at 0.2-0.4 sec. The 16000 Hz mp3 conversion of the stereo file shows more spikes which could indicate the ringing. The mono file, suggested by Trebor, shows less structure and is listenable, at least for lecture notes. A signal processing guru colleague suggested the stereo mp3 spikes could have resulted from the channels stored as L+R and L-R. Or is there another explanation?
If combining the stereo tracks to mono in audacity is creating a ringing effect, (possibly a comb effect if one mic is further from the lecturer than the other),
then instead of combining the stereo tracks (L+R) to produce a mono one, just discard either L or R track and make the remaining one mono.
e.g. use “split stereo to mono” on your original stereo track then discard one of the two mono tracks produced. If one track is inferior, (quieter or noisier), then bin that one, otherwise delete either one. Clicking on the [x] on the top left corner of a track deletes it, (this deletion is reversible using “ctrl” +“z”).
