Since your goal is to preserve the audio in as high quality form as the original, and to compress sufficiently to save storage space, I would suggest finding out what bandwidth is present in the original recordings. Record a few samples into Audacity at 48k or at 44.1k (or 88.2/96k if that is possible),  select short segments choose ANALYZE/PLOT_SPECTRUM. If the audio was bandlimited when recorded to AM radio quality, you will see no energy above 4kHz, FM quality would have energy out perhaps 14kHz. Make sure you sample at rate above twice the highest frequency energy in the signal. Sampling at below this “Nyquist Sampling Rate” will result in aliasing distortion that cannot be repaired by any means. The energy above HalfNyquist will fold back into the main signal spectrum.

 

BTW, digital voice over the telephone is sampled at 8000 samples/sec with 8 bits per sample è 64kbits/sec. A Dolby-like compression is applied to 13-bit samples to yield the 8-bits, so the dynamic range of telephone voice is 6 dB/bit x 13 bits = 78dB (expressed in audio marketing numbers).  Telephone quality voice is barely adequate, so we need to be aware of a need to maintain about 80dB of dynamic range and Signal to Noise ratio and also to have enough bandwidth to have good intelligibility. Telephone bandwidth is 300Hz to 3400Hz and we can’t tell the difference between “sss” and “fff” because the information to discriminate those sounds is above the telephone passband  -- 5000-6000 Hz.  A sampling rate used in commercial radio is 32,000 samples/sec which gives a practical bandwidth of 12k to 14kHz.  That is plenty good for voice, but you also need to preserve 13 or 14 bits of signal which is awfully close to the 16 bits of CD audio or 16-bit ADCs. More on that in a minute.

 

I don’t see a benefit to choosing a low sampling rate for the source material given that you intend to compress using MP3 – your total storage volume will depend on the choice of output bit rate of the MP3 encoder. MP3 encoders subscribe to the GarbageIn-GarbageOut maxim, so you might as well create best quality inputs. 

 

The next issue will be to get the best possible Signal to Noise ratio in the final product. To make this as straight forward as possible, I would suggest that you choose an audio converter that has 24-bit Analog to Digital Converters and set Audacity to store the samples in 32-bit floating format. That is wasteful of initial storage, but makes any post processing less susceptible to introduced artifacts from post processing effects.

 

Next select each audio clip that will become a track and choose EFFECT/NORMALIZE to amplify the signal to within 3dB of saturation, and then export the selected track into MP3. You can set up the MP3 encoder to have a lesser quality, but it defaults to 128kbits/sec which is more than good enough for voice. The suggestion to use 24bit ADCs is just to make it less important to have the recording levels set perfectly. One thing to be avoided in digital recording is clipping.  To avoid clipping, we need to set the recording gain low enough so that there is plenty of headroom. But that puts the bottom of our signal down close to the quantizing noise floor of 16-bit converters. Better to use 24 bit converters and push the converter noise floor down so low we don’t have to be concerned. Now if the normalization adds 12dB of gain, the converter noise floor will still be below the threshold of 16-bits.  If you export to .wav with 16-bit samples at 44.1 (CD-format), be sure to go to EDIT/PREFERENCES/QUALITY and select export dither as shaped or triangle. Dithering adds some random noise to the bottom bits and contrary to intuition, actually improves the quality of the result. Not a big deal for voice, but it is so painless to check a couple boxes that you may as well do it.

 

Audacity is quite amazing in its completeness. Kudos to the authors.

 

Good luck

Gary Nelson

 


From: audacity-users-admin@lists.sourceforge.net [mailto:audacity-users-admin@lists.sourceforge.net] On Behalf Of Moon Caine
Sent: Wednesday, March 29, 2006 3:10 PM
To: audacity-users@lists.sourceforge.net
Subject: Re: [Audacity-users] Voice Tapes to MP3 files ??

 

I'd pick 64KHz for sampling rate, too, and maybe try 22.5KHz for the sampling rate ... but whatever you do, RUN A TEST of your entire planned procedure, using maybe just a 1-minute audio clip, to find out if it works well and sounds good to you. If you decide to change a setting, such as bumping sample rate back up to 44.1KHz, then run another short test. In other words, don't do a lot of work that you might regret later. Test first.

I'm guessing that the recordings are quite important to you, so you may want to consider preserving the recordings as is, without editing, in some durable format like CD, before you start editing. Maybe your late Aunt Tootie's constant erm and hrmph sounds annoy you, but her nieces like to hear it because it's a warm and friendly memory of the way Aunt Tootie talked. It'd be great to have the original for such situations -- I think.

--moonie

On 3/29/06, Sarah <kales2@cox.net> wrote:

No I don't see a problem. In fact I'm going to do just that later this
summer. I'd thing about 44.1k for your sampling rate and 64kbps mono for
your bit rate.
----- Original Message -----
From: "Richard Folkerth" < rfolkerth@sbcglobal.net>
To: <audacity-users@lists.sourceforge.net>
Sent: Wednesday, March 29, 2006 1:59 PM
Subject: [Audacity-users] Voice Tapes to MP3 files ??


GREETINGS, ALL

I am a relatively inexperienced Audacity user and I am about to
digitize some old family voice tapes using Audacity 1.2.4 on my Mac.
But I have some questions.

I prowled thru the list archives for the last several months and
picked up some good hints and tips . but everyone seems to be dealing
with music, rather than with voice.

Here is my plan: digitize the old voice tapes, do a light edit on
them to remove long silences and terrible redundancies and false
starts . then export 'em as MP3 files . and burn an archival CD.
Then go thru them again to add explanatory voice inserts or
voiceovers ( to put the 1960s talk in perspective ) and then do a
heavier edit.  I believe it would be best to capture all the stuff in
digital form before my reel to reel tape player dies . and then do
the heavier edit and cleanup . and finally burn CDs for all my
children of the final product.

I understand that MP3s will not play on yesterday's CD players and
that is OK.  I have convinced myself that MP3 will be adequate
quality for the end product . since they are voice tapes NOT music.

The questions;
1. What sampling frequency and bit depth make sense for this voice
application when the final product will be MP3 files ??
2. Are there any important settings in Audacity while saving such a
voice recording as an MP3 ??
3. Does anyone see a problem with this approach ??

DICK FOLKERTH



-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
--
Mailing list: Audacity-users@lists.sourceforge.net
To UNSUBSCRIBE, use the form at the bottom of this web page:
https://lists.sourceforge.net/lists/listinfo/audacity-users



-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
--
Mailing list: Audacity-users@lists.sourceforge.net
To UNSUBSCRIBE, use the form at the bottom of this web page:
https://lists.sourceforge.net/lists/listinfo/audacity-users




--
-- I can believe it's not butter.
mooncaine@gmail.com