Without the benefit of Rich's knowledge, I was about to say that I found that 24KHz result suspicious, and I guessed it was almost all tape hiss and other noise introduced at the time of playback.
I feel sure you can "get away with" a sampling rate of 44.1KHz, which would give you plenty of overhead, plus it would already be at a suitable sampling rate for conversion to audio CD [44.1KHz, 16-bit]. Sure, you *can* probably work with 96KHz, 24-bit audio, but for the purpose you describe, it's a waste of file space -- the files will be larger, but they won't sound better.
Of course the point of my suggestion to test a small segment still applies: do a test of the whole, anticipated process, using your 30-second snippet. Try recording it at 44.1KHz, 16-bit, listen, and compare that to recordings made a 22.5KHz or 48, or even 96, KHz. Let your ears help you decide. Shoot for the lowest sample rate that sounds fantastic to your ears. That will become your "archive" format, probably.
Next, put your test recordings through whatever process you hope to use for sharing that audio -- such as MP3. Listen to a few versions at different MP3 bitrates, compression quality, etc. Choose the one that you think best balances your desire for audio quality against the size of the file.
By hearing all the end products, side by side, you can more easily decide which setting is the best one to commit to hours of recording work.
I just want to point out that it's absolutely, 100% impossible for a 1962
tube-driven consumer reel-to-reel tape recorder to have usable information
at 24KHz. In fact, it's equally impossible for it to have information at
20KHz. In 1962, a $50,000 studio recorder operating at 15 ips (inches per
second) didn't go that high in frequency. Due to bandwidth limitations of
the tube circuitry, the tape heads, the tape medium itself, the age of the
tape, and the inexpensive microphone most likely used, I'd guess that you
have no meaningful information beyond, say, 12KHz to 14KHz. Anything above
that is tape hiss, or spurious digital noise that you don't need to
preserve. A 48KHz sampling rate is truly overkill, especially for voice
recordings. When you said "BUT", and that the sample had "significant
noise", can you describe the nature of that noise? Also, Audacity's noise
removal function will do so much damage to the original sound that you may
as well record in 8 bits at a 22KHz sampling rate. I think you would be
better of preserving the original sound that came off the tape, in all its
noisy historical glory. Keep in mind too that your vintage Akai is not
simply playing back the recorded sound on the tape, its circuitry is also
introducing all sorts of noise of its own into the signal. In fact, I'd
almost suggest assessing the noise and distortion level of your Akai by
recording a well-recorded CD to it (not on the same tape!!), and listening
to the playback compared to the CD. You might be very surprised at what you
----- Original Message -----
From: "Richard Folkerth" <firstname.lastname@example.org>
To: < email@example.com>
Sent: Friday, March 31, 2006 5:21 AM
Subject: Re: [Audacity-users] Voice Tapes to MP3 files ?? Spectral content
test results ...
> Thanks for all the good responses: Gary's in particular. I took his
> advice and ran a bandwidth test to see what the spectral content of
> my voice tape actually was. The tape I tested was recorded on a
> 'portable', tube-driven, reel-to-reel machine in 1962. This thing
> was portable in the sense that I could lug it around when I was young
> and strong. It died long ago and I am playing the tape on my prized
> purchase from a pawn shop: a functioning Akai 1710 machine from about
> that same era.
> I hooked the tape player to a USB iMic and thence into my Mac dual 1
> Ghz G4 with 1.5 Gb of RAM. I chose a 96 kHz sample rate and 32 bit
> I recorded a little more than 30 seconds worth to check the spectral
> content. I checked several sections of the sample, each around two
> or three seconds long. The Plot Spectrum menu is available under the
> View menu in my version of Audacity, so I used it to examine several
> different output formats until I found one that made sense to me .
> the Log Frequency plot. Here it was clear that the signal on the
> tape extends as high as 24 kHz but has a significant roll-off
> shoulder, 50dB down, at about 20 kHz.
> Following Gary's suggestion to use sample rates at least twice the
> highest frequency on the signal, I plan to proceed at a sampling rate
> of 48 kHz.
> But. There is always a 'but', it seems . the sample had significant
> noise. I used noise removal to knock it down but lost signal in the
> process. I'll post a fresh message about my noise adventures when I
> understand what happened a bit better than I do just now.
> Again, thanks to all who responded.
> DICK FOLKERTH
> This SF.Net email is sponsored by xPML, a groundbreaking scripting
> that extends applications into web and mobile media. Attend the live
> and join the prime developer group breaking into this new coding
> Mailing list: Audacityfirstname.lastname@example.org
> To UNSUBSCRIBE, use the form at the bottom of this web page:
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!
Mailing list: Audacityemail@example.com
To UNSUBSCRIBE, use the form at the bottom of this web page: