Update: After getting things to sound reasonably normal when playing our podcasts, I thought I’d try the same thing from other computers. It turns out that the older sermons sound pretty good, and the “fixed” ones sound a little distorted. I think the problem was not in the recordings themselves, but in the machine playing those recordings. The computer I’m typing on now seems to play all digital audio so that it sounds a bit chipmunkish. If you’ve been hearing a similar effect on the older audio archives, then your computer may have the same problem. In light of this, I’ve changed the encoding process again to use the 44,100 Hz sample rate in our recorder. If you think this is not adequate, feel free to let me know.
I’ve finally looked into the slight distortion in our audio sermons. It made my voice sound slightly chipmunkish, but otherwise OK. Though I suspected the problem was a mistake on my part, it turns out to be a mistake on the part of our microphone recording device. The non-technical will be glad to know that I can apparently correct it, so hopefully future recordings will sound closer to normal.
The technically-inclined may want to know what the problem actually is. My recording device, a Zoom H2, seems to be using a sample rate somewhat lower than the rate it claims to use. Since I record this audio to compact disc, I use a sample rate of 44.1 kilohertz, meaning there are 44,100 digital samples taken per second. That’s the standard for audio CDs. However, when the CD has been made, or when your computer plays a file made with the incorrect assumption that the source audio was truly recorded at 44.1 khz, it results in some distortion of the sound. The effect is the same as with a tape recorder that runs slower than it should. When the tape is played on a correctly-running tape player, voices will be chipmunkish.
The solution for now is to have the recorder use a sample rate of 48 khz (an industry standard for DVDs and other media), which turns out to be a little lower than that as well. Then, when I encode that audio data for compression, I override the incoming audio file’s reported 48 khz sample rate with a sample rate of 44.1 khz. The result sounds much better in testing.
On a related note, I apparently upgraded some of the software needed to add a service to the sermon audio archives on our web site, and the new version changed something. It rendered unusable the mp3 files from the first four Sundays in May, but I was able to restore the most recent one after finding and fixing the problem. The Speex files were unaffected.