|
ABN Antioch is an Old-time Radio Shows station
located in Antioch, IL, a mile from the Wisconsin border and 20 miles
inland from Lake Michigan currently transmitting at 1610KHz. The AM
transmitter is a micro-power part15 AM RangeMaster1000 transmitter.
Range is very limited due to just 100mW of power and competition with
many distant stations but it is sufficient for all the antique and
crystal radios in the house and can be heard well by the near neighbors
if they care.
About 14,000 shows are reviewed and approved to be automatically scheduled out of the whole library
which has about 70,000 as mp3 and about 30,000 on reel tape with considerable duplication between the two.
The source of the playout runs on a
Late 2010 Mac Mini paid for by listeners that year which has the job of
scheduling and playing all
shows with in-house custom software begun in 2003 and using iTunes.
Library Selection:
Collections are merged from various
sources. As of 2018, these sources include recently digitized reel tapes.
From that I compare and make judgements on air-worthiness, fix and adjust
audio to remove hum, reduce hiss, trim, speed-correct, and EQ. I confirm and identify and correct
episode names, dates and spellings where I can. Some misspellings by other collectors get past me sometimes.
Some collectors have already done their own
noise reduction giving a very artificial artifacted sound or gating.
For audio quality, I prefer hearing some surface noise to a loss of fidelity and I'd sometimes
prefer a more complete show if the audio is a little worse and
sometimes I'll even merge two files into one. When I do noise reduction I try to make sure
it doesn't become distracting. People's ears can adjust to a constant noise. When I get a chance, I
replace these if I have a better copy digitially or get the chance to get it myself from the reels.
Large collections take
years to listen to so this is an ongoing process. But I spot check
everything by listening to the first 30 seconds, sampling several
places in the middle and listening to the end. I sometimes have to
remove long annoying 30 second to 2 minute trailing silence. I've often
done pitch corrections and careful noise reduction where I can to
remove clicks and narrowband noise such as hum using notch filters and
low and high cut filters but where I have control over it I'm careful
not to make noise reduction a distraction such as with heavy noise
gates.
Getting a large collection isn't too difficult.
It's the work of sorting out a collection from multiple sources that
takes a lot of time. Even from the best "HQ" sources
there can be problems so having more than one HQ source is an
advantage. From the more commonly available mp3 sources I've heard the
worst examples of very bad
decisions at the digital stage such as encoding stereo at a mono bit
rate like 32Kbps stereo 22KHz sampling. I've heard noise reduction
techniques that make things sound underwater. I've heard AOL sign-in
and Windows desktop sounds (just lovely and embarrassing to hear played
on a Mac). I've heard mp3 encoding glitches that sound like hiccups.
I've heard large amounts of leading and trailing silence which is like
counting rings on a tree as I hear the different hiss sounds of many
generations of tape recordings. I've encountered lots of missing ID3
tags. I've encountered duplicate files just with different file names
claiming to be an episode they're not. I've heard audio drop-outs
consistently every few seconds for an entire 900-show Suspense
collection.
Scheduling and
Automation:
Scheduling of shows is an automated process where
shows are selected by the following priority:
- Shows that match today's date that have not played recently
- Shows that match today's date regardless of when they last played
- Shows that match yesterday's date that haven't played recently
- Shows that match the day before yesterday's date that haven't played
recently
- Randomly selected shows that haven't played recently.
Where recently = a few months to a few years
depending on category. In the case of Comedy shows, this can be 2 to 3
years whereas with Spy Stories it can be just a couple months. Serials
just rotate.
The dates are read as a date object and audio is
concatenated to present an introduction of original play date including
the day of the week. Station IDs and an "Up Next" file are compiled
along with music fill.
Since we're matching today's date and since most
of you don't stay awake listening for 24 hours, there are replays of
several sub-genre segments across the day so you don't have to worry
about missing much.
When there's room at the end of a segment, we
fill with public domain classical music from musopen.com.
Audio Processing
The source material is varied in quality so in
audio processing I tried to make the best compromise. First the tracks
are virtually normalized with iTunes Sound Check function. iTunes plays
the audio which uses Volume Logic to lightly process (AGC, 5-band
compression, and limiting) the audio in realtime. Nicecast digitally
captures the audio that iTunes is playing. I use a couple Apple Audio
Units and VST effects in Nicecast to roll off the extreme highs and
lows (such as rumble) not useful or desirable for AM or low sample-rate
streaming.
From there it goes to two places: 1. Nicecast
encodes the audio using LAME and passes it on to an icecast server on a
major internet trunk at fast-serv.com which serves the internet
streams. 2. Nicecast runs a parallel audio chain which processes the
audio more aggressively with audio units before sending it out through
an audio transformer, and then to the transmitter in the back yard.
|
