MP2 at 192 kbps shouldn't sound objectionably bad. That's actually near the top end for what European DAB stations are using -- some are as low as 64 kbps. The HE-AAC codec that DAB+ uses was supposed to improve quality, but instead stations have used it to make their bitrates even lower, often 32 kbps.
If you would like to experiment with making your own MP2 encodings toward that end,
https://www.rarewares.org/mp3-others.php has archival copies of several command line MP2 encoders for Windows:
https://www.rarewares.org/files/mp3/mp2enc.zip
https://www.rarewares.org/files/mp3/tooLAME0.2l.zip
https://www.rarewares.org/files/mp3/twolame0.3.12b.zip
https://www.rarewares.org/files/mp3/twolame-0.4.0-x86.zip
I've ordered the links above from oldest to newest as far as their refinement and evolution go, so allegedly, the best encodings would be possible with the final one, although it might still be fun to experiment with the earliest one, good old MP2Enc from 2002.
If you really want to dive into the world of the esoteric,
https://www.rarewares.org/rrw/programs.php even has a small handful of MP1 encoders available for tinkering.
There is a cost to streaming at a higher bit rate, but it's relatively small. Raw bandwidth at the wholesale level doesn't cost much. A radio stream at 192 kbps would cost about 1 cent per hour in bandwidth to serve one listener. At 96kbps, it becomes half a penny per listener-hour. That can add up if you have a ton of listeners.
Does anybody still run their own, dedicated servers with self-managed Icecast and Shoutcast installations? Many datacenters now offer unmetered bandwidth at speeds of 1, 10, or more gigabits per second. Besides iHeart's ihrhls.com and streamtheworld.com, and Audacy's abacast.net, the most common large-scale audio streaming hosts I regularly see radio stations using today are securenetsystems.net, streamguys.com, Voscast, and Live365. But those all offer "premium" levels of service, beyond "plain vanilla" Icecast and Shoutcast daemon hosting. So you end up paying hourly listener pricing with them.
But even still, the music royalty (SoundExchange) is much larger, at about 3.5 cents per listener-hour for a music format.
That adds up even faster -- to insolvency. I thought rackets designed to price out small competitors were supposed to be illegal.
Theyre all on 128k mp3 now and might move to 256K aac plus at some point.
Did you mean AAC-LC? AACPlus (a.k.a. HE-AAC) will sound worse than AAC-LC at 256 kbit/s.
(Sometimes I think Fraunhofer was unwise naming the full fidelity version of AAC "low complexity." AAC-LC is only low complexity compared to the number of tricks AACPlus needs to implement to synthesize all of the treble. At 256 kbit/s, AAC-LC has plenty of bandwidth to convey the original treble with near-lossless quality -- definitely better sounding than a synthesized substitute.)
It comes down to one major thing... 99% of the potential listeners do not notice a difference.
99% is a cynical figure. I don't know what the actual percentage is that notices the difference. But I think it's enough to make the effort to satisfy them worthwhile. The proof is in the historical things radio did to satisfy that same audience segment. Like stations replacing their vinyl libraries with CD reissues. And eliminating carts for digital playout. And dumping suck-and-pump monoband processing for multiband Optimods (including spending big money as every new generation appeared, from the 8100 XT/2 on up through the 8700 and now beyond). Most listeners in noisy car interiors with non-premium sound systems, or listening through boomboxes or table radios, probably scarcely noticed any of the differences each of these incremental upgrades made. But the industry still made those changes to earn the extra ratings points from the listeners who did notice. In consideration of that, why would ditching the outmoded dial-up bitrates and codecs many stations still cling to today work out any less favorably? The people who're now eating radio's breakfast, lunch, and dinner, like Spotify and iTunes, have been delivering 256 kbit/s AAC-LC for a long time now. And in the last month or two, Spotify has been offering lossless streaming at no extra cost. Their research must be telling them these quality levels do matter to enough of their users to make their implementation and costs worthwhile. It should go without saying that given
their comparatively
massive global CDN and bandwidth bills, they wouldn't donate one solitary extra
bit to their listeners' ears if enough of them weren't noticing and sticking around for the difference.
And the streamers need those niche listeners to profit off of, but I guarantee the majority of those subscribers don't care. If they did, all audio would be distributed in .flac files only and not MP3.
Except that 256 kbit/s AAC-LC is the default for everybody who pays for just the entry-level tiers on platforms like Spotify. It isn't held back as a premium bitrate that costs extra. Even Spotify's new lossless option, again, is now available at no extra charge. This must mean they're seeing enough demand
among that majority to moot the increased costs.
Most people do not have a sensitivity to differences in bandwidth than those that are more "music obsessed". And those that are more "music obsessed" are also the ones more likely to be posting on a radio messageboard.
I don't think they're necessarily music obsessed in most of those cases; they're just ordinary listeners who happen to be more affected by old codecs and low bitrates than most. Often to the point of discomfort. That said, you don't have to cause people outright discomfort with your bitrate and codec choices for them to still notice the differences -- and for them to tune away from the inferior choices. For example, if you A/B compared synchronized 64 kbit/s HE-AAC and 256 kbit/s AAC-LC versions of the same songs through quality speakers to groups of people with undamaged hearing, most wouldn't understand
why the latter versions sounded better, but they would start picking up on the differences between the A (64) and B (256) versions as they continued listening. Eventually they would decide they preferred the latter versions. This is no different than when those old electronics chains like Circuit City and The Good Guys had big showrooms filled with "real" stereo systems with large speakers on display in big acoustic listening rooms. People would spend decent chunks of time standing in front of those systems, comparing them all against each other until identifying the one with "the best sound." Most of those customers never knew anything about the reasons their choices sounded the best -- stuff like amplifier classes and phase linearity. But they knew which one was "the best" after enough listening, and barring things like missing critical features, they bought it. It's no different with streaming. With time and in the aggregate, people will notice that Spotify and iTunes sound "the best" compared to radio stations who're shooting themselves in the feet with discount bitrates and obsolete codecs. Eventually, that will contribute to lost listenership.
I'm one of those that cannot tell the difference in a stream bitrate. It also takes a lot for a difference in radio station processing to be noticed. Only the really good or the really bad such as a Voltaire turned up to 25 can I tell.
Everyone is different. One of the big kerfuffles that previously happened in the world of radio audio -- the clipper aliasing distortion problems the Optimod 8200 allegedly had -- was never discernible to my ears. Lots of people were swearing up and down during the 8200's heyday that its clippers sounded like digital grunge. But with almost every form of popular music, I couldn't hear anything wrong with the 8200's sound. And this was at the point in my life where I could still hear to 20 kHz and distinguish lossless CDs from 320 kbits MP3 encodes almost all of the time.
I think the lesson is that whenever psychoacoustic things are done to audio, groups rather than individuals must listen for audible defects, because no single set of ears can ever be trusted to decide if the audio is okay. If even 1 in 10 people hears something odd, you can assume that 10% of your audience will hear something odd too, and that some subset of that 10% will hear something
awful and bail for a better-sounding competitor.
UK: hold my beer. We have a couple of music stations at 24kbps. At that point, you have to assume the business case is for them to exist for discoverability and to point people to a stream - nobody is going to listen to that crap day in, day out.
US: beer held. I can't even summon the words to describe how insane 24 kbit/s MP2 for music is. But I doubt discoverability is the motivation. How many listeners would realize the cause was a low MP2 bitrate DAB broadcast as opposed to just assuming "the station" sounded that way, whether listening on-air or via the internet stream?
ashlandcountyscountry.com
