Monday, March 21, 2005

An idea on how to get more sound out of IBOC HD Digital radio.

I need to develop this, but just had a great idea for making the best of the fewest bits.

Everyone knows the best quality audio encodes take a long time. Given enough information, a good acoustic compliler could be written (Yes I will define that) that would significantly outperform more real-time codecs.

One problem IBOC radio has is the limited bitrate. 96Kbps max is just not enough to convey CD quality audio without significant effort. And that's the other problem, being time. We have an 8 second latency as it is. Adding more is not going to fly. So the choice is either more compute (expensive, particularly on the receiver side), or more time.

Why not pre-compile program segments? Bumpers, imaging, commercials, and preset music sets could be encoded to a much higher standard than is currently possible in an 8 second window. This is where the 'acoustic compilier' comes into play.

This special piece of software would do the following:

1. Pre-process source material. The output of this stage is a collection of information about the source material that can greatly help the codec make it's best moves. (Determining noise sensitive segments, silence, commonalities, etc..)

2. Enhanced encoding stage where both the source material and hinting information from stage one are combined for a greater encode.

3. Analysis. The resulting stream from stage 2 is compared with the original source material. Flaws are noted and further catagorized according to some information from stage 1 and some other information to be determined by the broadcaster. (Acceptable tradeoffs, priorities, etc..)

4. Tweak. The bitstream is then tweaked to make the most of the receiver codec according to the analysis stage. This stage and the previous one form a loop that interatively approaches the source quality until no further returns can be made.

If this isn't clear, consider early computers and sound. Playing sounds through limited hardware does get better once the limits of the medium are well understood. As people learned to "play" the system, the resulting sound got better. They did this by spending time and making tradeoffs. The IBOC / HD Radio system is a limited system in that it cannot just reproduce full-on audio as an analog one would. This idea is simply expanding on that process in order to get better sound out of the system than that intended.

I got this idea from hearing different mp3 codecs and players. Some combinations sound really good, others bad even though the basic process is the same. Also, some mp3 encoding software, such as LAME, already do this to a degree, resulting in a better sounding program by bending the rules a bit to the listeners benefit.

(I'll put up a picture of what I am thinking later.)

Once created, these pre-encoded streams would then exist on some sort of storage system, at the ready, to be inserted into the main broadcast bitstream when it makes sense to do so.

These 'compiled' streams could also be re-distributed to interested listeners to be played on other equipment. Their professional quality nature, compared with real-time and or consumer level codecs could make them attractive enough to be worth paying for, or at the very least provide another means of getting additional ad revenue through an archive service of sorts.

The core idea behind this is the common elements to radio. There is a lot of material broadcast today that really can be built ahead of time, with the idea being to limit the real-time material as much as possible.

Here are a couple examples I could think of that might clarify this idea further:

Archives: Many talk stations will recycle their shows during off hours. Those stations replaying syndicated content could simply receive compiled bitstreams ready to play at the highest quality possible. Lets say a talk station has one additional audio stream running in addition to their main hybrid digital / analog one. Compiling a broadcast in this way would likely bring the secondary stream audio quality on par with the primary one. Lets further say the secondary stream time-shifts the live broadcast for off-hours listening. These listeners would experience a much higher quality program.

Station imaging, bumpers, jingles and other "already cooked" bits. I think almost everyone agrees on the quality being a necessary element of a strong station identity. The extra clarity would make these elements stand out in a subtle way that is powerful but not annoying or tiring to the listener.

Highly automated evening sets, music bundles and special program segments. Stations here do "three of an artist" or "focus on whatever" or simply play daily sets with very similar rotation. All of this material can be compiled ahead of time for the best quality reproduction possible. This combined with already build imaging elements would make for compelling radio when combined properly. Long sets, such as those used by oldies stations would also fit into this catagory. With storage being cheap, several variations on these would be possible. --Just rotate sets instead of specific tunes.

I suppose that's enough for now. I just thought I would put this out there for discussion. --You read it here first!

(This post will see some serious edits when I have the time.)


Post a Comment

<< Home