Hacker News Reader

Comment by jaylevitt | Hacker News Reader

I'm sure that's part of the evolution of FaceTime technology, but the article's author doesn't seem to really understand the technology he's touting. If cell phone antennas result in poor audio quality, they will also result in poor data quality, and thus poor audio-over-data quality. Either the antenna works well or it doesn't.

Similarly, ALAC would be a horrible way to send voice calls over wireless. Lossy codecs will require less bandwidth and thus be less susceptible to dropouts on poor-quality connections.

And while I know almost nothing about codecs, I'm pretty sure you have to design one specifically for streaming - you can't just take a write-to-disk format like ALAC or FLAC and decide to stream it wirelessly.

View on HN

Not only you need a codec designed for streaming, you also need it to be resistant to bit errors. This is a big technical issue that most people do not understand when they compare voip targeted codecs with cellular codecs. In regular voip communication, you might encounter an occasional dropped packet, but by and large, most packets make it to their destination and they arrive there intact. Voip codecs are designed with this scenario in mind. In cellular voice communication, a device has to accommodate for frequent bit errors in the packets it receives, and codecs are designed to cope with this (otherwise you'd get an inordinate number of bad packets that have to be dropped). This is why you need significantly more bits to achieve the same perceived audio quality with a cellular codec that with a traditional voip one. Yes, you can do regular voip over a mobile data connection, but the results will be substandard in less than perfect connectivity scenarios.

maybe this is a dumb question, but why are so many packets dropped in cellular technology? is it an issue of the way the network is built, or is there some sort of physical constraint? or something else?

I doubt phone antennas are the bottleneck, they're a pretty minor part of a huge, lumbering voice-transmission infrastructure. Much of that infrastructure is presumably expensively, archaic, and built assuming that audio traveling through it is compressed all to heck.

You can stream tunes in real time over data with quality far in excess of what a high-def voice call would need. There's bandwidth to spare on the data side of things.

This is a meandering article. Marco Arment's response at http://www.marco.org/2011/10/27/high-definition-audio conveys the point much more clearly and succinctly.

Audio and video codecs are generally stream/packet based. You'd really have to screw up the container format to make a codec completely incapable of streaming. (Like putting the last frame of video at the start and the first frame at the end of the file)

If you use a player like VLC you can generally watch videos while they are downloading.

Low-latency is required for realtime applications, which many codecs are incapable of.

The issue is more so whether the codec can handle dropped packets. It wouldn't surprise me if FLAC could not recover in such a case; usually lossless codecs employ something like Huffmann coding which places a large dictionary at the start of the file.

Reply on news.ycombinator.com