“A dropped stream is the nemesis of any regular Public Radio Tuner user. Nothing is worse than being caught up in a great public radio program and have it suddenly cut out….”

Some challenges can be expected:

• loss of signal while roaming from cell to cell. Networks are optimized for voice calls but not for data yet.
• minimal bitrate like 32 kbps is desirable but connection is still not guaranteed and sound quality is no great
• buffers have to be implemented in receiver to mitigate signal loss.

Results of a survey made by PC World suggest that 3G coverage may not be adequate for the delivery of sustained bitrates in major cities in USA. Like this table shows, networks speeds can be impressive but their reliability vary greatly so that live radio transmissions may be hard to achieve.

There is certainly a lot of room for experimentation here in this new area but I tend to believe that it could take a while before we see 3G replace true “physical layer” broadcast networks for live transmissions.

“At one point cern was toying with patenting the World Wide Web. I was talking about that with Tim one day, and he looked at me, and I could see that he wasn’t enthusiastic. He said, Robert, do you want to be rich? I thought, Well, it helps, no? He apparently didn’t care about that. What he cared about was to make sure that the thing would work, that it would just be there for everybody. He convinced me of that, and then I worked for about six months, very hard with the legal service, to make sure that cern put the whole thing in the public domain.”

The least we can say is that the strategy worked. The Web is now ubiquitous.

Is there a lesson here for creating the mobile broadcast system of tomorrow?

As a first step in this reflexion, I’d like to estimate the total bandwidth of Twitter, that is, how many kilobits per second are being Tweeted on average.

I made a similar exercise some time ago with regards to the blogosphere in a post titled “Broadcasting the Blogosphere: 30 million voices for the price of one!”.

So I found some twitter services that provide relevant data. For example, TweeSpeed is an instant speed meter that shows the current number of tweets per minute. A graph showing the speed per hour during the last week is also available. A quick look at that graph now suggests that 700.000 tweets per hour would be a reasonable approximation for last week’s average, excluding the peek caused by the “Michael Jackson Effect”. Twitpocalypse currently reports 221 tweets per second which results in a similar value (221*60*60 =795.600 tweets per hour ). On another front, the recent HubSpot State of the Twittershpere report provides similar amounts on a daily basis instead of per hour. I suspect that this is a mistake. I’ll be pessimistic and take the largest number. The Hubspot report also informs on the distribution of actual tweet length. I’ll average the tweet length to 110 characters per tweet.

So the math goes like this:

110ch * 1byte/ch * 700k/hour = 77 Mbytes/hour

or

TOTAL TWITTER BANDWIDTH = 170 kbps !