Let's start estimating the technical challenge here (we're supposed to be techies, after all).
Bandwidth: quick googling shows that a flight data recorder records between 64 and 256 12-bit words a second, i.e. 1-3 kbps. Browsing http://en.wikipedia.org/wiki/Bit_rate shows that another 2-8 kbps is needed for voice recording (from one microphone). Let's call it 10 kpbs all together.
Number of simultaneous connections: there are around 20,000 (ballpark) airliners in the air at the peak time each day.
Compare this to the rather old Iridium satellite network: 2400 (uplink) baud rate, 1100 calls simultaneously per satellite (of which there are 68). Not quite enough, since airplanes are not evenly distributed over the Earth or under the satellites, but not outrageously far from our requirements.
Cost: several billion dollars (under 10, I think) have been invested into building the Iridium network, and their operating expenses are around 100 million per year (they are a public company).
Clearly, it would not be unrealistically expensive, in the world of 500 million dollar airliners, to build and maintain a network of satellites to record all black-box voice and data in real time.
How many aircraft are lost that satellite link "crash data" would have been useful? I recall three, MH370 would make four. I imagine there are more, but compared to the number of aircraft that don't crash, the total is a rounding error.
Of the three, TWA flight 800 is the only one whose cause is controversial, and in-flight data transmission likely would not have helped identify the cause of that crash.
Of all the above accidents, the causes are understood and have not been repeated. In fact, the cause of Air France 447's instrumentation failure was understood before the crash - there was an an active airworthiness directive to fix the pitot icing.
Satellite could just be a fallback only used when you're not over ground. This would remove the load of all US/European intra-continental flights (the vast majority).
Part of the reason Iridium needs a lot of satellites is because they're in low earth orbit, to reduce latencies (800km instead of 42000km for geostationary orbits means round trip latency of 5ms instead of 280ms).
You might be able to lower costs by tolerating increased latency and using higher orbits. You'd still need something for pole coverage, of course.
Why use satellites? There is plenty of long-range terrestrial radio tech that could do the job. Land-based and or even cargo ship based listening posts would be much cheaper. Aircraft-to-aircraft comms could also potentially do the job.
Granted, aircraft fly high in the air, but there are no line-of-sight problems as with terrestrial radio links, so even high frequency comms are possible.
Keep in mind that over-the-horizon radio communication is subject to the vagaries of the ionosphere and is not totally reliable. Further, there is a day/night difference, and a dependence on latitude. Line of sight at 30000 feet is about 200 miles, so there are huge areas where everything is over your horizon.
But don't most actual aircraft flight paths tend to stay near land? For those flights that must go over that horizon, couldn't they fail over to a lower freq / lower bandwidth channel, when needed; while still keeping near real-time status updates?
Or connect via another aircraft who is further behind them but on a similar flight plan.
I had intended to use that to support my notion that aircraft tend to stay near land. I've flown that route several times and each time we flew an arc that kept us near or overland. Like this, but never in Russian airspace. http://flightaware.com/live/flight/CCA985/history/20140309/0...
Man, these guys are all over the place. I may have to reform my ideas about flight paths.
Well, for the ones that travel within sight of land, we don't have much of a problem. And that may be "most". But "most" flights don't end in disaster. It is the ones that spend long hours over the horizon that are of the most concern.
A significant number spend lots of time more than 200 miles from land. And "land" here means some land with a data link endpoint.
With respect to airplanes forming some sort of mesh network, consider the bandwidth requirements throughout this link.
>With respect to airplanes forming some sort of mesh network, consider the bandwidth requirements throughout this link.
Right, but all that is really necessary is groundspeed, altitude, and equipment number. That gets you enough information to find a crash / forced landing site quickly, and to know where to search for wreckage. So, even for those parts of the flight path that are "dark" It still may be potentially superior to low-bandwidth satellite links.
You can't just "fail over to a lower freq". Propagation beyond the radio horizon at all is at the whim of the ionosphere, which is the only reason long-distance radio works in the first place (it reflects back down, when the conditions are right)
And what I am saying is that in bad HF conditions, it won't hear anything, and thus won't say anything. Putting the likelihood of a missed communication higher.
Keep in mind that over-the-horizon radio communication is subject to the vagaries of the ionosphere and is not totally reliable.
Non-line-of-sight radio communications are short wave (well, also medium wave and long wave, but those modes require significant power). If ionospheric conditions are bad, as they are at many times of day (worse at night for some bands), at many times during the sunspot cycle, and during a solar storm, and sometimes during atmospheric events, then you won't get a radio signal through. One famous example of this was just prior to the Pearl Harbor attack, communications via radio between US and Hawaii were very poor or nonexistent for key parts of the time.
So to put a number to your question "likely better than half the time".
>So to put a number to your question "likely better than half the time".
I'm not sure how you get from 'because communications between arbitrary point A and and arbitrary point B are sometimes impossible on band C' that "likely better than half the time" communication will be impossible on every useful band to any useful place.
Bandwidth: quick googling shows that a flight data recorder records between 64 and 256 12-bit words a second, i.e. 1-3 kbps. Browsing http://en.wikipedia.org/wiki/Bit_rate shows that another 2-8 kbps is needed for voice recording (from one microphone). Let's call it 10 kpbs all together.
Number of simultaneous connections: there are around 20,000 (ballpark) airliners in the air at the peak time each day.
Compare this to the rather old Iridium satellite network: 2400 (uplink) baud rate, 1100 calls simultaneously per satellite (of which there are 68). Not quite enough, since airplanes are not evenly distributed over the Earth or under the satellites, but not outrageously far from our requirements.
Cost: several billion dollars (under 10, I think) have been invested into building the Iridium network, and their operating expenses are around 100 million per year (they are a public company).
Clearly, it would not be unrealistically expensive, in the world of 500 million dollar airliners, to build and maintain a network of satellites to record all black-box voice and data in real time.