I worked in the cellphone industry with wireless data 3 years ago, for a service provider. Back then, the critical numbers on the "voice side" were:
10 Mhz total bandwidth cap, per FCC
100-500 sectors per metro area * 5-20 radios per sector = ~4000 radios per metro area
19 kHz per radio (analog, dedicated one radio per phone conversation)
typical "frequency reuse" yield: 10-to-1
average usage per radio per day: 3 hrs
$0.10 cents per minute average charge => 0.10x60x3x4000 = $72000 gross revenue per day per metro area
Actually I'm not sure that revenue figure looks right; I was on the engineering side so I didn't pay much attention to the financials.
Anyways, we had a wireless data network which used some of of the capacity of the voice network for data. It was totally hopeless. We could provide 14.2kbps service, but because of the financial reality of the company, we had to charge somewhere close to $0.10 per minute to justify the borrowed capacity.
10 cents a minute for a lousy 14.4 connection is ridiculous. That means for a lousy 1MB porn MPEG you'll have to wait 10 minutes and pay over a dollar. Way too pricey for almost anyone except those who REALLY NEEDED mobile data access and were willing to pay through the nose for it! We kept wanting to price it lower, since it cost us very little to maintain, but the thing was that for every single kilohertz we gave to a "data" customer, well there was a "voice" customer perfectly willing to pay 10 cents a minute for that same bandwidth.
A ten-minute 20khz cellphone call is worth a dollar to many people. A one-megabyte ten-minute data transmission is worth a dollar to only a very few people.
We tried to move to a packet-switched service "CDPD" which would frequency-hop to avoid actually using up any voice-side dollars. But that was quite expensive in terms of equipment. And even as we approached viability for CDPD, the voice side switched from analog to CDMA digital, which makes more efficient use of the voice-side bandwidth, squeezing the opportunity cost of providing data service up even higher.
The numbers today are not that different. "PCS" bandwidth has allowed carriers to increase their total bandwidth by about 4x. (note: the fact that this bandwidth is in the Ghz part of the spectrum is of fairly immaterial). Conversations are now compressed and consume about 4x less bandwidth. Prices have probably dropped about 2x-4x from where they were at 3 yrs ago. BUT it's still true that a voice transmission is "worth" more per kHz than a data transmission, to most people.
I think high-speed mobile data access is a great idea. Kudos to the Sprint team for demonstrating such a high bandwidth connection on a cellphone. But the MONEY AIN'T THERE. No cellphone service provider in their right mind is going to stream 300kbit video to one person when they could be streaming 3kbit voice to 100 people.
At some point, total network bandwidth will become so huge that it swallows up the entire voice market. After all there are only 6 billion of us on the planet, so there are only 3 billion cell phone calls to be served:-).
Once there's room left over from that market, we'll start seeing wireless data prices that are "worth it" to many people. I'd give it about 10 more years though. You'll know it's coming when everybody you know has a cellphone and considers it "real cheap". It will happen in Europe and Asia before it happens here in the States, thanks to the vast sluggish power of the cell phone industry here.
---
One thing that could make wireless data cheaper than voice faster is its non-realtime nature. Typical data can survive latency much better than typical voice. Networks which are build from the ground up to take advantage of this fact (large buffers throughout the topology?) may become more practical sooner. But the cellphone industry won't be providing these. Some other industry will.
This should do a good job of a lot of the things that JPEG (even at high-quality q) is currently failing on for us: photographs that contain a lot of bluescreen / greenscreen; pencil / pen artwork, etc. With JPEG, I get nasty haloes around foreground elements; these are very visible on a flat background such bluescreen or paper. JPEG also forces a mottled effect on the bluescreen due to color quantization I believe. DjVu appears to reduce or eliminate these undesireable side effects. I will be following its progress with interest (as long as it doesn't get the kiss of death from proprietary interests...)
Slashdot Mirror
10 Mhz total bandwidth cap, per FCC
100-500 sectors per metro area * 5-20 radios per sector = ~4000 radios per metro area
19 kHz per radio (analog, dedicated one radio per phone conversation)
typical "frequency reuse" yield: 10-to-1
average usage per radio per day: 3 hrs
$0.10 cents per minute average charge => 0.10x60x3x4000 = $72000 gross revenue per day per metro area
Actually I'm not sure that revenue figure looks right; I was on the engineering side so I didn't pay much attention to the financials.
Anyways, we had a wireless data network which used some of of the capacity of the voice network for data. It was totally hopeless. We could provide 14.2kbps service, but because of the financial reality of the company, we had to charge somewhere close to $0.10 per minute to justify the borrowed capacity.
10 cents a minute for a lousy 14.4 connection is ridiculous. That means for a lousy 1MB porn MPEG you'll have to wait 10 minutes and pay over a dollar. Way too pricey for almost anyone except those who REALLY NEEDED mobile data access and were willing to pay through the nose for it! We kept wanting to price it lower, since it cost us very little to maintain, but the thing was that for every single kilohertz we gave to a "data" customer, well there was a "voice" customer perfectly willing to pay 10 cents a minute for that same bandwidth.
A ten-minute 20khz cellphone call is worth a dollar to many people. A one-megabyte ten-minute data transmission is worth a dollar to only a very few people.
We tried to move to a packet-switched service "CDPD" which would frequency-hop to avoid actually using up any voice-side dollars. But that was quite expensive in terms of equipment. And even as we approached viability for CDPD, the voice side switched from analog to CDMA digital, which makes more efficient use of the voice-side bandwidth, squeezing the opportunity cost of providing data service up even higher.
The numbers today are not that different. "PCS" bandwidth has allowed carriers to increase their total bandwidth by about 4x. (note: the fact that this bandwidth is in the Ghz part of the spectrum is of fairly immaterial). Conversations are now compressed and consume about 4x less bandwidth. Prices have probably dropped about 2x-4x from where they were at 3 yrs ago. BUT it's still true that a voice transmission is "worth" more per kHz than a data transmission, to most people.
I think high-speed mobile data access is a great idea. Kudos to the Sprint team for demonstrating such a high bandwidth connection on a cellphone. But the MONEY AIN'T THERE. No cellphone service provider in their right mind is going to stream 300kbit video to one person when they could be streaming 3kbit voice to 100 people.
At some point, total network bandwidth will become so huge that it swallows up the entire voice market. After all there are only 6 billion of us on the planet, so there are only 3 billion cell phone calls to be served :-).
Once there's room left over from that market, we'll start seeing wireless data prices that are "worth it" to many people. I'd give it about 10 more years though. You'll know it's coming when everybody you know has a cellphone and considers it "real cheap". It will happen in Europe and Asia before it happens here in the States, thanks to the vast sluggish power of the cell phone industry here.
---
One thing that could make wireless data cheaper than voice faster is its non-realtime nature. Typical data can survive latency much better than typical voice. Networks which are build from the ground up to take advantage of this fact (large buffers throughout the topology?) may become more practical sooner. But the cellphone industry won't be providing these. Some other industry will.
This should do a good job of a lot of the things that JPEG (even at high-quality q) is currently failing on for us: photographs that contain a lot of bluescreen / greenscreen; pencil / pen artwork, etc. With JPEG, I get nasty haloes around foreground elements; these are very visible on a flat background such bluescreen or paper. JPEG also forces a mottled effect on the bluescreen due to color quantization I believe. DjVu appears to reduce or eliminate these undesireable side effects. I will be following its progress with interest (as long as it doesn't get the kiss of death from proprietary interests...)