[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: telco's vs x-phones



At 10:32 PM 8/21/96 -0400, John Brothers wrote:
>
>I think you both have good points here.  I will redo the math.
>
>At 09:19 AM 8/21/96 -0700,  Brian D Williams <[email protected]> wrote:
>>In a reply to Vipul Ved Prakash, Jim Bell wrote:
>>Siecore plain vanilla 36 fiber singlemode list $ 1.82 a foot, $5.46
>>a yard.  
>
>Jim said $10/meter, so about 50% of his estimate.
>
>
>>
>>>Each fiber pair should be able to handle approximately 1 million
>>>conversations at current data rates, or a total of 18 million
>>>conversations for that 18-pair cable, or 9.5 trillion 
>>>conversation-minutes.
>>
>>At current data rates (OC-48 Sonet) 32,256 voice channels per
>>fiber, 580,608 total for the fiber.
>
>Ok.. I think we're talking apples and oranges.  Jim uses the term 
conversation,
>which I assume is a 3 minute duration of time.  Your OC-48 handles almost
>precisely 1 million conversations per day.   (32k DS0 x 24 * 60 * 20)

My error was being off a deciminal point.  I was thinking around 6.4 
gigabits per second, which works out to 100,000 channels at 64,000 bits per 
second per channel, not 1 million.  

However, even my selection of 6.4 gigabits per second is arbitrary:  I have 
seen at least two ads for 80 gigabit (8 channels/fiber at 10 gigabits per 
channel) for fiber optic systems.  But my goal is not to make fiber look 
unrealistically good, but to generate a ballpark figure.


>>>Multiply this cost by 10 for right of way, trenching, repeaters,
>>>and other auxiliary hardware, or $100 per meter.  This is probably
>>>just a ballpark estimate, but...
>>
>>Off by a factor of at least 10 not counting switching equipment.
>
>So, assume that he's a long haul carrier, and doesn't have to buy any
>switching equipment.   Still, based on  $500/meter, that means 
>that this network cost about 2.5 billion to build.

Even long-haul carriers must have SOME switching equipment.  But as you 
note, my point is that the difference between local and LD costs is simply 
the fact of the insertion of some arbitrarily large amount of fiber trunk 
between the local switches.  LD companies still deserve to be paid, but the 
AMOUNT that they are paid maybe should drop to an arbitrarily low level as 
their costs drop.


>So if the system was running at full capacity, 24/7/365, the LD will have
>to charge $0.50/conversation, or $0.15/minute to make 2.5 billion per year.
>
>Even reaching 10% of that capacity is unlikely at $5/conversation.
>
>Which, clearly, isn't happening.  I assume that either I have made a 
>mistake in my calculations, or the costs of laying fiber has been 
>grossly overestimated...


Yes, I think the costs laying fiber are still a bit opaque.  The main reason 
one of the first alternative telephone LD services (Sprint) was that it was 
the creature of Southern Pacific Railways, who (obviously) owned a large 
amount of (ta da!) RIGHT-OF-WAY. And, more particularly, right of way that 
was well-connected, easily accessible, and easily diggable to install the 
fiber.  They decided they were going to put in a fiber communications system 
for the railroad, and at some (early?) point realized that since the cost of 
the trenching and fiber is basically fixed, so turning it into an LD service 
was going to be relatively straightforward.


>Or, that since no-one is going around and buying
>up 5000km right of ways, Sprint, MCI, AT&T and co. can lay fiber
>much more cheaply.  If we assume $50/meter, it will cost 250 million, and they
>will have to charge $0.015/minute to break even at 100% capacity, or 
>$0.15/minute at 10% capacity - and we know Sprint charges less than that,
>but if they can get more than 10% of the capacity --- although it is doubtful.

(I think you still need to go back and re-check some of your figures.  Yes, 
I did make that decimal point error, but I think you've added a few 
(different) errors of your own.) 

Now, keep in mind that a little perusing of these numbers (even if they are 
not exact) reveals an few interesting facts:

The cost of doing the right-of-way/trenching is essentially independant of 
the number of fibers that you intend to lay.  Moreover, the cost of the 
fiber itself is very low.  In other words, everything being equal, you want 
to pack as many fibers into a trench as you can.  Add to this, the fact that 
you can provide a way for fiber in a buried conduit, and later (years?) blow 
in new fiber.  The line I saw installed, "up close and personal," had three 
separate ways for cable, and the one cable being installed immediately was 
36 fibers.  Had two similar fibers been installed in the other ways, that's 
108 fibers, and at 2.4 gigabits per second, that's about 54x40,000 
simultaneous bidirectional conversations, or 2.16 million.  There are 
probably very few communications corridors that need more capacity than this.

 The obvious limit of this is that there is only a fairly limited demand for 
bit transmission (at least, compared with the capacities possible with 
fiber).  In addition is the fact that once you have modern-quality fiber 
installed, you can increase capacity by changing the optics/electronics.  It 
wasn't too many years ago that 400 (?) megabits/second or so was a fairly 
standard rate.  2.4 gigabits/sec supplanted it a few years ago, and speeds 
substantially greater than this are straightforward today.  AT+T claims to 
supply fiber now, and electronics/optics within a few months, to allow 80 
gigabits per fiber, or about 32 times more than 2.4.  

This should immediately explain why you aren't seeing a lot of trenching:  
Except for the natural desire among engineers to duplicate trenches along 
major routes to protect against "trencher fade" (accidental cuts), it won't 
really pay to put in more than a couple trenched routes between most major 
cities, if they're separated by "boonies."  It's far more economical for 
multiple companies to share a single route, or a pair of routes.  (there is 
an obvious exception if the whole area is developed, such as extended 
corridors like Washington/New York/Boston.)


>My best assumption at this point is that we don't have to make back the entire
>cost of the fiber per year to make a profit.  if we only have to earn 20%
>of its value to make a 10% profit, it would bring the price down to
>$0.075/minute, and if we were to get OC-768 systems invented, we could reduce
>that to $0.005/minute, which is approaching the values that Jim derived.

The value of an investment is compared to the average interest rate paid.  
However, that's probably too conservative with a fast-changing market like 
telecom today.  Even so, if a system paid back profit at 10% of the 
investment per year, it would probably be considered a very profitable 
investment, and I agree with your figure of 20% before fixed costs.

>So, I hate to say it, but it looks like long distance is more expensive
>than Jim (or I) thought.  But, in a few years (say, 5) it will approach
>ridiculously cheap.  We can expect at that point that either
>a) All (or most) long distance will be over the Internet
>b) Internet phone will vanish as regular long distance becomes a flat fee
>   like local service.

Probably closer to the latter.

The main lesson from these calculations is that fiber's costs are almost 
entirely based on their one-time installation cost, not a continuing 
maintenance cost.  True, the installation must be paid for, but that 
translates into an interest charge that is very low.

In view of the fact that the billing for telephone is supposedly half of  
the phoneco's cost, it will make sense that all telephone service will turn 
into a flat monthly fee, "enforced" by the possibility of Internet telephone 
for the few recalcitrant companies/countries who resist the transition.



Jim Bell
[email protected]