[time-nuts] Sound cards

[Hal Murray]

> The optical link commonly being used for S/P-DIF is TosLink and it
> seems  like it can be the cause of many problems. It seems like some
> care in  doing the optical link setup is needed. I have never digged
> into why the  optical links have that problem. I can only guess, but
> bad optical  coupling seems reasonable. The multimode "fiber" seems to
> be leaving one  or two things to ask for. 

It's been a while since I did any serious work with fibers.

There are 2 limitations.

One is signal to noise.  You have to get enough light in the transmit end so 
that after attenuation there will be enough coming out for the receiver to be 
able to find the bits.  Attenuatiion is linear with length with a constant 
for getting in and out of the fiber.  Add some more for splices/connectors.

The other is dispersion.  If you have a multi-mode fiber, some of the photons 
bounce around more than others which results in a longer path and increased 
transit time.  Simple geometry is a good approximation.  The net result is 
that the photons get smeared in time.  If your pulses are too narrow (bitrate 
too high), the smearing will cause adjacent bits to overlap and you can't 
easily sort things out at the receiver.

Single mode fibers don't have modal dispersion.  But they do have chromatic 
dispersion.  Long distance telco links use very narrow bandwidth lasers.

One characteristic of dispersion is that there is a trade-off between 
distance and bandwidth.  Fibers have ratings in megabit-miles.  Typical 
multi-mode fibers were 300-500 megabit-miles.

Single mode fibers are roughly 7-9 microns dia for the active region.  
Multi-mode fibers were 50 or 62.5 microns.

Roughly 10 years ago, there was a sweet spot at 155 megabits (OC-3) and 2 km 
using LEDs for the transmitter and multi-mode fibers.  Since then, they are 
using low cost lasers (from CDs) so things have changed.  If you wanted 
faster or farther, you used a laser and single mode fibers.

The engineering/specsmanship on the overall link was super conservative.  It 
was essentially impossible to measure the error rate.  The trick is to insert 
enough attenuation so you get enough errors to measure, then compute what you 
would get without the attenuation.

I haven't worked with plastic fibers.  I'd expect the engineering to be 
conservative so it should just work.  If it doesn't the obvious problems are 
dirt/mud at the connectors or cracked/broken fibers.  (I'm assuming a sane 
length.)

One disadvantage of conservative engineering is that a system that's broken 
might actually work well enough to act like a flaky system.  I'm thinking of 
something like a broken fiber that is sometimes held in place close-enough by 
the jacket.




-- 
These are my opinions, not necessarily my employer's.  I hate spam.