Pure Music to DAC - again
Audio Empire wrote:
On Fri, 2 Sep 2011 06:46:53 -0700, Dick Pierce wrote
Audio Empire wrote:
At some length, fiber requires in-line repeaters, but 30 feet is NOT
that length. Fiber. like everthing else, is lossy. Light leaks out the
sides
of the cable, there are internal reflections that can cancel or otherwise
compromise signal integrity,
Actually, in media like Toslink, it is the internal
reflections that make it work by reducing the leakage
out the sides.
Those are not the kinds of reflections I was talking about.
What kind were you talking about? End-to-end?
This introduces the same non-problem that reflections
in a cable present. Remember, stuff's happening at
the speed of light in the medium, and that's no picking
of daisies.
Let's do a little gedanke: say the ends are really badly
terminated, such that you have a 50% reflection from
the cable end. And assume you need to drop the level
60 dB to drop below the level you're generating receiving
problems. And let's further assume a 10 ft cable.
Each round trip looses 12 dB: 6 dB for each 50% loss in
reflected intensity. So you have 5 round trips, which,
assuming an RI of 1.62, and thus a propogation velocity
of 607,000,000 odd ft/sec, it takes 98 nanoseconds for
the reflected energy to drop below the 60 dB threshold.
In fact, assuming a badly terminated fiber with pure
air gap, worst-case Fresnel reflection is on the order
of 15%:
R = ((RIf -Ria) / (Rif + Ria))^2
So, that represents 16 dB per end or a 32 dB round-trip
loss, which means 2 round trips and we're below our
arbitrary 60 dB threshold, under 40 nanosecods.
And, in fact, the threshold is substantially more
generous than that.
If not end-to-end, what kind of reflections were you
talking about?
Mainly optical's strengths are much wider bandwidth (an optical
signal can carry much more information than wire without nearly
as much loss because light is a much higher frequency than an
electrical signal).
Well, elements of this are true, but, as a whole, it's quite
incorrect. The fact that the transmission medium uses a carrier
with a very high frequency (light) is completely irrelevant to
the system's bandwidth. The fact is the bandwidth limit is NOT
set by the freuqency of the light: is is, in this case, set by
the bandwidth of the transducers at each end of the link:
the transmitter and receiver. Their bandwidth is FAR less than
what the intervening cable might or moght not support. We could
double the frequency of the light: go into the near UV as
opposed to the near IR, and the bandwidth of the system will
not change on iota unless the transducers' bandwidth changes,
and that's almost totally independent on the frequency of the
carrier.
I said that in reality, the carrier frequency advantage
was inconsequential in this case.
So it's a true in some case but orrelevant here sort of
thing?
Indeed, one of the big problems with Toslink is NOT the optical
cable, but the crappy electro-optics at either end: unsymmetrical
thrsehold hysteresis, rise and fall times and more in the detectors
can lead to all sorts of problems, effectively reducing the bandwidth
to far less than what might otherwise be obtained.
Very true, but again, at audio sampling frequencies, probably not of any real
consequence.
Actually, for those DACs that depend heavily upon timing
accuracy i the individual bits to recover the sample
clock (an all-around BAD design which, unfortunately,
a number of high-end companies gleefully implemented),
it has a VERY real and VERY significant consequence,
and is one of the few verifiable causes of very large
amounts of sample jitter.
To increase the accuracy of the sattement, I might
be inclined to have said, "In equipment designed
competently to properly manage sample output clocking,
at audio sampling frequencies, probably not of any real
consequence," but this makes a lot of assumptions about
high end audio designs which are not supportable in a
practical sort of way.
It's also curious to note that those modules they designate
for digital audio purposes, despite running at a wavelength
650nm (equivalent to a free-air frequency of 461 terahertz),
the specificied data rate is only 15 Mb/s: we were spec'ing
pulse transformers and transmitter/receiver pairs for our
project that had data rates well over an order of magnitude
higher, and this was routine. This sinmply demonstrates that,
in practice, Toslink DOES NOT have anything like a wider
bandwidth than electrical transmission.
That's probably correct. My comments were about optical in
general, not TOSLINK specifically.
But the properties of things like single-mode dark fiber
are so totally removed from those of Toslink, even though
they operate under the same physical principles, whatever
is true of single-mode fiber is completely irrelevant
for Toslink. Single mode 9 micron fiber can go 60
kilometers witgout a repeater.
I have no experience
with TOSLINK other than as a user,
but in the Polaris Trident project, we used glass fiber interconnects to do
virtually all of the rocket guidance and internal navigation communications.
THat's the difference between 9 uM glass and the soda straws
used in Toslink. TOslink is cheap but at the cost of low
bandwidth and short distance. Single-mode glass can run 40 Gb/s
over kilometers, audio toslink is limited to 15 Mbs for runs
of a few meters at best.
We were able to replace literally over half a ton of mil-spec cabling with
several light, thin strands of glass optical cabling carrying hundreds of
different digital signals consisting of everything from audio frequencies
for the in-engine vectoring fins to near microwave for radar and guidance
control signals, all at the same time.
No doubt. And had you done it with Toslink, I doubt it
would even boot up.
--
+--------------------------------+
+ Dick Pierce |
+ Professional Audio Development |
+--------------------------------+
|