Carranza, the GPRS "caveat" is those sloppy GPRS networks
some deliver, always almost good enough, every month
sure to be fixed next month. (until the year runs out)
Not the first time, nothing new, but the way some company
cultures are.
HSCSD bypasses all of that, the all-digital channel is there
for those who connect GSM to GSM calls, part of
international GSM roaming. (the bits won't be converted
to analog speech at some weird point, will reach
the destination correctly, in most cases)
However, the 2-3-4 bitstreams might be hooked up,
connected slightly differently, different delays,
if some stuff is not "HSCSD aware", so it must be the
terminals who figure out when to ask for a re-transmit
(whaddyasay error correction??)
Note that voice is OK with some errors here and there,
not necessarily in the airintereface, but because of
sloppy stuff inbetween base stations, still OK for
voice.
GSM (voice) obviously has channel coding to cope with
deep fading,etc, but if the terminal is fairly
pedestrian (like the good doctor) this is not needed.
(additionally the gaussian FSK has some inbuilt
eror correction, channel coding, compared to QAM, ask
Trellis-Viterbi and FSK-Wozencraft)
I'm sure you understand that voice cannot tolerate
varying transmission delays, needs to use convolutional
channel codes, but "data" does not need those guaranteed
delays, can use full raw data speeds using that old
method of whaddyasay when garbage is sometimes received.
(without getting into the grey area where neither one is
good enough, like with regular, mobile CDMA)
One example I often use for educational purposes is
this interesting way of transmitting data using the
tail of meteorites as a reflector.
When that meteorite falls, one can transmit Gbytes of
data for a split second, inbetween nothing.
Not good for full duplex voice transmission, but
great for (military) data, as long as a star falls
every now and then.
Additionally extremely good for understanding the
extremes, meaning of channel coding, interesting case to
have to wait for the next meteorite to be able to say
which data didn't get through, packets still
missing, numbering the missing ones,etc (protocols)
Another good case, including redundancy, is the wife talking
in the kitchen while trying to catch to the news and the
kids are running around the house.
Ilmarinen
That is, a "raw channel" might transmit just garbage
when on the limits, and even if asking for a
retransmit (whaddyasay) it will still be just
the same garbage (static for old radio guys).
Channel coding, or forward error correction, is
used to lift that signal above the "statics"
by kind of switching to a lower bitrate.
(this big shannon thing, almost)
Backward <g> error corrections is adding some checksums
to the data, the receiver can find out if the
(whole block, packet of) data is correct or not.
If OK, no problem.
If not correct, ask the other guy to repeat what he said.
The fear of any system like this, in marriage leading to
divorce, is wasting more time on claiming what was said
correctly and what wasn't (this part of telecom 101 is
mostly targeted on more mature participants)
The beauty of GSM (gaussian FSK) is that users,channels
are fairly isolated from each other, as well as
those in neighboring-neighboring cells on the same
channels.
That is, channel error rates can be fairly well predicted,
are isolated, independent, especially for pedestrian users
for GSM, not so for CDMA.
For fast moving users it gets more complicated,
turns more into that meteorite example, blast
data when data really goes through, don'twaste time when not possible, change fast between high speed, capacity
trucks and slow tractors which eventually will arrive,etc...
Ilmarinen
That is, this thing which is finally hitting the
academic fan, that
- shannon capacity limit can be achieved with infinity
delay in receiving the data. Ie. most of the shannon data
hasn't arrived yet, few will be around when it does.
- considering both latency (delay) limits and
channel limits is still to difficult for regular
academics, but regular folks have always found solutions
for this academic problem (whaddyasaid) |
