On Thu, 27 Feb 2003 19:18:59 -0000, Tony Raven <
[email protected]> wrote:
> Tim Woodall <
[email protected]> wrote:
>>
>> Round trip (one way) is roughly 1/4 second, going up to about 1/3 second when you include the
>> delays in the transmitter and receiver.
>>
>
> Best case round trip for a geostationary satellite is 240ms one way (up to satellite and back down
> = 44,600 miles @ 186,000 miles per second) Round trip is 480ms. Normal latency accounting for the
> angle factor since the satellite is not directly overhead you (usually on the equator) and the
> groundstation plus the electronic and ground distances generally make the delay 2-4x longer. So
> latency for sending a data packet and receiving the response is typically 1-2 seconds on
> bi-directional satellite.
>
I would be surprised at 1-2 seconds. I have sat in a satellite TV control room with a direct feed
from the studio and a direct feed from the transmitter with one bounce in between and the delay was
near enough
1/3 second. (I'll accept 1/2 second as I could only see the screens, I had no way of freezing them
so I could read the timecode properly but I would have said that there were no more than 10 frames
between them at 25fps. Interestingly the delays on digital cable TV weren't much different
although this must presumably have been encoding/decoding overheads)
If the delay is 2 seconds for satellite data transmissions (and I have never used a satellite link
for this) then I would think at least half the delay is due to encoding and decoding stuff at the
transmitter and receiver and maybe the satellite as well. Presumably the return link is transmitted
at much lower powers so the satellite will probably have to apply error correction before
retransmitting to the ground link rather than the satellite just blindly retransmitting everything.
If you are sending multiple packets then TCP doesn't require a reponse to every packet and the
sender doesn't have to wait for each packet to be acknowledged before sending again.
On something like VNC I would swap latency for bandwidth for almost any connection, certainly
anything less than 100Mb/s[1]. OK I don't want to play games across VNC and I can accept that what
you are seeing being a second or more behind what the computer is doing can be tricky but you can
easily cope. Waiting 5-10 seconds while the screen repaints itself is much more annoying.
[2] Stupid latencies excepted. I don't care what bandwidth a carrier pigeon can manage, an hours
latency isn't practical for anyone
I have programmed interactive games for use on live satellite TV and you can cope with one bounce
easily, two bounces comfortably and three bounces with great care and planning. (The one problem we
did have was convincing the producers that you couldn't play one person from home against a
celebrity in the studio. In theory you can simulate the delay in the studio but it doesn't seem to
work well in practice.)
(Incidently, I suspect you actually know this but it's not apparent from your comment above, for a
geostationary satellite it has to be over the equator. geosynchronous satellites will draw a figure
of 8 in the sky, the lobes depending on the precise orbit - geostationary being the special case
where the lobes collapse completely. The correction for the extra distance due to the satellite not
being overhead is fairly small as well - I'm not sure if the 22,000 miles is measured from the
surface or the centre of the Earth and I can't be bothered to work it out but even in the worst case
the round trip is only 8,000 miles further or about 50ms )
Regards,
Tim.
--
God said, "div D = rho, div B = 0, curl E = - @B/@t, curl H = J + @D/@t," and there was light.
http://tjw.hn.org/ http://www.locofungus.btinternet.co.uk/