I think the original question had to do with how fast
one could send messages...anyway, that's where we are now...
I forgot about the 'throttling' I needed in my radio tests,
which bear out your assertion that messages get lost if you
go as fast as possible. I throttled on the PC side with a
command/response mechanism, so my minimum was 10ms. I never
looked any deeper, why do you think the receiver is the bottleneck?
MS
Jeongyeup Paek wrote:
Was Kishore's original question about
"how fast can you send packets?" ? or
"how fast can you trigger timer?" ?
For the Timer, as Michael said, 3ms is the min.
Otherwise you 'might' lose a timer.fired event.
For the packet rate, it depends on the pkt size.
But one thing I know:
although sending immediately after sendDone events
without any timer is fastest logically,
it loses far more packets than slowing it down a little.
If the sender sends packets back-to-back,
the receiver is not fast enough to process them and drops some packets.
For example, in one of my experience (I forgot the pkt size),
sending every 9ms achieved far greater goodput than
sending back-to-back.
Also, if you do it with Timer firing every say 3ms,
then at least some of calls to Send.send will return FAIL
b/c it is processing the previous pkt. So your actual
sending interval can be 6 or 9 or 12...
So... you need to figure out your self for you pkt size.
Thanks
- jpaek
Michael Schippling wrote:
The TimerC period is in "binary" milli-secs, I think 1024/second,
which comes out to about 1.4 sec/min skew...using that you should
be able to do accurate time measurements within the unpredictability
of interrupt interleaving, etc...
For T1 versions above .10, the min Timer count is 3. I'm not clear
why that was hacked in at that point, but probably something to do
with the micaz...
To send messages as fast as things can go, skip the timer thing and
just start a new message from the sendDone() of the previous.
MS
Tao Wu wrote:
According to TinyOS1.x TimerC and ClockC' s implementation,
for a repeated timer, the min Timer interval is about 2 or 3 ms.
And my experiment measure shows that for 1000 ms interval,
there are about 1 sec bias per minute.
If you want to do it more accurately, try to use ISR directly.
Tao
On Nov 15, 2007 9:10 AM, RamKishore <[EMAIL PROTECTED]> wrote:
Hi all,
I fire timer at 100msec.For each firing, I transmit a packet.To
increase the Rate of transmission, Timer's time-period can be
decreased.what factors should be included in estimating the timer's
time
period? and Whats the min. Time period that can be achieved?
Regards,
Kishore
_______________________________________________
Tinyos-help mailing list
Tinyos-help@Millennium.Berkeley.EDU
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
_______________________________________________
Tinyos-help mailing list
Tinyos-help@Millennium.Berkeley.EDU
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
_______________________________________________
Tinyos-help mailing list
Tinyos-help@Millennium.Berkeley.EDU
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
_______________________________________________
Tinyos-help mailing list
Tinyos-help@Millennium.Berkeley.EDU
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
