I did some test lately as well. I checked rather influence of layers on transmission speed and I used 32kHz timer too, but now I am interested in more precise time testing, ie. CPU cycles. However it's hard for me to go deep into TinyOs to get the lowest layer - very close to physical. I gave up and I think other OS like LiteOS (without all that abstraction) is better for testing device.
2011/3/29 <prasannakart...@cse.iitb.ac.in> > hi, > I did some measurements recently. I quantified the time taken by the > different components. Here are my results: > > before Amsend.send() statement 64420 > Send.send() 64426 > send() 64427 > AcquireSpiresource() 64431 > before loadtxfifo() statement 64432 > Loadtxfifo() 64433 > before TXFIFO.write() statement 64440 > after TXFIFO.write() statement 64446 > TXFIFO.writedone() 64519 > AttemptSend() 64521 > CaptureSFD.capture() 64655 > ReleaseSpiresources() 64538 > after loadtxfifo() statement 64447 > after Amsend.send statement 64448 > Amsend.Senddone() 64662 > > All time values are based on 32Khz timer. CCA has been disabled. Also, DMA > has been enabled. This is for 100 Byte packets on telosb platform. > Tx time : 135 ticks => 4.2 msec > Spi copy time without s/w overheads: 73 ticks => ~2.5 msec > Hope this helps. > > I would like to know how you are measuring the time values. If you are > using printf's(like i did a few days back), then the values are likely > blown up. > > > Hi > > > > Does it happen in every packet transmission? > > If no, maybe some packets are retransmitted by PacketLink Layer? > > > > 2011/3/28 蔡沅峰 <starry1...@yahoo.com.tw> > > > >> Dear all, > >> > >> > >> We have a high-data-rate application and we need to decrease the > >> transmission delay (between send.send and send.sendDone). > >> > >> The payload of our packet is 80 bytes. ( Total packet size is about 100 > >> bytes in Tinyos ) > >> > >> We've adjusted the initial backoff time defined in CC2420csmaP.nc to the > >> range of 10~60 ticks (under 32khz clock) > >> > >> The above two added is about 100*8/250(kbits/sec) + 35/32768 = 4.27ms > >> > >> But when we measure the time between send.send and send.sendDone, the > >> result is 7.82ms, much bigger than 4.27ms > >> > >> Even though we take the congestion backoff time into account( 10~80 > >> ticks > >> ----> 45/32768 = 1.37ms ), the result is still less 7.82ms > >> > >> But the inteval between our testing data packest is 3 seconds, the > >> congestion backoff time shouldn't exist. > >> > >> Our application is broadcast, so we don't consider ACK wait time. > >> > >> Does anyone know why the larger transmission delay in our experiment? > >> > >> > >> We also record the time when the packet is send out ( by the SFD > >> interrupt > >> ), the time between this SFD time and send.sendDone is about 1.99ms > >> > >> What confuse us is this time is much less than the data transmission > >> needed > >> (4.27ms), does anyone know why? > >> > >> > >> Many thanks!! > >> > >> > >> Best regards, > >> > >> > >> Sean > >> > >> _______________________________________________ > >> Tinyos-help mailing list > >> Tinyos-help@millennium.berkeley.edu > >> > https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > >> > > > > > > > > -- > > Pozdrawiam, > > Damian Rusinek. > > _______________________________________________ > > Tinyos-help mailing list > > Tinyos-help@millennium.berkeley.edu > > https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > > > -- Pozdrawiam, Damian Rusinek.
_______________________________________________ Tinyos-help mailing list Tinyos-help@millennium.berkeley.edu https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
