CCS/LAUNCHXL-CC1312R1: 6lowpan TI code example - forever loop from probably wrong version of ROM image

Hi Fredrik,

TI sent me pre-released example of 6lowpan, and i managed to setup 6lowpan network.

my setup includes: 1 Poll Root and 2 poll End-hosts.

The problems:
1. for some reason it takes more than 1 min for the EV Boards to start communicate, and to discover network changes.
2. I've added a new socket between the end-nodes and sending to fixed ipv6 address of the nodes , but for some reason i'm unable to communicate between end-nodes until Root start send them messages.
3. is it possible this problem exist cause i used the IP address the ROOT uses? where can i find the local fixed ipv6 address of the hosts?

do you have an idea how to make it faster? or what is the reason for such delay?

Thank you,
vera

Thank you Wonsoo,

1. i've changed the SCAN_TIMEOUT_VALUE to 10000, but it didn't helped me, i think it made the root node to stop sending keep alive messages to one of my nodes.

2. what about the TRICKLE_TIMEOUT_VALUE  ? if i change it will it increase the discovery of new nodes?

3. i have continuous problem regarding post to mailbox when i get "FALSE" response while posting, is there a proper way to change the mailbox size? 

from app.cfg:

/*
  =================================
  network stack mailbox
  =================================
*/
var mailboxTcpipParams = new Mailbox.Params();
mailboxTcpipParams.instance.name = "tcpip_mailbox";
Program.global.tcpip_mailbox = Mailbox.create(16, 20, mailboxTcpipParams);

 this is the configuration, and when i change it the device get in to cont loop. 

typedef struct __TCPIP_cmdmsg
{
    uint8_t event_type;
    uint8_t index;
    void *data;
} TCPIP_cmdmsg_t;

i guess the size 20 is regarding to tne msg size. is there any description regarding this mailbox function used in app.cfg?

4. another question regarding to the UDP buffer size UIP_CONF_BUFFER_SIZE - for some reason it configured to 300B , from the notes it contain udp header(8B) and the rest is for the upd payload. Q: is it posible to increse this buffer? when i'm changing it to 600B i hace compilation errors regarding memory limit.

Thank you,

Vera

Hi Vera, 

Please, find my comments. 

1. Did you make the changes for both root and end-node? If not, it should be. Can you clarify more on "didn't help"? Did you see the end-nodes join to the root but didn't see any improvement on joining time? Is this consistently happened? 

2. The TRICKLE_TIMEOUT_VALUE is another parameter that can impact on the joining time. It's the same story. With a smaller value, it would increase collision probability so it might not fasten the joining process in a large network. But in a small network it would be effective. Depending on your network size and topology.

3. Please, review TI-RTOS user guide for the API. The first parameter is for msg size. 

4. It's possible. I see a couple of issues for your changes though.  

4.1. The specific object (.data) with the increment falls into the CACHE_RAM section. The software example configured the 8KB CACHE for general-purpose RAM. You can find the memory allocation in the cc13x2lp.cmd file (in the .nwm_data section). The corresponding .data allocation with the size increment should be removed from .nwm_data section, which means that the .data allocation moves to 80KB SRAM area. To optimize the memory usage, I would recommend you to allocate different .data in objects that fit into the 8KB RAM. To figure out this, you can take a look at the .map file generated with your binary. 

4.2. To increase the UDP packet size, please increase the following buffer sizes for optimal performance. The following buffer sizes are related to the fragmentation and reassembly at 6LoWPAN layer. The fragmentation size is 128B. The 600B application data (+lower stacks headers) requires 5 fragments. If you consider to hold two application packets it requires 10. In the same way, the reassembly buffers hold receive packets. With a bigger size data, it will receive more numbers of packets in burst. 

Unless you manage this, some packet drops would happen. The first two are for fragmentation and the last one for reassembly. Please, set the same number for  QUEUEBUF_CONF_NUM and QUEUEBUF_REF_NUM. 

#define QUEUEBUF_CONF_NUM  6

#define QUEUEBUF_REF_NUM 6

#define SICSLOWPAN_CONF_REASS_BUF_SIZE 12

Hope that this answer your questions. 

Regards,

Wonsoo

Hi Wonsoo,

thank you for your quick replay.
what is the best was to achieve higher throughput from the project?
is it possible to change the PHY configuration?

p.s:

where can i find this "design guide" ?

thanks,
Vera