• State Resolved
  • Locked Locked
  • Replies 13 replies
  • Subscribers 57 subscribers
  • Views 2090 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Large Scale Zigbee Networks

Sam Elia
Prodigy 140 points
Other Parts Discussed in Thread: Z-STACK, CC2530

Hello,

I am trying to create a 250 node zigbee network (36 routers and 214 end devices).  Each node sends it sensor readings every minute (payload = 30 bytes).  Each router sends it Network Link Status every 15 seconds.  End Device Poll Rate messages have been disabled.

It seems that when my networks gets to the size of about 60 devices, I start seeing packet corruption.  From the packet sniffer, It looks like the traffic is very heavy to the point where it sometimes takes a node up to 5 minutes of continuously trying to send its packet before it is successful.

According to my calculations, in a 60 device network of 36 routers and 24 end devices:

Sensor Reading Packets: 60 

+

Acknowledgments:             60

+

Router NWK LinkStat:        36*4    (every 15 sec)

=  264 packets per minute

= 4.4 packets per second

Is 4.4 packets a second too much for Zigbee to handle?  I ran the PER sample program and sent 100 packets per second with 0% error.

I know there are other factors that go into packet corruption such as interference but I was hoping you could give me some advice specific to the Z-Stack that would help with implementing a 250 node network.  Is there a general rule of thumb for maximum packets per second?

ex. Change router NWK_LINK_STATUS PERIOD to 30 seconds to reduce traffic?


Also, I noticed from the packet sniffer that after leaving the 60 node network over the weekend, some nodes have switched PAN IDs.....  How is this possible when I specified the PAN ID in f8wConfig.cfg file.

  • 0
    Guru 31280 points

    Hi,

    I don't know yet what might be the solution in your case, however you have to answer several questions first:

    • What is the version of Z-stack?
       
    • What about security, is it enabled or not (my guess it's disabled)?
       
    • Have you implemented many-to-one routing, I believe that all these 60 devices sends
      data to some sort of sink (Coordinator)?
  • 0 in reply to Igor Sherer
    Prodigy 140 points

    Hi Igor,

    I am using the latest Z-stack 2.5.1a.

    Yes, security is disabled.

    The devices only need to send their data to the coordinator which serves as a gateway to a server.  They do not need to communicate with each other except to route packages.  The application is built off the GenericApp sample.  How can I implement many-to-one routing?

  • 0 in reply to Sam Elia
    Prodigy 140 points

    Sorry to waste your time, I should have just done a search before replying.  This is what I found you said in another forum:

    1. Schedule a Route discovery request (this is actually a "many to one" routing property of ZigBee - PRO),
      by calling NLME_RouteDiscoveryRequest( 0xFFFC, 0x03, 0 ); every 15 + jitter seconds .
    2. Add this line in project options (Alt+F7)->C\C++ compiler->Preprocessor (tab)->Defined symbols (sub window):
      CONCENTRATOR_ENABLE=true
    It looks like these settings are only for the coordinator?  I am using TI cc2530 routers and end device with a Digi x4 gateway.  Is there anything I can set on the cc2530 nodes to enable many-to-one routing?
  • 0 in reply to Sam Elia
    Guru 31280 points

    Sam Elia said:
    How can I implement many-to-one routing?

    In coordinator FW project add CONCENTRATOR_ENABLE=TRUE (preprocessor->defined symbols)

    Additionally, in your application, once device status changed to DEV_COORDINATOR,

    set a reload timer which invokes a MANY_TO_ONE event (it's your own event), let's say 10 seconds +

    some jitter. In this event call NLME_RouteDiscoveryRequest( 0xFFFC, 0x03, 0 );

  • 0 in reply to Sam Elia
    Guru 31280 points

    Nice, I've written this before. :)
     

    Sam Elia said:
     Is there anything I can set on the cc2530 nodes to enable many-to-one routing?

    Not sure if you can do something with cc2530 and many-to-one feature while the Digi x4 used as gateway.

    However, the "sink" can be every router in your network, it's not only for coordinators, therefore, if you have some

    router that can be hooked up to a PC, in other words, used by you as coordinator concentrator, you can use

    the same technique in this specific router.
     

    You know, coordinator is used only to start a network, after the network is up, coordinator behaves as other routers

  • 0 in reply to Igor Sherer
    Mastermind 9802 points

    Another trick is to be sure to stagger the transmissions in time. This way if all the devices are powered on at the same time (like after a power outage) then the network doesn't get jammed with everyone talking at once. I sometimes have devices delay their periodic reporting by a number of milliseconds equal to the last byte of the MAC address * 10 or something like that. You could also use a random number generator, but the MAC address way is simple.

  • 0 in reply to Igor Sherer
    Intellectual 465 points

    HI Igor:

    i  have some questions now,

    1、Do you mean that i have to set a reload timer  if i want to  use the many to one feature,like using a WatchDog? and when to set it,and if i can stop it?

    2、Are there some docoments introducing how to use the many to one feature? or what's best ,can you email some sammple codes to my mailbox? 371471521@qq.com

  • 0 in reply to Guanqiang Chen
    Guru 31280 points

    Guanqiang Chen said:
    1、Do you mean that i have to set a reload timer  if i want to  use the many to one feature,like using a WatchDog? and when to set it,and if i can stop it?

    U can use a reload timer (OSAL timer), which periodically sets an event which in turn invokes the many

    to one request (say every 10 seconds). You can stop timer run whenever you want, which will effectively

    stop the many to one requests.

    However if "you" know when a network reached it's final formation (no more router will be added from

    this point on), a gateway can initiate the many to one routing request only once, when the formation is

    done.

    Guanqiang Chen said:
    2、Are there some docoments introducing how to use the many to one feature? or what's best ,can you email some sammple codes to my mailbox? 371471521@qq.com

     

    I have mentioned in the other post where you can read regarding many-to-one property, however I don't

    know if there a document describing the exact how to.

    You can use the isntructions I've posted in this thread and the other thread (same in both threads), it should

    pretty straight forward. The extra effort from you would be defining a new event and add a call for reload timer.

    If you asks yourself at which point in the code you should start a new reload timer, well, it's hard to answer,

    you probably should analyze your application needs.

  • 0 in reply to Igor Sherer
    Intellectual 465 points

    ok,i see!thank you very much!

  • 0
    Genius 4525 points

    Your network size and traffic seems OK, just be careful for two issues based on my experience

    1. end device broadcast, that is too much traffic

    2. router discovery, some configure it as 30s, but if your routers are stationary, I guess you can change this into a much larger value

    Rui

  • 0 in reply to Rui Zhang69778
    Prodigy 140 points

    Thanks Rui,

    I faced the same thing.  Much better results when changing broadcast mode to 16 bit destination mode and destination to 0x0000 for coordinator.

    Sam

  • 0 in reply to Sam Elia
    Genius 4525 points

    Do you turn on APS level Ack?

    Rui

  • 0 in reply to Rui Zhang69778
    Prodigy 140 points

    Rui Zhang69778 said:

    Do you turn on APS level Ack?

    Rui

    No, I don't think so. Just the NWK level Ack.