Connection parameter optimize: nRF52840(central) multilink with 8x52832 peripherals

[Avadacadabara]

Hi all,

I'm developing a multilink system with 52840 dongle as central and 8 52832 devices as peripherals, which is similar to previous project in thread part I & Part II.

8 peripherals are exactly the same and generates 172Bytes/each device every 100ms, so total data rate would be 13.76kB/s. 

↑According to power profiler, I foud that data packets generated every 100ms, which is 8(peripheral devices) * 251(PDU size) * 1(packet per device) could be transmited within 25ms.  So I set central connection parameters as follows：

But here is my problem: the throughput could reach the requirement, but the connection stability seems very sensitive to distance. When peripheral devices are placed closely to central, the connection is very stable. Then I move peripherals 1 meter away from the central, the devices start to disconnect & reconnect frequently which cause severe packet loss.

I've tried different parameters in previous thread, for instance, 100ms connection interval or 80ms interval. I also tried 52840 module with PA/LNA to raise TX/RX sensitivity. Seems no help. So I think there might be some faults in parameter setting?

Thnaks.

Ava

[Sigurd]

Hi!

1) Are you testing this with custom board or DK ? If you are using custom board, do you see the same behavior with nRF52840-DK?

2) What is your NRF_SDH_BLE_GAP_EVENT_LENGTH value in sdk_config.h?

3) 

stability seems very sensitive to distance.

Does this happen when there are only 1 device connected? or do you need to connect all 8 peripherals for this to happen?

[Avadacadabara]

Hi，

Yes I tried the same setting on 52840 DK, 52840 dongle and custome board.

Sigurd said:

1) Are you testing this with custom board or DK ? If you are using custom board, do you see the same behavior with nRF52840-DK?

Current GAP_EVENT_LENGTH is 80(this is for 80ms connection interval condition I previous set, and remain untouched when change to 30ms interval). 

Sigurd said:

2) What is your NRF_SDH_BLE_GAP_EVENT_LENGTH value in sdk_config.h?

If TOTAL_LINK_CNT is set to 1 then it won't happen. This only happens in multilink applications.

Sigurd said:

Does this happen when there are only 1 device connected? or do you need to connect all 8 peripherals for this to happen?

[Sigurd]

Hi!

Do you have NRF_SDH_BLE_GAP_EVENT_LENGTH set to 80 on the central as well ?

If you have a connection interval of 30ms, and 8 links, try setting it to 30/8= 3.75 ms. It's in 1.25 ms units, so 3,75/1,25=3

so try to set NRF_SDH_BLE_GAP_EVENT_LENGTH  to 3 on the central's sdk_config.h, and see if the behavior changes.

See this link for more info: https://infocenter.nordicsemi.com/topic/sds_s140/SDS/s1xx/multilink_scheduling/suggested_intervals_windows_s132.html

[Avadacadabara]

Peripherals are set to 6(7.5ms) and Central is set to 80.

Sigurd said:

Do you have NRF_SDH_BLE_GAP_EVENT_LENGTH set to 80 on the central as well ?

So you were saying I should set Central NRF_SDH_BLE_GAP_EVENT_LENGTH to 3. Then should I remain Peripherals to 6 or change them to 24(which equals to 30ms)?

[Sigurd]

Avadacadabara said:

Then should I remain Peripherals to 6 or change them to 24(which equals to 30ms)?

If the peripherals only connect to 1 central, then you could increase it to 24.

PS: Connection Event Length Extension could also be helpful for the throughput:

https://infocenter.nordicsemi.com/topic/sds_s140/SDS/s1xx/multilink_scheduling/extend_connection_event.html

    ble_opt_t  opt;

    memset(&opt, 0x00, sizeof(opt));
    opt.common_opt.conn_evt_ext.enable = 1;

    err_code = sd_ble_opt_set(BLE_COMMON_OPT_CONN_EVT_EXT, &opt);
    APP_ERROR_CHECK(err_code);

[Avadacadabara]

Hi,

I tried to set Central GAP_EVENT_LENGTH to 3 but seems like the disconnection is getting worse. Then I set both Central and Peripheral GAP_EVENT_LENGTH to 24 and it could hold connection quite a long time in range of 2 meters. But when I try to move further (3m), they could still hold for several minutes. Then first come the throughput unstablization, followed with massive disconnection and reconnection. 

 

I've already turn on DLE for all Central and peripherals.

Sigurd said:

PS: Connection Event Length Extension could also be helpful for the throughput:

[Avadacadabara]

Plus following snippet is current scan parameters on my Central, is there any conflicts between connection event and scanning event?

[Sigurd]

With NRF_SDH_BLE_GAP_EVENT_LENGTH=3 (in 1.25ms units), you have 8*3,75 = 30ms, then you have the scanner on top of that.

NRF_BLE_SCAN_SCAN_WINDOW = 16 (in 0.625 ms units) -> 10ms , with tScanReserved ~1ms, you should have a connection interval of at least 41ms when you have 8 connections running + the scanner running.