nRF52840 BLE Transfer speed issues

[masterLee]

Hi all,

I have a custom nRF52840 board that transmits 32-byte packets to mobile device and I want to achieve maximum transfer speed.

The code is based on the ble_app_uart project

The BLE settings in my project are as follow:

//sdk_config
NRF_SDH_BLE_GAP_DATA_LENGTH 251
NRF_SDH_BLE_GATT_MAX_MTU_SIZE 247
NRF_SDH_BLE_GAP_EVENT_LENGTH 6

//main
MIN_CONN_INTERVAL   MSEC_TO_UNITS(8, UNIT_1_25_MS)
MAX_CONN_INTERVAL   MSEC_TO_UNITS(8, UNIT_1_25_MS)
 
ble_stack_init();
gap_params_init();
gatt_init();
services_init();
advertising_init();
sd_ble_gap_tx_power_set(BLE_GAP_TX_POWER_ROLE_ADV,m_advertising.adv_handle,8); // Max  ADV Power = 8dB
sd_ble_gap_tx_power_set(BLE_GAP_TX_POWER_ROLE_CONN,m_advertising.adv_handle,8); // Max CONN Power = 8dB

The transmission data rate from the custom device to the mobile is 8kB/sec +/- ~400B

I want to reach maximum transfer speed but I don't know what the upper limit is.

Based on table 2 of the following link, I see that the max transfer speed is ~165kB/sec. So I try to figure out what I am doing wrong and my device's transfer speed is so low.

(https://infocenter.nordicsemi.com/index.jsp?topic=%2Fsds_s140%2FSDS%2Fs1xx%2Fble_data_throughput%2Fble_data_throughput.html)

From the debug session I get the following Information where the connection is established with DLE (247Byte / BLE_packet)

So I assume that something else keeps the transmission speed from my board to the mobile so low.

Also, I have a few questions :

1. How can I increase NRF_SDH_BLE_GATT_MAX_MTU_SIZE from 247 to 512? Because it is a 1-byte value and 512 needs 2-bytes  will need to mess with the SD libraries
2. Is it possible to increase the NRF_SDH_BLE_GAP_EVENT_LENGTH  higher than 6? Because when I set it to 7 or 8 the whole system stops.

If there is something else that I have not taken into account please correct me.

Thank you for your time

Regards

[Sigurd]

Hi,

1. How can I increase NRF_SDH_BLE_GATT_MAX_MTU_SIZE from 247 to 512? Because it is a 1-byte value and 512 needs 2-bytes  will need to mess with the SD libraries

 It should be enough to just set it to 512. But, 247 will give you the max throughput. By setting NRF_SDH_BLE_GAP_DATA_LENGTH = 251 (this is max for this setting), and NRF_SDH_BLE_GATT_MAX_MTU_SIZE = 247, we avoid fragmentation of the ATT packet into several on-air data packets.

2. Is it possible to increase the NRF_SDH_BLE_GAP_EVENT_LENGTH  higher than 6? Because when I set it to 7 or 8 the whole system stops.

 Yes, but you would likely need to allocate more RAM to the SoftDevice. How this is done depends on what toolchain/IDE you are using.

Also, as long as the phone does not limit the amount of BLE packets per connection interval, you might be able to get slightly higher throughput with a higher connection interval.

Also, I recommend enabling connection event length extension. You can enable Connection Event Length Extension with a function like this:

void conn_evt_len_ext_set(bool status)
{
    ret_code_t err_code;
    ble_opt_t  opt;

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

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

}

Then call conn_evt_len_ext_set(true) after you have enabled the SoftDevice.

If you still are having issues with throughput, please do a sniffer trace with nRF Sniffer.

[masterLee]

Hi, I will try what you suggest.

Now, regarding the max packets per connection interval,I read in post that the S140 has a maximum of 6 packets.

Does it mean that  6 packets are transferred to one direction such as from the peripheral (nRF52840) to mobile or, 3 packets from each direction meaning three packets from the peripheral side and three packets from the central side?

I am confused with the concept of data packets because if I get it right the data packets for one connection interval increase if I increase the connection interval. But from comparison tables and the fact that android devices can handle up to 6 packets per interval I understand that this is not true.

Also I haven't found some definition in the softdevice documentation that specifies the exact number of maximum packets that it can handle.

[masterLee]

Yes I am using SES but now I am not at the office. I will do it first thing tomorrow.

But even if I increase the GAP_EVENT_LENGTH the one packet per connection interval will still remain the same. This does not change the number of packets in one connection interval. Correct?

[Sigurd]

Correct, not directly at least, this config decides the time set aside for the connection on every connection interval in 1.25 ms units. Most likely you would need to do a sniffer trace to see what's happening on-air, and see what is limiting the throughput. https://www.nordicsemi.com/Software-and-tools/Development-Tools/nRF-Sniffer-for-Bluetooth-LE

[masterLee]

The tablet has the latest android version and Android supports up to 6 packets per connection interval.

Also, before and after calling the ble_nus_data_send() I turned on and off  an output pin and I measured the time with the oscilloscope for the function to complete and it matches exactly the connection interval I have set. This means that even if I call the ble_nus_data_send() 6 times for example, it will take 10msec for each one. This means that I send one 244-bytes packet every 10msec and not six 244-bytes packets in 10msec (one connection interval).That is why the data rate stays the same. 

What I want is to transfer more than 244 bytes in one ble_nus_data_send() and this cannot be done as long as the function cannot accept more than 244 bytes in one call. Is there some other function that can send more than 244 bytes in one call ? Or can I modify the existing function to do that?

[masterLee]

I know what limits the throughput. Check my last message explaining the steps I take to check it. I will test the sniffer but I dont think it will show something different. And by the way , since the max size for one packet is 244 bytes why the ble_nus_data send () size argument is of type uint16_t and not uint8_t?

[masterLee]

This is the nRF sniffer output while peripheral device was streaming to the tablet device

(The Connection Interval is set to 10msec and DLE is ON). The 244-bytes packet is transfered in 150usec and then the connection is idle for the remaining part of the interval...

So the issue remains how do I include more BLE packets in one connection interval instead of one(current)?

What am I doing wrong and the speed is

[Sigurd]

Could you save and upload the log here?

[masterLee]

0488.log.pcapng

[Sigurd]

The slave sets the MD(MoreData) bit to 1 when it sends the first packet, so it has more data to send. The master continues the connection event by sending a new empty packet, but the slave is not sending the packet in reply to that, and the connection event ends. In a 1-to-1 connection setup, I would have expected the slave to send another packet.

Is the peripheral connected to several master devices when you tested this ?

[masterLee]

Hi,

No the peripheral is connected to a single central device (tablet). The tablet app is configured such a way that it can connect up to 5 peripherals. The good thing is that the speed is increasedto 50kBps by changing  the GAP_EVENT_LENGTH from 5 to 10. Further increments did not increase the speed. 

Also, I noticed that although I set the PHY to 2M (inside ble_evt_handler) the tablet keeps it at 1M all the time.

Any suggestions?

[Sigurd]

Regarding 2MPHY, did you e.g. add something like this to the BLE_GAP_EVT_CONNECTED event ?

            ble_gap_phys_t const phys =
            {
                .rx_phys = BLE_GAP_PHY_2MBPS,
                .tx_phys = BLE_GAP_PHY_2MBPS,
            };
            err_code = sd_ble_gap_phy_update(p_ble_evt->evt.gap_evt.conn_handle, &phys);
            APP_ERROR_CHECK(err_code);

If yes, then you should be able to see a LL_PHY_REQ packet from the slave, and a LL_PHY_UPDATE_IND packet sent from the master in the sniffer log, indicating if it accepts the 2MPHY request or not.