Throughput ATT_MTU

[Ridhi]

We are using nrf52832 soc for our product development and using that we are trying to increase the throughput of our device using the information from the following forum:
Novel Bits throughput guide 
In this forum it states that the Maximum GATT MTU SIZE possible is 247 bytes in which 244 bytes of data is the actual payload size and if the data is greater than the limit, then it will split into multiple packets. So we tried to send more that 247 bytes of data just to check how the data split in multiple packet is working. We tried to sent 256 bytes of data and we were able to get 255 bytes of data without any splitting, which is more than 247 bytes. So we are wondering what is the Maximum GATT MTU SIZE possible in nrf52832 soc. 
We also looked into the throughput example provided by the nordic (ble_app_att_mtu) and in that MTU size is set to 247 and data length is set to 251 which is less that what we are able to send from nrf52832 to phone, is there any possible explanation for this?

[Simonr]

Hi

1. Maximum number of packets per connection interval = [connection interval / Data_Packet_Time] where [] rounds to the highest whole number. So for example [7.5*1,000 microsecs / 2,468 microsecs] = 3 packets.

2. I'm sorry, but I'm not sure I understand what you mean. Feel free to use this formula to calculate throughput on your own.

3. Indeed, the 2MBPS PHY should result in ~2x the throughput, do you enable the 2MBPS in your application code? If so, and if both devices in the connection supports 2MBPS you should see an increase in throughput.

Best regards,

Simon

[Ridhi]

2). i have used :

27 bytes of ATT MTU 

158 bytes of ATT MTU 

247 bytes of ATT MTU 

my data length is 12 bytes and i took 2000 values of 1bytes each after 2 secs interval

Then what will be the throughput considering ATT_MTU , Connection Interval and number of packets per connection interval

3). how to enable 2M phy in connected mode in peripheral side?

[Simonr]

Hi

2. I'm not sure I understand what you're asking here. If the ATT MTU changes but the data length stays the same you won't increase the throughput much I'm afraid, as the data length should also be increased according to what you increase the ATT MTU with. The 2000 bytes over 2 seconds would equate to ~1 kB per second (8 kbits per second).

3. Once connected, the central will be in charge of deciding the PHY that is used, but you can set the preferred PHY in the BLE_GAP_EVT_PHY_UPDATE_REQUEST event to BLE_GAP_PHY_2MBPS instead of BLE_GAP_PHY_AUTO as .rx_phys and .tx_phys like below (this must be added to/edited in the ble_evt_handler() function of your main.c file:

        case BLE_GAP_EVT_PHY_UPDATE_REQUEST:
        {
            NRF_LOG_DEBUG("PHY update request.");
            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);
        } break;

Then you need to initiate a PHY update request from the central, and as long as the central supports the 2MBPS PHY, the connection should move to the 2MBPS PHY.

Best regards,

Simon

[Ridhi]

Simonr,

I have tested BLE 5.0 LE 1M phy and BLE LE 2M Phy

i took 10,000 logs of 12 bytes data at 2 secs interval.

I am able to receive the logs

#define MIN_CONN_INTERVAL MSEC_TO_UNITS(1000, UNIT_1_25_MS) /*

#define MAX_CONN_INTERVAL MSEC_TO_UNITS(1000, UNIT_1_25_MS) /*

NRF_SDH_BLE_GAP_DATA_LENGTH 31/162/251

NRF_SDH_BLE_GATT_MAX_MTU_SIZE 27/158/247

WITH these combinations on sniffer i found:

My issue is that why in LE 2M same data is taking more time compared to LE 1M?

[Simonr]

Hi

That sounds very strange. What I'm most skeptical to is the fact that the 1MBPS PHY with 247 byte MTU size lets you do 9 packets per conn. interval, and this is halved for the 2MBPS version of the transmission. At what range are you testing this throughput?

Best regards,

Simon