PHY1M is slower than PHY2M while forwarding data.

[TGuanLa]

Hello All,

I have a problem when I build my project from sample central_uart &  peripheral_uart.

I have two nrf5340 called A and B. B scans the broadcast to A to establish a connection. Then my phone(C) connects B.

A sends data to B, and B forwards data to C.

Speed of retransmission  is 160kbps when B works on PHY1M. However, speed drops to 40kbps when B works on PHY2M. 

I wonder what cause this problem.

Is there any ways to increase forwarding data speed？

Thank you.

static uint8_t ble_data_received(struct bt_nus_client *nus,
						const uint8_t *data, uint16_t len)
{
	ARG_UNUSED(nus);
	int err;

	err = bt_nus_send(peripheral_conn, data, len);
		// printk("len:%d\n",len);
		if (err) {
			printk("Send Error(err %d)\n", err);
		}
	
	return BT_GATT_ITER_CONTINUE;
}

B is connected with A as the center and C as the periphery.

[Einar Thorsrud]

Hi,

The theoretical maximum throughput is doubled when going from 1 Mbps to 2 Mbps PHY. However there are many factors that affect throughput. What are the connection parameters, what about the event length? Is the device (A) able to provide data quickly enough to fill the connection event with more packets? It could be interesting to know which parameters you use, and see a sniffer trace of both the cases (1 and 2 Mbps) to see what the differences are on air.

[TGuanLa]

Hi，

The parameters I use are below.

CONFIG_BT_CTLR_DATA_LENGTH_MAX=251

CONFIG_BT_BUF_ACL_TX_SIZE=256

CONFIG_BT_BUF_ACL_RX_SIZE=512

CONFIG_BT_CTLR_PHY_2M=y

CONFIG_BT_CTLR_RX_BUFFERS=2

CONFIG_BT_CTLR_SDC_MAX_CONN_EVENT_LEN_DEFAULT=50000

I use the default parameters for the rest.

Moreover, the speed of data could reach 1.2Mbps when A only connects B without advertising. 

Thanks.

[TGuanLa]

In 2Mbps PHY mode, I use this function of B.

static void data_len_updated(struct bt_conn *conn, struct bt_conn_le_data_len_info *info)
{
	int err;

	LOG_INF("Data length updated. RX length: %d, TX length: %d", info->rx_max_len,
		info->tx_max_len);

	err = bt_conn_le_phy_update(conn, BT_CONN_LE_PHY_PARAM_2M);
	if (err) {
		LOG_ERR("bt_conn_le_phy_update() returned %d", err);
	}
}

static void phy_updated(struct bt_conn *conn, struct bt_conn_le_phy_info *param)
{
	LOG_INF("PHY updated. RX PHY: %d, TX PHY: %d", param->rx_phy, param->tx_phy);
}

static struct bt_conn_cb conn_callbacks = {
	.connected = connected,
	.disconnected = disconnected,
#if CONFIG_BT_USER_DATA_LEN_UPDATE
	.le_data_len_updated = data_len_updated,
#endif
	.le_phy_updated = phy_updated,
};

In 1Mbps PHY momde, I don't use data_len_updated().

[Einar Thorsrud]

Hi,

TGuanLa said:

Moreover, the speed of data could reach 1.2Mbps when A only connects B without advertising. 

It is expected that advertising is relevant for throughput, as we discussed in this other thread.

TGuanLa said:

In 1Mbps PHY momde, I don't use data_len_updated().

I see. That should not matter. (In the 2 Mbps case you initiate a PHY update after a DLE for some reason, I do not see a problem with that though I do not immediately find it logical either, as these are independent things).

I do not see here why you get lower throughput. As mentioned, a sniffer trace of both the 1 Mbps and 2 Mbps case would be a good place to start. Then we can compare and see if anything sticks out (event length, interval, packet lengths, perhaps MD field is not set indicating not enough data queued fast enough, etc). Note that you do not need a expensive sniffer in order to collect sniffer traces. As long as you have an additional DK or dongle you can use nRF Sniffer (see the documentation for details on installing and using).

[TGuanLa]

Thanks for your answer.

Is it possible for B and A to work at 2Mbps in PHY, and for B and mobile phones to work at 1Mbps in PHY?

Must B only work in one PHY mode?