How to achieve NRF52840 ESB Maximum data rate

[JamieHartshorne]

As part of a university project I am trying to read an analog signal and transmit it at the maximum data rate over ESB. But I cant seem to get above 1700kbps without any of my payload writes failing. I have disabled acking, I am using the highest bitrate, lowest CRC and fast ramp up. Are there specific timing settings that I am missing or does anyone else have any experience trying to achieve this feat. (at the moment my timing settings are linked to ADC acquisition rate and increasing the rate above a certain amount causes payload_write issues). Aplogies if this has been properly answered before but I have tried searching.I am using the NRF Connect SDK. I am looking to get as close to 2000kbps with minimal latency.

Included is my ESB config and conf settings.

config.protocol = ESB_PROTOCOL_ESB_DPL;
config.retransmit_delay = 10;
config.retransmit_count = 1; // Default retransmit count
config.tx_output_power = ESB_TX_POWER_8DBM;
config.bitrate = ESB_BITRATE_2MBPS; // Fixed bitrate
config.event_handler = event_handler;
config.mode = ESB_MODE_PTX;
config.selective_auto_ack = true;
config.payload_length = 252
config.crc = ESB_CRC_8BIT; // NRF_ESB_CRC_OFF
config.use_fast_ramp_up = true;

CONFIG_NCS_SAMPLES_DEFAULTS=y
CONFIG_ESB=y
CONFIG_DK_LIBRARY=y
CONFIG_CLOCK_CONTROL=y

CONFIG_ESB_NEVER_DISABLE_TX=n
CONFIG_ESB_RX_FIFO_SIZE=256
CONFIG_ESB_TX_FIFO_SIZE=256
CONFIG_ESB_MAX_PAYLOAD_LENGTH=252

# Enable logging subsystem
CONFIG_LOG=y


# Enable logging for the ESB module
CONFIG_ESB_LOG_LEVEL_DBG=y
# Enable heap memory pool for dynamic memory allocation
CONFIG_HEAP_MEM_POOL_SIZE=16384

CONFIG_NRFX_SAADC=y
CONFIG_NRFX_PPI=y
CONFIG_NRFX_TIMER0=y
CONFIG_NRFX_TIMER2=y

[Håkon Alseth]

Hi,

 

You will never reach the throughput equal to the on-air data rate, as there will be overhead in general (addressing, ramp-up etc).

When you setup your ESB configuration, this needs to match on both PRX and PTX.

Have you set the .noack member to the struct esb_payload variable that you're using? To enable noack, you need to set this in the tx_payload variable:

tx_payload.noack = true;
err = esb_write_payload(&tx_payload);
...

 

Kind regards,

Håkon

[JamieHartshorne]

Hi Hakon, thanks for your reply. I set .noack equal to 1 before each esb write and I have selective auto ack enabled on the PRX aswell, when i was troubleshooting the issue i tried using 32 byte packets and sending them faster, a friend has been able to achieve rates of 1650kbps using an NRF24L01 and an FPGA with 32 byte packets, but I could only send around 900kbps with that packet size so I'm wondering if the issues are with my configuration/code or if it is just the difference in overhead between the old and new devices. Please let me know if there is any information i can give you to help. 

Doing some quick maths, 252 payload + crc byte + 4 prefix/address bytes = 257bytes

at 2000kbps on air time = 1010.5us, with 40us ramp up, this becomes 1050.5us period or 0.952khz

0.952*252*8=1919.2 kbits per second theoretically, I am just trying to see where i could find the extra 200 or 100kbps to squeeze every last bit out of the protocol. Again I appreicate your help, I am an embedded novice so I could easily be doing something very simple wrong.

My code is based on the PRX and PTX examples and has addresses as follows so maybe I am just making the addresses too long?

	uint8_t base_addr_0[4] = {0xE7, 0xE7, 0xE7, 0xE7};
	uint8_t base_addr_1[4] = {0xC2, 0xC2, 0xC2, 0xC2};
	uint8_t addr_prefix[8] = {0xE7, 0xC2, 0xC3, 0xC4, 0xC5, 0xC6, 0xC7, 0xC8};

[Håkon Alseth]

Hi,

 

JamieHartshorne said:

a friend has been able to achieve rates of 1650kbps using an NRF24L01

Yes, but in this case you need to really really fine-tune the setup, by holding the nRF24 in TX for several milliseconds before the carrier drifts (open loop radio design in 24L-series). 

If you ramp up/down the radio with each TX, you will get approx. 130 is + time-on-air-for-32-byte-user-payload (4 us per byte => 160 us for 40 bytes), where 900 kBit then is expected.

 

Just to avoid any confusion around this subject:

In your current test-case, you need to use nRF52-series or newer devices on both PRX and PTX. With fast ramp-up, you are not backwards compatible with older nRF51- and nRF24-series devices.

JamieHartshorne said:

Doing some quick maths, 252 payload + crc byte + 4 prefix/address bytes = 257bytes

at 2000kbps on air time = 1010.5us, with 40us ramp up, this becomes 1050.5us period or 0.952khz

0.952*252*8=1919.2 kbits per second theoretically, I am just trying to see where i could find the extra 200 or 100kbps to squeeze every last bit out of the protocol. Again I appreicate your help, I am an embedded novice so I could easily be doing something very simple wrong.

My code is based on the PRX and PTX examples and has addresses as follows so maybe I am just making the addresses too long?

The full data format will look like this:

preamble 1 byte + RF address + PCF (packet control field) 11 bit + user_payload + crc.

That gives approx. 259 bytes, -> 1036 + 40 us -> 1076 us (0.92937 kHz).

.92937*252*8 = ~1873 kBit/s.

 

That is your theoretical max throughput, provided that your configuration works as you describe.

What is the range between your prx and ptx? This should be kept approx. 1 meter for testing purposes.

Have you checked the dynamic current consumption of your PTX device? This should show more-or-less "TX-current" all the time.

Kind regards,

Håkon

[JamieHartshorne]

Hi Hakon, I dont currently have a way to measure the dynamic current, using a regular multimeter and a chopped up usb c able with the XIAO nrf52840 ptx board, i get a current of 19mA at a DC voltage of 5.12V, but i will look into the oscilloscope method listed in the docs, at the moment the maximum data rate is 1727.7kbps when the DK and SEED XIAO nrf52840 are quite close together 5cm? At 1m the data rate drops below 1500kbps. This is still using a 4 byte address.

[Håkon Alseth]

Hi,

 

Remember that there is processing that is required here as well, in terms of executing ISRs/setting up timers, etc.

I setup a test, with a 1 second timer interrupt, where I count the amount of bytes sent.

I am seeing 219240 byte per second, which is 1753920 bit/s (1712 kBit/s).

 

A bit less than you report.

 

Kind regards,

Håkon

[JamieHartshorne]

Thanks Hakon, at least I have a sanity check on my results. I have ordered an NRF54L15 which I hope will give me faster rates.

[Håkon Alseth]

Hi,

 

Happy to help out!

nRF54L15 can do 4 MBit, so this should help wrt. the overall throughput you need.

 

Kind regards,

Håkon