nRF52382 SparkFun Breakout Bluetooth Transmission wrong

[danilbso]

Hello!

I'm having a problem and a hard time trying to understand what's wrong. Currently, I'm using the SDK 14.2.0, SD 5.0.0 and Segger on macOS 10.14.2.

I'm using the Bluetooth Low Energy example "Relay", from peripheral and central. 

I started developing the software on the Development Kit Board (PCA10040), and it's working fine. My application only needs to transmit advertisement packets, with modified manufacturer data (I'm using those bytes). Also, it scans for those packets, that's why I needed the relay example.

The thing is: the firmware is working fine on the Development Kit Board, modifying the data as I need it to, but when I try to program the sparkfun board, it seems to work as usual but some packets have wrong data, always on the same bytes.

It seems an memory issue, but I couldn't fine more resources to fix that.

Here's an example: This is a text file I made to log the packets. It was captured with a Ubertooth One.

The packet with ID 19 has the wrong company ID. 

id=18 systime=1548453092 freq=2402 rssi=-35 data=42 14 db f4 1c 3b 85 e7 02 01 04 0a ff 59 00 89 01 41 ea 26 88 01 b3 4f af 
id=19 systime=1548453092 freq=2402 rssi=-35 data=42 14 db f4 1c 3b 85 e7 02 01 04 0a ff 49 00 8f 01 41 ea 26 8f 01 f1 3c d7 
id=20 systime=1548453093 freq=2402 rssi=-41 data=42 14 db f4 1c 3b 85 e7 02 01 04 0a ff 59 00 91 01 41 ea 26

The address of my board is DB:F4:1C:3B:85:E7. This error also happens on the company ID byte, changing to 0xEF or 0xFB.

It happens on a considerable percentage of the packets, like 5%-10%.

Sometimes the error occurs on the first byte of this "data" I'm printing, changing from 0x42 to 0x62.

id=15 systime=1548453091 freq=2402 rssi=-37 data=42 14 db f4 1c 3b 85 e7 02 01 04 0a ff 59 00 81 01 41 ea 26 7b 01 05 53 62 
id=16 systime=1548453091 freq=2402 rssi=-41 data=62 14 db f4 1c 3b 85 e7 02 01 04 0a ff 59 00 85 01 41 ea 1e 83 01 f0 73 ab 
id=17 systime=1548453091 freq=2402 rssi=-41 data=42 14 db f4 1c 3b 85 e7 02 01 04 0a ff 59 00 87 01 41 ea 1e

This is what I'm printing for this data:

int lell_get_data(const lell_packet *pkt, uint8_t *data)
{
	int i, length;
	length = pkt->length - 6;
	for (i = 0; i < length; ++i)
		data[i] = pkt->symbols[12 + i];
	return length;
}

All configurations are the same as for the DK, and it's working perfectly on the DK.

Thanks in advance,

Danilo.

[run_ar]

How often are you updating the manufacturer specific data? Does this happen 5-10% of the time even if you do not update the data? Are you checking the crc on the sniffer? How many failing boards do you have and what is the failure rate? What sparkfun board are you using? and how are you powering this?

[danilbso]

1. How often are you updating the manufacturer specific data?

I'm updating the manufacturer specific data every 110ms, through this define statement:

#define APP_ADV_INTERVAL    MSEC_TO_UNITS(110, UNIT_0_625_MS)

2. Does this happen 5-10% of the time even if you do not update the data?

Yes, I just tested with a string of static data, like "1234567890" and the same effect happened.

3. Are you checking the crc on the sniffer?

Yes, I'm using the ubertooth library with CRC verify as 'true'. The adv packets of the Nordic DK doesn't show any disturbances.

4. How many failing boards do you have and what is the failure rate?

Actually I have only one board here to test. My application will envolve several boards, but if this error continues I'll need to look for another board.

5.  What sparkfun board are you using?

https://www.sparkfun.com/products/13990

6. and how are you powering this?

I offen power with the DK board itself, using the 5V pin (DK) to the Vin pin (SparkFun). I thought this might be the problem, but I tested with alkaline batteries on the 3V3 pin as well, but the problem persists. 

[run_ar]

What is the complete laser marking on top of the chip that is failing?

I assume you are going to sleep (calling power_manage in the main loop) in between sending the advertisements? Could you try to comment that part of the code so you do not go to sleep, and see if that helps?