Reed Solomon error

I am currently using DW1000 for design, but during testing, I found that when dw1000 communicates with dw1000, packet loss will occur, and the probability of packet loss will be relatively large, about 0.5%.The log of 5000 sending and receiving packets is shown as:
PHE err: 63
RSL err: 21
CRCG err: 127
SFDTO err: 113
The actual number of lost packets is 24,This is like packet loss caused by RSL err.
Due to the need to increase the number of labels, this will have a serious impact.
The configuration used is:preamble 128,PAC 8 ,Data rate 6.8M,channel 2
May anyone help me ?

Hi lwl9855000,

Under which conditions are you performing those measurement ?

What is the distance, environment between the devices?


Thanks in advance for your reply
Our test environment is tested in a shielded room. The environment rack is set to use two PCBs, and use high-frequency cable cables, use an adjustable attenuator between the two PCBs, then use one of them to send 5000 packets, and then use the other board to receive packets. Then when adjusting the adjustable attenuator 20-70db, packet loss always exists.
Thank you

Ok so you never reach an attenuation for which packet loss is null and CRCG is 100% ? This is not expected indeed.

Are you using a custom design ? What dcdc do you use ?

Thank you,

Yes, it is custom design. Then DC to DC uses SYV887B, both 1.8V and 3.3V. I have also verified testing with a dc source, but the behavior is consistent.

I am not sure I understand your initial graph.

Please note that CRCG is actually successful reception. If you send 5000 frame, at the sensitivity level, you should have 99% of CRCG. So you should basically tweak the attenuation until you have 99% CRCG and this gives you the sensitivity of the device.

Sorry, this is a mistake I organized.

This is my refreshed chart, one column of CRCG is using the secondary axis
It should be noted that: in fact, when we output the log, we will not display any log when we correctly accept a full frame of signals. Only if an error occurs, such as PHE err, RSL err, etc., then we will first display an RX err (“RX err all” in the chart is the sum of all errors), and then the value of the internal counter will be displayed. So if there is a dot in the chart, then it is err once. Of course there are multiple err happening at the same time.

Thanks again for your reply.

This is 5000 packets. The last error log occurred during the 4945th frame.

The number of items is as follows:
PHE = 62 (1.24%)
RSL = 15 (0.3%)
CRCG = 4921 (98.42%)
CRCB = 1 (0.02%)
SFDTO = 112 (2.24%)

The actual actual loss of 5,000 frames is 24 (frames judged unusable by dw1000)
Rx_Real loss = 24 (0.48%)