在无线网络中,一个设备(如eNodeB)是按块(block)向另一个设备(如UE)发送数据的。发送端使用块中的数据计算出一个CRC,并随着该块一起发送到接收端。接收端根据收到的数据计算出一个CRC,并与接收到的CRC进行比较,如果二者相等,接收端就认为成功地收到了正确的数据,并向发送端回复一个“ACK”;如果二者不相等,接收端就认为收到了错误的数据,并向发送端回复一个“NACK”,以要求发送端重传该块。如果在某个特定的周期内,发送端没有收到接收端的回复,则发送端假定之前发送的块没有到达接收端,发送端自动重发该块。(MAC层的HARQ处理) BLER(block error rate),即误块率,是出错的块在所有发送的块中所占的百分比(只计算初传的block)。在实际应用中,某一特定百分比(如:LTE中数据信道的BLER要求为10%以下)的BLER并不总是必须的,因为可以重传出错的块并通过特殊的处理(如软合并等),使得接收端正确解出收到的数据。需要测量和计算BLER时,在发送端就能够完成,因为可以通过收到的NACK数来计算BLER。 在LTE中,控制信道的目标BLER为1%,数据信道的目标BLER位10%。当BLER不超过10%时,UE将向eNodeB上报它所能解码的最高MCS。 LTE在无HARQ重传情况下误块率指标为10%,加入HARQ重传后误帧率(FER)大概为1%,再加上RLC层的ARQ后性能提升到10^-5数量级。 例:假设发送了500个block的数据,其中499个block回复ACK,1个block回复NACK,则BLER为1 / 500 = 0.002 * 100% = 0.2%(从这个例子可以看出,计算BLER时,是不把重传的block的ACK/NACK计算在内的)。
参考:How to Calculate the Block Error Rate When devices transmit data to one another, for example, over a wireless network, they transmit it in blocks.The sending device uses the data in the block to calculate a cyclic redundancy check (CRC), and sends that with the block. The receiving device determines if the transmission contains errors by calculating the CRC of the data received and comparing it to the sender's CRC. If they agree, then the receiving device replies with "ACK," or acknowledgement. If they are different, then the device replys with "NAK," or negative acknowledgement, and the sending device resends the block. Some devices will resend the block automatically if the receiving device doesn't respond after a fixed period of time, on the assumption that the block never made it to the device. The block error rate, or BLER, is simply the ratio of blocks sent with transmission errors to total blocks sent. A 0 percent BLER is not always necessary or practical, due to the extra time it takes to resend blocks with errors. Mobile phone conversations, for example, can tolerate a BLER above 0 and still be understood by listeners. To measure and calculate the BLER, you only need access to the sending device, because you can calculate BLER based on the number of NAKs received.
|