POLQA - MOS fail 解题

VzW 认证实验室TM测试过程中报了耳机通路的e-MOS fail。

这东西，怎么都难fail吧。大概率得找设置的原因。

1. HIB settings & 2. source signal adj level

ACQUA供应商支招，这个设置更改对响度影响挺大的。

错误：MOS=3.5

改正：MOS=3.78

3. lab delay

理论上电信号测试就不应该有多少delay，这水货实验室干到了500ms+，对于POLQA v1.1 time shift 来说信号对齐异常是致命的。v2.4发布的白皮书里面讲了三点改善，其中就有这条。

-Length dependency of the results in narrowband (NB) mode if the signal duration exceeded approximately 10 s in narrowband mode

- Transparency problems: Too often a comparison of the reference signal with itself resulted in MOS scores below the optimum (known as “the transparency problem”)

-Shift performance: Small shifts of the starting point of the degraded signal may have had unexpectedly large impacts on the measured MOS.

咱讲究的就是有错就改，山不改我改。lab delay有问题，直接给POLQA analysis version上新版。

改完之后按VzW要求 run 20 iterations, average score = 4.18

结果是好的，过程中有几个值得记录的对比：

离谱到家，洗洗睡了。