To check the validation of current PEM injection method, I analyzed the acoustic component in REFL table accelerometer signal.

In this result (plot), it is over estimated.

In my hyposesic, it's coming from the direct vibration from the speaker.

The microphone detects the enviromental sound (S_env) and the injected sound (S_inj).

The accelerometer detects the enviromental vibration (V_env) , the vibration from speaker (V_inj), and the total sound with coupling function C_A.

From injected data and background data for MIC and ACC, the coupling function C_A is caliculated as

.

If we ignore V_inj, the coupling function and acoustic component are overestimated as

The same situation occurs in other channels.

So it is difficult to evaluate any valid environmental noise budget from the acoustic injection by large speaker, and we need to do local shaking test.

However, anyway, the fact that the area around REFLEX is delicate against environmental noise (sound and seismic) is true.

c.f.)

I  tried to swap the main channel and sub channel  from (table ACC, booth MIC) to (booth MIC, table ACC).

The result is here. Overestimation is not occures. From this plot, the vibration of REFL table may be direct one, rather than acoustic coupling.