In order to improve the speed of the OMC shutdown to prevent the PD damage, here is a proposal of how to modify the OMC PZT driver.

General idea

When there is a lock loss of the main interferometer, we want to quickly change the length of the OMC so that it is kicked out of the locked state, preventing a large amount of energy from falling onto the DCPDs. This rapid OMC shutdown (ROS) should happen faster than 1msec.
Up to now, KAGRA used the HV output of the OMC PZT driver for this purpose. However, due to the LPF in the HV path, this method is too slow for ROS.

It turned out that the original scheme for ROS incorporated in the design of the PZT driver by LIGO uses the LV output.
LV output is usually used to dither the OMC length for OMC LSC. It also has 10V DC offset to set the working point of the dither.
When we want to perform the ROS, this 10V offset is rapidly brought down to 0V. This gives about 0.15um of PZT motion, enough to kick the OMC out of resonance.

How does it work

Page 6 of the attached PDF shows the LV circuit of the OMC PZT driver.

The output node "lv_omc" is driven by two lines. 

One is the DC offset provided by U5 LT1021U(Voltage reference IC). This voltage is buffered by U1 then goes through a voltage divider (R30/R34).
Then there is a 10Hz active LPF and two passive RC LPFs. Output of the filter chain is connected to lv_omc.

Another signal comes from the DAC for dithering. It is connected to the lv_omc through a coupling capacitor C30.

For ROS, lv_omc is also connected to the source of a MOSFET (Q2). The drain of Q2 is connected to the ground.
In the normal operation, the MOSFET is OFF, i.e. the source and drain are disconnected. Therefore, lv_omc is not connected to the ground.
When there is a trigger input, the MOSFET is turned on. In this state, lv_omc is connected to the ground through Q2. So the LV output is suddenly brought down to 0V from 10V.

How can we use this functionality?

We need to send a trigger singal to the trigger input of the PZT driver board, which is a BNC connector on the front panel.
LIGO uses an analog circuit to monitor the AS port power and send a trigger signal to the PZT driver for ROS.
https://dcc.ligo.org/E1200359

While KAGRA may want to implement this in the future, we currently don't have this circuit.
For the moment, I propose to use the DGS Binary Output to send triggers to the PZT driver.
To do so, we need to use the BIO converter to convert a binary output channel to a BNC.
Since the binary output of DGS is an open collector, we need to pull up the output. We implement 10kOhm to R13, which is omitted now.
We also need to remove R12. With this connection, making this bit 1 will trigger the ROS.

We also want to fine adjust the DC offset voltage so that after ROS, the OMC does not accidentally jump to the resonance of a higher order mode.
This adjustment can be done with R30 and R34.

Adjustment of dither gain

The dither signal is added to the LV output with the attenuation factor of C30/C25.
In the schematic, C30=100pF and C25 is 1uF. So the attenuation factor is 1/10000.
However, the actual implementation of the circuit is now C30=1000pF, so the attenuation factor is 1/1000.
This change was done to increase the OMC LSC signal so that we can lock it even in the air.
However, currently the dither amplitude is 3000cnts. 
We can increase this amplitude to 30000 and change C30 to 100pF, to achieve the same dither amplitude, while reducing the op-amp and DAC noise coupling by a factor of 10.
If 30000cnt is too large, C30 could be 220pF.
We may want to make this change along with the changes proposed above.

Summary

I propose to make the following changes to the OMC PZT driver board:

• R13 -> 10kOhm
• R12 -> remove
• R30 and R34: adjust appropriately
• C30 -> 100pF or 220pF