How much can we get the error voltage at following conditions?
REF input (3pin) : 1000MHz CW
VCO input (6pin) : 1000MHz plus (10us)
In this case, can we get 0Vdc under the phase differemce is 0 rad, -0.4Vdc under the phase differemce is -p/2 rad and +0.4Vdc under the phase differemce is p/2 rad as the error voltage in the term of 10us plus?
If I understand you're question correctly you want to know if you can achieve the stated error voltages within 10uS, right?
The 10uS would be part of the 'settling' time of the synthesizer which is ultimately limited by the modulation BW of the VCO and the loop filter bandwidth. With wide loop bandwidths it's possible to get settling times close to this so this seems feasible. This wasn't characterized Additionally, the fact that the HMC439 supports comparison frequencies up to 1300MHz seems to support this as well.
The voltages you list are slightly different than what is shown in the datasheet so I'm assuming these are approximations. Yes, this are the error voltages that would be measured. I would caution you that this part should NOT be used as a 'Phase Detector'. As a 'Phase Frequency Detector' it was designed specifically for closed loop operation within frequency synthesizer applications.
Please refer to the following document for additional details:
FAQ: HMC Phase Frequency Detectors
Is your error voltage the output of the measured filter? Or what part of the voltage is the error voltage?